Math Practice

An absolute-value equation is a distance problem — $|x-2|=5$ asks 'which $x$ is distance 5 from 2?' — two answers.

$|x-a|<r$ means stay inside a radius; $|x-a|>r$ means outside it.

$2+3=5$ is concrete. $a+b=b+a$ is abstract. 'Groups have associativity' is more abstract.

Just like $\frac{1}{3} + \frac{1}{4}$ requires a common denominator of 12, adding $\frac{2}{x+1} + \frac{3}{x-2}$ requires the LCD $(x+1)(x-2)$. Rewrite each fraction so both have the same denominator, then add the numerators. The process mirrors numeric fractions but with polynomial denominators.

Just as English has grammar, algebra has rules for combining symbols meaningfully.

Beyond numbers: what happens when ANY set has operations with certain properties?

$x^2 + y^2 = 1$ constrains $(x, y)$ to lie on a circle — not all points in the plane are allowed.

Identities are always-true shortcuts — no matter what values you substitute, both sides will always be equal.

The degree of a polynomial doesn't change when you multiply it by a non-zero constant.

It is like rearranging a sentence without changing its meaning.

$a^2 - b^2$ always factors to $(a+b)(a-b)$ — recognize the pattern once and apply it everywhere.

Translating 'the cost is \$5 plus \$2 per item' into $C = 5 + 2n$.

$x^2 + y^2$ is symmetric: swapping $x$ and $y$ gives the same expression.

Each term of $(a+b)^n$ picks '$a$' or '$b$' from each factor. $\binom{n}{k}$ counts how many ways to pick $k$ $b$'s.

Treat your answer as a hypothesis and test it by substituting back into the original equation to verify.

Imagine you have $x^2 + 6x$ and want a perfect square. A perfect square like $(x + 3)^2 = x^2 + 6x + 9$ needs that extra $+9$. So you add 9 and subtract 9 to keep the expression equal—then group the perfect square part.

The constraints don't contradict each other—there's some answer that works.

$\pi \approx 3.14159$ is always the same (constant). $x$ can be anything (variable).

Multiple conditions at once: '$x > 0$ AND $x + y = 10$ AND $y \leq 6$.'

$x + y = 5$ AND $x + y = 7$ can't both be true simultaneously — this is a contradiction.

Like a map with street numbers—the address $(3, 2)$ is 3 right, 2 up.

Place two arrows flat on a table. The cross product points straight up from the table, perpendicular to both. Its length tells you how much area the two arrows span—like the area of a parallelogram with the arrows as sides. If the arrows are parallel, they span no area, so the cross product is the zero vector.

If $x + y = 10$, you can choose $x$ freely, but then $y$ is fixed. One degree of freedom.

You choose the input (independent), and the function gives the output (dependent).

The determinant measures how a matrix scales area (in 2D) or volume (in 3D). If $\det(A) = 3$, the transformation described by $A$ triples all areas. If $\det(A) = 0$, the transformation collapses space into a lower dimension (like squishing a plane into a line), which is why the matrix has no inverse.

You can't add meters to seconds — dimensionally inconsistent equations don't make physical sense.

The discriminant is the expression under the square root in the quadratic formula. If it is positive, you can take the square root and get two answers. If it is zero, the square root is zero so both answers are the same. If it is negative, you cannot take a real square root, so there are no real solutions.

The dot product measures how much two vectors point in the same direction. If they point the same way, the dot product is large and positive. If perpendicular, it is zero. If they point in opposite directions, it is negative. Think of it as a 'similarity score' for directions.

A balanced scale: both sides must weigh the same. Solve by keeping balance.

Whatever you do to one side, do to the other — the balance stays true.

Plug in the number and compute: if $x = 3$, then $2x + 1 = 2(3) + 1 = 7$.

Open up the parentheses: $(x + 2)(x + 3)$ becomes $x^2 + 3x + 2x + 6 = x^2 + 5x + 6$.

Combine like terms, reduce fractions, apply identities to clean up expressions.

A recipe for calculating a value: '$2x + 3$' tells you to double $x$ and add 3.

Reverse distribution: instead of expanding $(x+2)(x+3)$, you compress $x^2 + 5x + 6$ into the same product.

Imagine four terms that seem unrelated. By cleverly grouping them into two pairs and factoring each pair separately, a common binomial factor often emerges—like finding a hidden pattern by rearranging puzzle pieces.

When you multiply $(a + b)(a - b)$, the middle terms cancel: $a^2 - ab + ab - b^2 = a^2 - b^2$. So any time you see a perfect square minus a perfect square, you can instantly factor it. Think of it as a rectangle whose area is the difference of two square areas.

Reverse engineering multiplication: 'What times what gives $x^2 + 5x + 6$?'

Look at what all terms share in common—like taking the common ingredient out of a recipe. In $6x^3 + 9x^2$, every term has at least $3x^2$ in it, so pull it out front: $3x^2(2x + 3)$.

You are reverse-engineering FOIL. If $(x + p)(x + q) = x^2 + (p+q)x + pq$, then you need two numbers $p$ and $q$ whose sum is $b$ and whose product is $c$ (when $a = 1$). When $a \neq 1$, use the AC method: find two numbers that multiply to $ac$ and add to $b$, then split the middle term and factor by grouping.

A function is like a machine: put a number in, get a number out. The table records what goes in and comes out, and the graph draws the picture.

Temperature determines ice cream sales—sales DEPEND ON temperature.

Use boundary lines and shading to show where conditions are true.

A parabola is a U-shaped curve (or upside-down U). Start by finding the vertex—that is the turning point. Then the axis of symmetry tells you the curve is a mirror image on both sides. Plot a few symmetric points and connect them in a smooth curve.

$a + a = 2a$ is always true (identity). $x + 3 = 7$ is only true when $x = 4$ (equation).

Instead of 'equals exactly,' it's 'at least,' 'at most,' or 'greater/less than.'

Parentheses mean the endpoint is NOT included; square brackets mean it IS included. For example, $(2, 5]$ means $2 < x \le 5$.

If matrix $A$ represents a transformation (like rotating 30 degrees), then $A^{-1}$ undoes that transformation (rotating $-30$ degrees). Multiplying by the inverse is the matrix equivalent of dividing. Just as $5 \times \frac{1}{5} = 1$, we have $A \cdot A^{-1} = I$.

Peel away everything around $x$ until only $x$ remains: $x =$ answer.

Every step right changes $y$ by the same amount—like climbing stairs at a constant pace.

You search the corners of an allowed region for the best score.

Two lines can cross (one solution), be parallel (no solution), or overlap (infinite solutions).

Adding matrices is like adding two spreadsheets cell by cell. If spreadsheet $A$ has sales for January and $B$ has sales for February, then $A + B$ gives total sales in each cell. Scalar multiplication is like giving everyone in the spreadsheet a 10% raise—multiply every entry by 1.1.

Think of a spreadsheet: rows go across, columns go down, and every cell holds a number. A $2 \times 3$ matrix is like a mini-spreadsheet with 2 rows and 3 columns. Matrices package multiple numbers into a single organized object so you can manipulate them all at once.

Imagine each row of $A$ as a question and each column of $B$ as an answer key. You 'grade' each row against each column by multiplying corresponding entries and summing. This is why column count of $A$ must match row count of $B$—the question and answer key must have the same length.

Turn a word problem into math: identify what's unknown, write relationships as equations.

A one-step equation is like unwrapping one layer of packaging. A multi-step equation has several layers: first simplify each side (distribute, combine like terms), then peel off operations one at a time until $x$ stands alone. Think of it as cleaning up a messy room before finding what you're looking for.

It works exactly like multiplying and dividing numeric fractions. To multiply: factor everything, cancel common factors across any numerator and any denominator, then multiply across. To divide: flip the second fraction and multiply. $\frac{a}{b} \div \frac{c}{d} = \frac{a}{b} \cdot \frac{d}{c}$.

Like giving directions: go 3 blocks east, then 4 blocks north — the order matters because (3, 4) is a different spot than (4, 3).

In $y = mx + b$, $m$ and $b$ are parameters — different values give different lines.

Think of like terms as the same type of object: $3x^2$ and $5x^2$ are both '$x^2$ things,' so you can combine them into $8x^2$, just like 3 apples plus 5 apples equals 8 apples. You cannot combine $x^2$ and $x$ any more than you can add apples and oranges.

Each term in the first polynomial must 'shake hands' with every term in the second. For two binomials like $(x + 3)(x + 5)$, the FOIL method (First, Outer, Inner, Last) organizes the four handshakes: $x \cdot x + x \cdot 5 + 3 \cdot x + 3 \cdot 5$.

A sum of terms like $3x^2 + 2x - 5$. The highest power is the degree.

When $x = 0$, $y = 0$. The line passes through the origin—no head start.

Each factor $(x - r)$ equals zero when $x = r$. So the factored form literally shows you where the parabola crosses the $x$-axis—plug in either root and the whole expression becomes zero.

When factoring fails, this formula always finds the x-intercepts.

The path of a thrown ball — rising then falling — traces a parabola opening downward.

Think of it as a template with three slots: $a$ controls the width and direction of the parabola, $b$ shifts it sideways, and $c$ slides it up or down. Every quadratic can be written this way by expanding and collecting like terms.

Imagine sliding a basic $x^2$ parabola around on the coordinate plane. The value $h$ shifts it left or right, $k$ shifts it up or down, and $a$ stretches or flips it. The vertex $(h, k)$ is the parabola's turning point—you can read it directly from this form.

A radical 'traps' the variable inside a square root. To free it, isolate the radical on one side, then square both sides to undo the square root. But squaring can introduce fake solutions (extraneous solutions) that do not actually satisfy the original equation, so you MUST check every answer.

Treat simplified radicals like variables: $3\sqrt{5} + 2\sqrt{5} = 5\sqrt{5}$ works just like $3x + 2x = 5x$. You can only combine radicals with the SAME radicand. Multiplication is more flexible since $\sqrt{a} \cdot \sqrt{b} = \sqrt{ab}$ always works.

Miles per hour, dollars per item, degrees per minute — change per unit.

A radical in the denominator is considered 'messy.' To clean it up, multiply top and bottom by the same radical (or conjugate). This works because $\sqrt{a} \cdot \sqrt{a} = a$, which eliminates the radical from the bottom. For binomial denominators like $3 + \sqrt{2}$, multiply by the conjugate $3 - \sqrt{2}$ to use the difference of squares pattern.

If equation 2 is just equation 1 doubled, it's redundant — the same constraint stated twice.

$2(x + 3)$ and $2x + 6$ look different but are the same—rewriting shows this.

You lend someone \$100 and they pay you \$5 every year as a thank-you — the \$5 never changes because it is always based on the original \$100.

Look inside the radical for perfect squares hiding as factors. $\sqrt{72}$ contains $36 \times 2$, and since $\sqrt{36} = 6$, you can pull the 6 out: $\sqrt{72} = 6\sqrt{2}$. Think of it as freeing numbers that are 'ready' to leave the radical.

Just like simplifying the fraction $\frac{6}{8} = \frac{3}{4}$ by canceling the common factor of 2, you can simplify $\frac{x^2 - 4}{x - 2}$ by factoring the top as $(x+2)(x-2)$ and canceling the common $(x-2)$ factor. But remember: you can only cancel FACTORS (things being multiplied), not TERMS (things being added).

How much the line goes up for every step to the right. Steeper = bigger slope.

The answer to 'what value of $x$ makes this equation true?' — found by solving, confirmed by checking.

Not just one answer, but ALL answers that work — an inequality like $x > 3$ has infinitely many.

Undo what's done to $x$ by doing the opposite: if $x + 5$, subtract 5.

Fractions make equations messy. Multiply every term by the LCD to 'clear' all the denominators at once, turning a rational equation into a simpler polynomial equation. But be careful—values that make any original denominator zero are excluded from the domain and must be rejected even if they appear as solutions.

Instead of juggling multiple equations with substitution or elimination, pack everything into a matrix and use systematic row operations. It is like organizing a messy desk—once the equations are neatly arranged in a matrix, a mechanical process (row reduction) reveals the answer. Each row operation is an allowed algebraic move (swap equations, scale an equation, add equations) performed on the matrix.

Seeing that $x^2 - 1 = (x+1)(x-1)$ is structural. Computing $7^2 - 1 = 48$ is computational.

If $y = 2x$, you can write $2x$ everywhere you see $y$—they're the same.

Instead of $2+3=3+2$ and $5+7=7+5$, write $a+b=b+a$ for ALL numbers.

'-' can mean subtraction, negative sign, or 'opposite of.' Context tells which.

Where two lines cross—the point that satisfies both equations.

'For any number $n$, $n + 0 = n$' works for ALL numbers, not just one.

Like a blank in a sentence: '$\_ + 3 = 7$' asks 'what number fits here?'

Like a box that can hold any number. '$x + 5 = 12$' asks: what's in the box?

Walk one arrow, then another; the single shortcut arrow is their sum.

Vectors are arrows with direction and magnitude. Adding two vectors is like walking along the first arrow, then continuing along the second—you end up at the tip of the combined arrow (tip-to-tail method). Scalar multiplication stretches or shrinks the arrow: $2\mathbf{v}$ is twice as long in the same direction, while $-\mathbf{v}$ points the opposite way.

Magnitude is how long the arrow is—like measuring the length of a stick. Direction is which way it points. A unit vector is a 'pure direction' with length 1, like a compass needle. To get the unit vector, shrink or stretch the vector until its length is exactly 1 while keeping it pointed the same way.

Fold the parabola along the axis of symmetry and both halves match perfectly. The vertex is at the fold—the very bottom of a U-shaped parabola or the very top of an upside-down one. It is the point where the function changes direction.

Word problems are stories in disguise. Your job is to find the main character (the unknown—call it $x$), figure out what's happening to it (the operations), and write down the punchline (the equation). 'Five more than twice a number is 17' becomes $2x + 5 = 17$.

The zeros are where the parabola crosses or touches the $x$-axis. A parabola can cross twice (two zeros), just touch once (one repeated zero), or miss entirely (no real zeros). You can find them by factoring, completing the square, or using the quadratic formula.

To find the area between two curves, subtract the lower curve from the upper curve and integrate. It's like finding the area under the top curve and subtracting the area under the bottom curve—the difference is the area of the 'sandwich' between them.

Add the same number each time — 2, 5, 8, 11,... (add 3 each step). This is constant-rate growth.

Derivative of outside times derivative of inside. Unpack layers.

Convergence means the infinite sum adds up to a finite number—each new term adds less and less, and the total stabilizes. Divergence means the sum either blows up to infinity or never settles down. The key question: does adding infinitely many terms produce a finite result?

The first derivative tells you whether the function goes up or down (like reading a speedometer). The second derivative tells you whether it's speeding up or slowing down (like reading an accelerometer). Together, they give you a complete picture of the curve's shape.

The signed total area under the curve from $a$ to $b$—positive above the $x$-axis, negative below.

How fast is the output changing right now? The slope of the curve at each point.

Shortcuts so you don't have to use the limit definition every time.

Integration undoes differentiation. They're two sides of the same coin.

Multiply by the same number each step — 2, 6, 18, 54,... (multiply by 3). This is exponential growth.

Sometimes you can't (or don't want to) solve for $y$ explicitly. Instead, differentiate the whole equation as-is. Every time you differentiate a $y$-term, attach $\frac{dy}{dx}$ by the chain rule (since $y$ secretly depends on $x$), then solve for $\frac{dy}{dx}$.

Can an infinite region have a finite area? Surprisingly, yes. The area under $\frac{1}{x^2}$ from 1 to infinity is exactly 1. Improper integrals extend integration to infinite intervals and unbounded functions by using limits to handle the 'improper' part.

If each term is a fixed fraction of the previous one, the terms shrink fast enough that the total sum stays finite. Imagine walking halfway to a wall, then half the remaining distance, then half again—you approach the wall but the total distance is finite (exactly the full distance to the wall).

Going on forever without end. Infinity is a direction or limiting idea, not a number you can reach or write down.

If derivative gives rate, integral gives total. Derivative of position = velocity; integral of velocity = position.

The product rule for derivatives says $(uv)' = u'v + uv'$. Rearranging and integrating gives integration by parts. The idea is to trade your original integral for a (hopefully easier) one. You're transferring the derivative from one factor to the other.

A continuous function can't skip values. If you start below a line and end above it, you must cross it somewhere. It's like driving from sea level to a mountaintop—you pass through every elevation in between.

An algebraic equation like $x^2 = 4$ asks 'what number satisfies this?' A differential equation like $\frac{dy}{dx} = 2x$ asks 'what function has this derivative?' The answer isn't a number but a family of functions: $y = x^2 + C$.

When both numerator and denominator go to zero (or both to infinity), the limit depends on which one gets there faster. Taking derivatives measures the rates at which they approach 0 or $\infty$, so the ratio of derivatives captures this 'race.'

What output do you get closer and closer to as you get closer and closer to some input?

If you drive 150 miles in 2 hours, your average speed is 75 mph. The MVT says at some instant during the trip, your speedometer read exactly 75 mph. The instantaneous rate must equal the average rate at least once.

Find where the function hits its peaks (maxima) and valleys (minima) by finding where the slope is zero.

Just as $\frac{7}{12}$ can be split into $\frac{1}{3} + \frac{1}{4}$, a complex fraction like $\frac{5x-1}{(x+1)(x-2)}$ can be split into $\frac{A}{x+1} + \frac{B}{x-2}$. The simpler pieces are each easy to integrate.

A power series is an 'infinite polynomial' centered at $c$. For each value of $x$, you get a number series that may or may not converge. The set of $x$-values where it converges forms an interval centered at $c$, and within that interval, the power series behaves like a well-defined function.

How much does the output change for each unit increase in input? That ratio is the rate of change.

A recursive formula is like step-by-step directions ('from where you are, go 3 blocks north'). An explicit formula is like GPS coordinates ('go to 5th Avenue and 42nd Street'). Both describe the same sequence, but explicit formulas let you jump to any term instantly.

If two quantities are linked by an equation, their rates of change are also linked. A balloon inflating: as the radius increases, the volume increases too. How fast does the volume grow if the radius grows at 2 cm/s? The chain rule connects the rates.

Imagine filling the area under a curve with thin rectangles. The more rectangles you use, the better the approximation. In the limit of infinitely many infinitely thin rectangles, you get the exact area—which is the definite integral.

If the rate of change factors into a piece that depends only on $x$ and a piece that depends only on $y$, you can sort them onto opposite sides of the equation—all the $y$-stuff on the left, all the $x$-stuff on the right—then integrate each side in its own variable.

A pattern of numbers: first term, second term, third term, and so on.

Add up all the terms: $a_1 + a_2 + a_3 + \ldots$ — an infinite series can still have a finite sum if terms shrink fast enough.

Sigma notation is shorthand for 'add these all up.' The letter below $\Sigma$ is a counter, the number below is where to start, the number above is where to stop, and the expression to the right tells you what to add each time.

Imagine a field with tiny arrows showing which way a river flows at each point. A slope field is the same idea: the DE tells you the slope (direction) at every point, and solution curves are paths that follow these directions everywhere. Drop a 'particle' anywhere and follow the arrows—that's a solution.

If $f$ is squeezed between two functions that both approach the same value $L$, then $f$ has no choice—it must also approach $L$. Like being caught between two walls closing in to the same point.

The tangent line is the unique straight line that best approximates the curve at a specific point — same value, same slope.

Approximate any smooth function with a polynomial by matching the function's value, slope, curvature, and all higher derivatives at a single point. The more terms you include, the wider the region where the polynomial closely matches the function. It's like fitting a polynomial glove onto the function's hand.

Continuous means you can draw the graph without lifting your pen. A removable discontinuity is a single hole you could fill in. A jump discontinuity is a gap where the function leaps to a different value. An infinite discontinuity is where the function shoots off to infinity (a vertical asymptote).

When you see a composite function inside an integral along with its inner derivative lurking nearby, substitution collapses the composition into a single variable. It's like un-nesting a function: replace the inner part with $u$, and the integral becomes simpler.

Spin a flat region around a line, like spinning a pottery wheel. The flat shape sweeps out a 3D solid. To find its volume, slice the solid into thin pieces (discs, washers, or shells), find the volume of each slice, and add them up—which means integrate.

You can only add like-sized pieces directly — $\frac{1}{4}$ and $\frac{1}{3}$ must be converted to twelfths before adding.

If you have $\frac{2}{5}$ of a pie and get $\frac{1}{5}$ more, you now have $\frac{3}{5}$—same size pieces, just count them up.

You can't add thirds and fourths directly—it's like adding apples and oranges. Convert both to twelfths first, then add.

To compare $\frac{3}{4}$ and $\frac{5}{6}$, rewrite them with the same denominator so the numerators can be compared directly.

Decimal operations follow the same rules as whole numbers, but you must track the decimal point carefully—like keeping track of dollars and cents.

Fractions and decimals are two ways to write the same number. $\frac{3}{4}$ and $0.75$ are the same amount—just different notation.

Money uses decimals: \$3.50 means 3 dollars and 50 cents (half a dollar).

Imagine you have 2 cups of flour and each serving of a recipe needs $\frac{1}{3}$ cup. How many servings can you make? You are asking 'how many one-thirds fit into 2?'—that is $2 \div \frac{1}{3} = 6$ servings. Division by a fraction counts how many pieces of that size fit inside the whole.

Half a pizza is the same whether cut into 2 or 4 pieces: $\frac{1}{2} = \frac{2}{4}$.

Like a regular line plot, but the number line uses fractions instead of whole numbers — each mark shows a measurement like $\frac{3}{4}$ inch.

$\frac{3}{4}$ of 20 means split 20 into 4 equal groups (5 each), then take 3 groups: $3 \times 5 = 15$.

Divide the space between 0 and 1 into equal parts. $\frac{3}{4}$ means go 3 of the 4 equal parts from 0.

A pizza cut into 4 slices—eating 1 slice means you ate $\frac{1}{4}$ of the pizza.

$\frac{7}{4}$ means you have 7 quarter-pieces—that's more than one whole (which would be $\frac{4}{4}$).

You ate 2 whole pizzas and $\frac{3}{4}$ of a third pizza—that's $2\frac{3}{4}$ pizzas.

Mixed to improper: multiply the whole number by the denominator, add the numerator, keep the denominator. Improper to mixed: divide numerator by denominator to get the whole part and remainder.

$\frac{2}{3} \times \frac{3}{4}$ means 'two-thirds of three-quarters.' Take $\frac{3}{4}$ of something, then take $\frac{2}{3}$ of that result.

Convert all fractions to a common denominator and then read off the order from the numerators.

A 20% tip on a \$45 meal: $0.20 \times 45 = \$9$ tip, so total is \$54. A 30% discount on \$80: save \$24, pay \$56.

If a price goes from \$50 to \$60, the change is \$10. Compared to the original \$50, that's $\frac{10}{50} = 20\%$ increase.

$25\%$ of 80 means 'one quarter of 80.' Convert $25\%$ to $0.25$ and multiply: $0.25 \times 80 = 20$.

Percent means 'per hundred.' $25\%$ means 25 out of every 100.

If 2 candies cost \$1, then 4 candies cost \$2—same proportion.

60 miles per hour tells you how many miles you travel for each hour — it compares distance to time.

A recipe that uses 2 cups flour for every 1 cup sugar has a $2:1$ ratio.

You have $\frac{5}{8}$ of a cake and eat $\frac{2}{8}$. Same size slices, so subtract the count: $\frac{3}{8}$ remains.

To find $\frac{3}{4} - \frac{1}{3}$, convert to twelfths: $\frac{9}{12} - \frac{4}{12} = \frac{5}{12}$. Same idea as addition, just subtract.

Amplitude is the maximum displacement from the middle of a wave — it is half the total height of a full oscillation from crest to trough.

Imagine depositing \$100 every month into a savings account. Each deposit earns interest for a different amount of time—the first deposit earns interest for the full term, the last deposit barely earns any. An annuity formula adds up all these differently-growing deposits in one clean expression, instead of computing compound interest on each payment separately.

The graph gets infinitely close but never touches—like chasing something forever.

Your calculator only has $\ln$ and $\log_{10}$ buttons. The change-of-base formula lets you compute ANY logarithm using whichever base you have available. It works because all logarithms are proportional to each other—changing base just changes the scale factor.

Changing rate means accelerating or decelerating progress — like compound interest where each year's gain is larger than the last because the base keeps growing.

Work from the innermost function outward — compute $h(x)$ first, then feed that result to $g$, then feed that to $f$. The order matters critically.

Simple interest pays you only on your original deposit. Compound interest pays you interest on your interest—your money earns money on the money it already earned. The more frequently you compound, the more you earn, because each tiny interest payment starts earning its own interest sooner. The ultimate limit of compounding more and more frequently is continuous compounding: $A = Pe^{rt}$.

Imagine a flashlight shining on a wall. Straight on: circle. Tilted slightly: ellipse. Tilted to match the cone's edge: parabola. Tilted past the edge: hyperbola. All four shapes come from the same geometric object (a cone), just viewed from different angles.

Constant rate means steady, uniform progress — like a car traveling at a fixed speed: every hour, the same number of miles is added to the total.

A continuous function can be drawn without lifting the pencil — there are no sudden jumps, gaps, or points that shoot to infinity.

Like a flowchart: A affects B, B affects C. Arrows show dependencies.

The domain is the list of valid "questions" you can ask the function — values outside the domain produce undefined or meaningless answers.

What if both angles in the sum formula are the same? Setting $A = B = \theta$ in the sum identities gives you the double-angle formulas. They answer: if you know the trig values for an angle, what are the trig values for twice that angle? The cosine double-angle formula is especially versatile because it has three equivalent forms, each useful in different situations—pick whichever one simplifies your problem.

Imagine pinning two ends of a loose string to a board (these are the foci), then tracing a curve with a pencil keeping the string taut. The resulting oval shape is an ellipse. A circle is just a special ellipse where both foci coincide.

A circle is the set of all points at the same distance (the radius) from a center point. The equation just says 'the distance from $(x, y)$ to the center $(h, k)$ equals $r$,' using the distance formula squared.

The 'natural' base for growth—what you get from continuous compounding.

Even means mirror across $y$-axis; odd means rotational symmetry through the origin.

Growth (or decay) that multiplies by a constant factor repeatedly.

Exponential growth means the amount added each period is proportional to the current amount — the bigger it gets, the faster it grows, creating an accelerating curve.

Microphone feedback: sound → speaker → microphone → more sound → louder...

Frequency counts how many complete cycles occur per unit of the horizontal axis — higher frequency means the wave oscillates more rapidly in the same space or time.

A machine: put something in, get exactly one thing out. Same input always gives same output.

Like a dictionary: every word maps to a definition. Every input maps to an output.

Chain two machines together—output of the first goes into the second.

$y = mx + b$ is a family of lines. Different $m$ and $b$ give different lines.

The notation $f(x)$ is not "$f$ times $x$" — it means "the output of function $f$ when the input is $x$." The parentheses contain the input, not a multiplication.

Moving or reshaping a graph without changing its basic shape.

Translate a situation into a function, then use math to analyze it.

Growth compounds: each period's increase is larger than the last. Decay shrinks: each period's decrease is smaller than the last, never quite reaching zero.

A horizontal line that crosses the graph at two points means those two inputs produce the same output — the function is many-to-one and has no inverse without domain restriction.

While an ellipse keeps the SUM of distances to foci constant, a hyperbola keeps the DIFFERENCE constant. This creates two separate curves that open away from each other, each curving toward (but never reaching) a pair of asymptotic lines.

Like a vending machine: put in selection (input), get out snack (output).

Shifting a parabola doesn't change that it's a parabola—shape is invariant.

If $f$ turns $a$ into $b$, then $f^{-1}$ turns $b$ back into $a$. Reverse the process.

Regular trig functions answer: 'Given an angle, what's the ratio?' Inverse trig functions answer the reverse: 'Given a ratio, what's the angle?' Since $\sin$ and $\cos$ are many-to-one (many angles give the same ratio), we must restrict their domains to make the inverse a proper function. Think of it like this: if you know the slope of a ramp is $0.5$, $\arcsin(0.5) = 30°$ tells you the angle.

In 2D, a line is defined by a slope and a point ($y = mx + b$). In 3D, slope doesn't work—there's no single number for direction in space. Instead, you specify a starting point and a direction vector (an arrow pointing along the line). The parameter $t$ acts like a slider: at $t = 0$ you're at the starting point, and as $t$ increases or decreases, you slide along the line in the direction of the vector.

Local is "zoom in on one spot"; global is "zoom out to see the whole picture." Near $x = 0$, $\sin(x) \approx x$ (local linear approximation), but globally it oscillates forever.

The exponent that produces a number. $\log_2(8) = 3$ because $2^3 = 8$.

Logarithms were invented to turn hard operations into easy ones. Multiplication becomes addition, division becomes subtraction, and exponentiation becomes multiplication. This is why slide rules worked—they added lengths (logarithms) to multiply numbers.

Multiple students can have the same grade—many inputs, one output.

Same function, different views: $y = 2x$ as formula, as table, as line, as 'doubling.'

If $e^x$ asks 'what do I get after growing continuously for time $x$?', then $\ln x$ asks 'how long do I need to grow continuously to reach $x$?' The natural log measures time in the world of continuous growth.

No two inputs share the same output—like social security numbers.

Every point on a parabola is exactly the same distance from the focus as it is from the directrix line. This geometric property is why satellite dishes and flashlight reflectors are parabolic—signals from the focus reflect off the curve in parallel lines.

Instead of saying '$y$ depends on $x$,' parametric equations say 'both $x$ and $y$ depend on time $t$.' Imagine an ant walking on a table—at each moment $t$, the ant has an $x$-position and a $y$-position. The path it traces is the parametric curve, and $t$ is the clock ticking forward.

To sketch a parametric curve, make a table of $t$, $x$, and $y$ values, then plot the $(x, y)$ points and connect them in order of increasing $t$. Arrows on the curve show the direction of travel. Alternatively, you can sometimes eliminate $t$ to get a familiar Cartesian equation—but you may lose information about direction and speed.

It is the original template shape you move, stretch, or reflect.

The same pattern over and over. Like a heartbeat or the seasons.

Think of the behavior as shifting gears — the function follows one rule until it hits a boundary, then switches to a different rule for the next region.

A piecewise function is like a rulebook: look up which rule applies to your input value, then use only that rule to compute the output.

Think of a plane as a perfectly flat, infinite floor that can be tilted at any angle in space. A horizontal floor is one plane; tilt it and you get another. To describe which tilt you have, imagine sticking a pole straight up out of the floor—that pole is the normal vector, and it captures the exact orientation of the surface. Any flat sheet in 3D, no matter how it's angled, is completely determined by where it sits and which direction its pole points.

Instead of 'go right 3, up 4' (Cartesian), polar says 'go 5 units in the direction of 53°.' It's how a radar works—distance and direction from a central point. Some shapes that look complicated in Cartesian coordinates become beautifully simple in polar.

As the angle $\theta$ sweeps around, the distance $r$ changes according to the equation, tracing out a curve. Think of it like a radar sweep where the blip's distance from the center varies with direction. This creates curves with stunning symmetry that would require complex implicit equations in Cartesian coordinates.

Sums of power terms with whole-number exponents. The building blocks of functions.

Would you rather have \$100 today or \$100 in five years? Today, obviously—because you could invest the \$100 and have MORE than \$100 in five years. Present value answers: 'How much would I need TODAY to have \$X in the future?' Future value answers: 'If I invest \$X today, what will it become?' Discounting is the reverse of compounding—it shrinks future money back to today's value.

Double the input, double the output. No offset—starts at zero.

On the unit circle, the point $(\cos\theta, \sin\theta)$ is always at distance 1 from the origin. By the Pythagorean theorem, $x^2 + y^2 = 1$ becomes $\cos^2\theta + \sin^2\theta = 1$. This single fact—that sine and cosine are tied to a circle—generates all three Pythagorean identities. Dividing through by $\cos^2\theta$ or $\sin^2\theta$ produces the other two forms.

Imagine wrapping the radius of a circle along its edge like a piece of string. The angle you've swept out is exactly 1 radian. Since the full circumference is $2\pi r$, a full turn is $2\pi$ radians. Radians measure angles in terms of the circle itself, which is why they're the natural unit for calculus and physics—no arbitrary conversion factor like $360$ is needed.

It ties angle directly to the circle’s geometry instead of degree counting.

The range is the set of all possible "answers" the function can give — some output values may be unreachable no matter what valid input you choose.

Rational functions are the "fractions" of the function world — they behave like polynomials except near the zeros of the denominator, where they blow up or have holes.

$-f(x)$ flips over x-axis (upside down). $f(-x)$ flips over y-axis (mirror).

You keep only the input interval where the function behaves one way.

Room fills until no more people fit. Growth can't continue forever.

Vertical scaling stretches or squishes the graph up/down; horizontal scaling stretches or squishes it left/right. Both change the function's measurements without altering its fundamental character.

A sensitive scale notices tiny weight differences. An insensitive one doesn't.

Shifting is like sliding the entire graph on the coordinate plane — the function's shape is completely unchanged, only its position moves.

When the variable is trapped in an exponent, logarithms free it. Taking $\log$ of both sides brings the exponent down to ground level where you can solve for it using algebra.

If logarithms trap the variable inside a $\log$, converting to exponential form releases it. The key insight is that $\log_b(\text{stuff}) = c$ means $b^c = \text{stuff}$—just rewrite and solve.

A ball in a bowl returns to center; a ball on a hill rolls away.

Imagine a staircase: the height is constant on each step, then jumps up (or down) at each transition. Postal rates, grade cutoffs, and floor() all create steps.

What happens when you combine two rotations? If you rotate by angle $A$ and then by angle $B$, the result involves both angles interacting. The sum and difference formulas tell you exactly how the trig values of two separate angles combine. They're like a multiplication rule for rotations—the result isn't simply adding the trig values, but mixing sines and cosines together.

Even functions are symmetric about the y-axis: $f(-x) = f(x)$. Odd functions have 180° rotational symmetry about the origin: $f(-x) = -f(x)$.

If you track the $y$-coordinate of a point moving around the unit circle and plot it against the angle, you get the sine wave. It's the shape of ocean waves, sound waves, and alternating current. The general form $y = a\sin(bx - c) + d$ lets you control four properties: how tall the wave is ($a$, amplitude), how fast it repeats ($b$, affecting period), where it starts ($c$, phase shift), and its vertical center ($d$, vertical shift).

Angles have numbers associated with them—sin, cos, tan capture different ratios.

Imagine walking around a circle of radius 1. Your $x$-coordinate is $\cos\theta$ and your $y$-coordinate is $\sin\theta$. Instead of being limited to right triangles, the unit circle lets you define sine and cosine for ANY angle—even angles bigger than $360°$ or negative angles. Every point on the circle is at distance 1 from the center, so the hypotenuse is always 1, and the trig ratios simplify to just coordinates.

It translates shapes into equations so algebra can solve geometry problems.

A protractor is like a ruler for turns — it tells you exactly how much one line has rotated from another.

Think of opening a book flat on a table—the two pages form supplementary angles (they add to a straight line, $180°$). Now think of the corner of a room where two walls meet the floor—those two angles are complementary (they add to a right angle, $90°$). When two lines cross like an X, the opposite angles are always equal—those are vertical angles.

Opening a door wider makes a bigger angle; a corner of a book is $90°$.

Imagine walking along a circular track but only covering a portion of the full loop. The arc length is how far you actually walked. If you walk a quarter of the circle ($90°$), you cover a quarter of the circumference. The fraction of the full circle you cover determines the fraction of the circumference you walk.

How many unit squares would you need to tile the inside of the shape completely, with no gaps?

Imagine cutting a pizza into many thin slices and rearranging them into a shape that looks like a rectangle. The 'height' of that rectangle is the radius $r$, and the 'width' is half the circumference ($\pi r$). So the area is $r \times \pi r = \pi r^2$.

Cut a triangle off one end of the parallelogram and slide it to the other end — you get a rectangle with the same base and height.

Two identical trapezoids fit together to form a parallelogram. The trapezoid is half of that parallelogram.

Every triangle is exactly half of a rectangle with the same base and height — cut the rectangle along the diagonal.

Shapes are like cookie cutters—circles are round, squares have 4 equal sides.

A fence around a yard—it marks exactly where 'inside the yard' ends and 'outside' begins.

Imagine standing at the center of a clock face. The angle between the hour and minute hands is a central angle. The arc between the two numbers the hands point to is the intercepted arc, and its measure (in degrees) equals the angle you see.

Spin around with your arm fully outstretched—your fingertip traces a perfect circle.

Imagine wrapping a string tightly around a circular jar lid, then straightening the string out. That length is the circumference. No matter the size of the circle, the circumference is always $\pi$ times the diameter—roughly $3.14$ laps of the diameter around the edge.

Order matters, like doing rotate then reflect versus reflect then rotate.

If you could pick up one shape and place it exactly on the other, they're congruent.

Imagine building a triangle from sticks and hinges. If you fix all three side lengths (SSS), there's only one triangle you can make. If you fix two sides and the angle between them (SAS), the triangle is locked in place. You don't need all six measurements—just the right three.

Instead of arguing with angles and congruence marks, drop the shape onto a grid and let algebra do the heavy lifting. Want to prove a quadrilateral is a parallelogram? Calculate all four slopes—if opposite sides have equal slopes, they're parallel, and you're done. Coordinates turn visual intuition into airtight calculation.

Every point has a unique numerical 'address' like $(3, 4)$ that locates it exactly on the plane.

Slice an orange—the cut surface is a cross-section (a circle).

Imagine slicing a loaf of bread—each slice reveals a 2D shape. The shape you see depends on the angle and position of your cut. Slice a cylinder straight across and you get a circle; slice it at an angle and you get an ellipse. Slice a rectangular prism and you can get rectangles, triangles, or even hexagons depending on the cut.

A tight turn has high curvature; a gentle bend has low curvature.

Like zooming in or out on a photo—everything gets bigger or smaller proportionally.

0D = point (no direction). 1D = line (one direction). 2D = plane. 3D = space.

North, south, east, west—or the way an arrow points, regardless of how long the arrow is.

Where you ended up relative to where you started—direction and distance combined.

'As the crow flies'—the straight-line separation between two locations.

Imagine two points on a grid. Draw a horizontal line from one and a vertical line from the other to form a right triangle. The horizontal leg is the difference in $x$-coordinates, the vertical leg is the difference in $y$-coordinates, and the hypotenuse—the direct distance—comes from the Pythagorean theorem. The distance formula is just $a^2 + b^2 = c^2$ in coordinate clothing.

Draw a right triangle between the two points — the horizontal and vertical distances are the legs, and the straight-line distance is the hypotenuse.

Imagine standing at one corner of a triangular park and looking along one side. The exterior angle is how far you'd turn to look back along the other side. That turn combines the 'bends' at the other two corners—it equals their angles added together.

A map isn't the territory—it abstracts away most details to show what matters.

A door hinge constrains the door to swing in an arc, not slide sideways.

What stays exactly the same when you move, rotate, or flip a shape? Those unchanging things are invariants.

Modeling a house as boxes and triangles; a planet as a sphere.

What rectangle with fixed perimeter has the most area? A square!

It is a legal argument where each line needs a valid reason.

Moving, rotating, flipping, or stretching a shape to produce a new image of that shape.

Use a smaller, measurable shadow to infer a taller object’s height.

Instead of measuring sides and angles, show that one shape can be moved, flipped, or resized to land exactly on another.

Imagine sitting on the edge of a circular stadium and looking at two players on the field. The angle your eyes make is an inscribed angle. No matter where you sit on the same arc, that viewing angle stays the same—and it's always half of what you'd see from the center. It's like the circle is 'halving' your perspective compared to the center's view.

A closed fence divides the world into two zones: the yard inside and everything else outside. Any closed curve does the same—splitting the plane into an interior region and an exterior region.

Where two roads cross—that single crossing point is their intersection.

A perfectly straight edge that goes on forever in both directions.

Think of filling a water bottle — liquid volume tells you how much water fits inside.

Mass tells you how heavy something feels — a paperclip is about 1 gram, a textbook is about 1 kilogram.

Finding the midpoint is like finding the average position. If two friends live at different addresses on the same street, the midpoint is the house number exactly halfway between them—the average of their two house numbers. In 2D, you just average both coordinates independently.

Picture a triangular picture frame hanging on a wall. Stretch a rubber band between the midpoints of two sides. That rubber band runs perfectly parallel to the bottom of the frame, like a miniature shelf—and it spans exactly half the width. No matter how you reshape the triangle, that halfway connection always mirrors the opposite side at half scale.

Unfold a 3D solid like a cardboard box—the flat connected pattern you get is a net of that solid.

Which way is up? Which way are you facing? That's orientation.

How many oranges can you stack in a box? How to arrange them?

Parallel tracks run side by side; perpendicular streets form a plus sign.

Railroad tracks—they stay exactly the same distance apart and never meet, no matter how far they extend.

If an ant walked around the edge of a rectangle, perimeter is how far it walked.

The corner of a book or a room—the two edges meet at precisely $90°$.

No matter how big or small the circle, circumference $\div$ diameter always equals $\pi$.

An infinite sheet of paper with absolutely no thickness, extending forever in every direction.

The tip of a pencil or a dot on a map. Position only, no width or length.

Connect-the-dots that closes into a shape—no curves allowed.

A shadow cast on the ground is a projection—a 3D object mapped down to a 2D silhouette.

Similar triangles have proportional sides: if one side doubles, all sides double.

If you draw squares on each side of a right triangle, the two smaller squares fill the big one exactly.

Think of quadrilaterals as a family tree. The most general is any four-sided shape. Add one pair of parallel sides and you get a trapezoid. Add two pairs and you get a parallelogram. Make the angles right and it becomes a rectangle. Make the sides equal and it becomes a rhombus. A square is the 'royal' member—it has every property: parallel sides, equal sides, and right angles.

Like looking in a mirror—left and right are swapped, but size and shape are perfectly preserved.

Imagine a ramp leaning against a wall. The steepness depends on the ratio of how high the wall is to how long the ramp is. Trigonometry gives names to these ratios: sine is how high compared to the ramp, cosine is how far along the ground compared to the ramp, and tangent is how high compared to how far along the ground. No matter how big or small the ramp, if the angle is the same, these ratios stay the same.

A triangle made of sticks is rigid. A rectangle made of sticks can collapse into a parallelogram.

Like a Ferris wheel turning around its center hub—every seat traces a circle, staying the same distance from the axle while sweeping through the same angle.

If you turn it and it still fits exactly, it has rotational symmetry.

A map is a scale drawing of the real world. If 1 inch on the map equals 10 miles in reality, the scale factor is $1:10\text{ miles}$. Every distance on the map uses the same ratio, so the shapes stay accurate—just smaller. Enlarging a photo works the same way in reverse.

Double the size: length $\times 2$, area $\times 4$, volume $\times 8$.

Imagine cutting a pizza into slices. Each slice is a sector. If you cut the pizza into 4 equal slices ($90°$ each), each slice has $\frac{1}{4}$ of the pizza's total area. The sector area is simply the fraction of the full circle determined by the central angle, applied to the total area.

On a flat surface the straight line is always the shortest path between any two points.

One figure is an enlarged or reduced copy of another—same shape, same angles, but possibly different size.

A photo and its enlargement are similar—same shape, different size.

Think of a photo and its enlargement. They look the same but are different sizes. For triangles, you only need to check that two angles match (AA)—if the angles are the same, the shape is the same, even if the size differs. It's like verifying two buildings have the same blueprint, even if one is a scale model.

A ramp's steepness—the ratio of how high it rises to how far it goes horizontally.

Imagining how furniture will fit in a room before physically moving any of it.

Cut an equilateral triangle in half and you get a 30-60-90 triangle. Cut a square along its diagonal and you get a 45-45-90 triangle. These two cuts give you exact side ratios you can memorize forever—no calculator needed.

The 'skin area' of a perfectly round ball—the amount of material needed to cover it with no overlaps.

How much wrapping paper would you need to completely cover every face of a gift box?

Imagine peeling the label off a can of soup. The label is a rectangle whose width is the circumference of the can ($2\pi r$) and whose height is the can's height ($h$). Add the two circular lids (top and bottom), and you have the total surface area.

Imagine unfolding a cereal box and laying it flat—you get a net of six rectangles. The surface area is the total area of that flattened cardboard. For any prism, you always have two identical bases plus a 'belt' of rectangles wrapped around the middle.

A butterfly's wings are symmetric—fold it down the middle and both sides match.

A basketball resting on a flat floor—the floor touches the ball at exactly one point.

Imagine a ball sitting on a flat floor. The floor touches the ball at exactly one point—that's tangency. The floor (tangent line) is perfectly perpendicular to a line from the ball's center to the contact point (the radius). No matter how you tilt the flat surface, if it only touches at one point, it must be perpendicular to the radius there.

Like a bathroom floor tile pattern that fits together perfectly and could extend forever in all directions.

Bathroom tiles cover the floor perfectly—no gaps between them.

A coffee mug and a donut are 'the same' topologically—both have one hole.

Sliding a chess piece straight across the board—every point moves the same amount, same direction.

Imagine a ladder leaning against two horizontal rails (the parallel lines). The ladder is the transversal. At each rail, the ladder makes the same pattern of angles—like a stamp pressed in two places. Corresponding angles are in matching positions at each crossing, and they're always equal when the rails are parallel.

Tear off the three corners of any paper triangle and line them up—they always form a straight line ($180°$). No matter how pointy or flat the triangle is, the angles always add up the same way, like three puzzle pieces that always complete a half-turn.

Try to build a triangle with two short sticks and one very long one—you can't. The two short sticks can't reach across to close the shape. It's like trying to take a shortcut: the direct path (one side) is always shorter than going around (the other two sides combined).

The simplest polygon—you need at least 3 sides to enclose space.

An arrow: how long it is (magnitude) and which way it points (direction).

How many cubic centimetre blocks would it take to completely fill the inside of the object?

Imagine filling a cone-shaped paper cup with water and pouring it into a cylinder of the same width and height. You'd need to fill the cone exactly three times to fill the cylinder. A cone is a cylinder that 'tapers to a point,' losing two-thirds of its volume in the process.

Imagine stacking hundreds of identical circular coins into a tall tower. Each coin is a thin circle with area $\pi r^2$, and stacking $h$ units high gives you a cylinder. The volume is just the area of one coin times the height of the stack.

Imagine filling a sphere with water, then pouring all that water into a cylinder that has the same radius and a height equal to the sphere's diameter ($2r$). The sphere fills exactly two-thirds of the cylinder. Archimedes was so proud of discovering this relationship that he had it carved on his tombstone.

Imagine filling a box with small cubes — the total number of cubes is the volume.

We use 3.14 for $\pi$, knowing it's not exactly right but close enough.

We group things by tens—probably because we have 10 fingers.

Which is bigger? Which is smaller? Are they the same? Comparison answers these questions with precision.

Extending numbers into a second dimension to solve equations like $x^2 = -1$.

Numbers that can be built by multiplying smaller numbers together.

Like pointing to each toy and saying '1, 2, 3...' to know how many toys you have.

$\sqrt[3]{27}$ asks: what number times itself times itself equals 27? Answer: 3, because $3 \times 3 \times 3 = 27$. For negatives, $\sqrt[3]{-8} = -2$ because $(-2) \times (-2) \times (-2) = -8$.

Just like $234 = 200 + 30 + 4$, we have $2.34 = 2 + 0.3 + 0.04$.

No matter how close two numbers are, you can always find one between them.

People come in whole numbers (discrete). Height can be any value (continuous).

Can you share 12 cookies equally among 4 people? Yes, 3 each. 12 is divisible by 4.

Like a balanced scale—both sides weigh the same. If you add weight to one side, you must add to the other.

Quick mental math to get 'close enough'—is $48 \times 52$ closer to 2000 or 3000?

Since $a^3 = a \cdot a \cdot a$ and $a^2 = a \cdot a$, multiplying them gives $a \cdot a \cdot a \cdot a \cdot a = a^5$. You just add the counts. All the other rules follow the same logic of counting how many times you multiply.

Factors are the 'building blocks' you multiply together to make a number.

A jar of 100 marbles is finite. The counting numbers are infinite.

The biggest 'piece' size that fits evenly into two numbers—like the largest tile that covers both a 12-unit and 18-unit floor.

Numbers never stop—there's always a bigger one. Infinity isn't a number, it's a direction.

Temperature can go above or below zero—integers include both directions.

More workers = less time to finish. Double the workers, halve the time.

$\pi$ and $\sqrt{2}$ go on forever without any pattern—you can't write them as a fraction.

The first number that appears in both times tables—where two counting sequences land on the same value.

How big something is, regardless of which way it points—5 miles east and 5 miles west are the same distance.

Like comparing piles of blocks—the taller pile has more. Or compare two rows one-to-one; the row with leftover has more.

Skip-counting produces multiples: counting by 3s gives 3, 6, 9, 12... — those are the multiples of 3.

If zero is sea level, negative numbers are depths below the surface — temperature $-5°$ is 5 degrees below freezing.

Numbers aren't just for counting objects—they tell us 'how much' of anything.

Numbers live in order on a line—smaller to the left, larger to the right.

Knowing that 100 is way more than 10, or that 7 is between 5 and 10.

Numbers aren't random—they have deep structure (primes, factors, operations).

Numbers live on a line—you can always put them in order from left to right.

Can you split it into two equal groups? Yes = even, no = odd.

'Per cent' means 'per hundred'—$25\%$ means 25 out of every 100.

In 352, the 3 is worth 300 because it's in the hundreds place.

How many decimal places matter? Measuring in inches vs. millimeters.

Break a number into building blocks that cannot be split further (primes).

Primes can't be broken down further—they're the 'atoms' of multiplication.

If you double one, you double the other. Triple one, triple the other.

Before we count, we notice there's 'some amount' of something—quantity is that raw sense of how much.

Any number you can write as a fraction, including decimals that end or repeat.

Any number you can point to on an infinitely precise number line.

Simplifying for easier calculation or communication—\$19.87 becomes 'about \$20'.

Zooming in or out—everything gets bigger or smaller by the same factor.

Instead of writing out all the zeros in 93,000,000 or 0.000042, you slide the decimal point and count how many places it moved. The exponent on 10 keeps track of the shift.

Multiplying and dividing are straightforward: multiply or divide the coefficients and add or subtract the exponents. Adding and subtracting require matching the powers of 10 first, like finding a common denominator.

Think of them as the digits you can trust from a measuring tool.

The building blocks of fractions—$\frac{1}{2}$ is one of two equal parts, $\frac{1}{4}$ is one of four.

Zero is the placeholder that makes '10' different from '1'—it marks empty positions.

$-5$ and $5$ are both 5 units from zero, so $|-5| = |5| = 5$.

Think of money: $\$3.75 + \$2.50$. You line up the dollars with dollars, the dimes with dimes, and the pennies with pennies. The decimal point is the anchor that keeps everything in the right place.

Think of putting groups together—3 apples plus 2 apples gives 5 apples.

When you pour two cups of water together, you get their combined amount.

$(2 + 3) + 4 = 2 + (3 + 4)$. How you group the additions doesn't matter.

An equation is like a balanced scale—add weight to both sides equally.

Think of buildings on a city skyline—taller buildings stand out. In a bar graph, taller bars mean bigger numbers. You can compare at a glance without reading every number.

Temperature tomorrow will be between 60F and 75F. Those are bounds.

$\frac{6}{8} = \frac{3}{4}$ because we can cancel the common factor 2 from top and bottom.

$3 + 5 = 5 + 3$ and $3 \times 5 = 5 \times 3$. Swapping the order doesn't change the answer.

If $y$ is always 3 times $x$, the constant of proportionality is 3.

You can't spend more money than you have—that's a constraint.

Just as moving left of the decimal point makes each place 10 times bigger (ones, tens, hundreds), moving right makes each place 10 times smaller (tenths, hundredths, thousandths). It's like zooming in—each step splits things into 10 equal pieces.

Distance varies directly with time at constant speed: $d = 60t$.

Three packs of (2 red + 4 blue) = ($3 \times 2$ red) + ($3 \times 4$ blue) = 6 red + 12 blue.

If you want to split \$7.20 equally among 3 people, you're dividing a decimal. The trick for harder problems is: if the divisor is $0.4$, multiply both numbers by 10 to get $72 \div 4 = 18$. You haven't changed the answer—just made it easier to compute.

Sharing 12 cookies equally among 4 friends—each gets 3. Or: how many groups of 4 fit into 12?

If $3 \times 4 = 12$, then $12 \div 4 = 3$. Division reverses the multiplication.

12 cookies shared among 4 kids—each gets 3. Division tells us the share size.

Imagine a movie starts at 2:15 PM and ends at 4:45 PM. Elapsed time is like counting how many minutes the movie lasted—you hop forward from the start time to the end time.

$3 + 2 = 5$ doesn't mean '3 + 2 makes 5'—it means they ARE the same.

$\frac{1}{2}$, $0.5$, and $50\%$ are equivalent—different forms, same value.

$2^3$ means $2 \times 2 \times 2 = 8$. The exponent tells you how many times to multiply.

Imagine stacking blocks in a staircase—each step is one block taller than the last. The pattern grows by a rule: $+1$ block per step. If the rule is $+3$, the staircase grows faster.

Adding 0 leaves any number unchanged; multiplying by 1 also leaves it unchanged. Both are 'do-nothing' values.

If $5 < 7$, then 5 is somewhere to the left of 7 on the number line.

Think of a number line with zero in the middle. Positive numbers go right, negative numbers go left. Adding a positive moves right; adding a negative moves left. Multiplying two negatives gives a positive because reversing a reversal brings you back to the original direction.

Rearranging an equation keeps both sides equal—equality is the invariant.

Adding 5 then subtracting 5 brings you back to where you started.

More workers means less time: if 4 workers take 6 hours, 8 workers take 3 hours.

Measuring length is like asking 'how many of this unit fit end-to-end along the object?' Lay paper clips along a pencil—the number of clips is its length in paper-clip units.

Add the same amount each step. Like paying \$10/month—increase is constant.

Long division is like distributing items into groups one place value at a time. If you have 156 stickers to share among 12 friends, you first figure out how many groups of 12 fit in 156 by working from the biggest place value down: how many 12s in 15? Then bring down the next digit and repeat.

If a toy costs \$3.75 and you hand the cashier \$5.00, making change means figuring out the gap between what you paid and what it costs—like counting up from \$3.75 to \$5.00.

Each coin is like a shortcut for counting—a nickel is a bundle of 5 pennies, a dime is 10 pennies, and a quarter is 25 pennies. Counting money is like skip counting with different-sized jumps.

Your age is monotonically increasing—it only goes up, never back down. A timer counting down is monotonically decreasing.

Imagine stacking blocks in columns for ones, tens, and hundreds. When the ones column adds up to more than 9, you bundle 10 ones into 1 ten and carry it over—just like exchanging 10 pennies for a dime.

Think of a rectangle with sides 23 and 47. You can break it into smaller rectangles: $20 \times 40$, $20 \times 7$, $3 \times 40$, and $3 \times 7$, then add the pieces. That's partial products—the standard algorithm just organizes this neatly.

If you have 4 bags with 3 apples each, multiplication tells you the total: $4 \times 3$.

A $3 \times 4$ rectangle has 12 unit squares inside—multiplication counts them.

Multiplying by 2 doubles something; by 0.5 cuts it in half; by 3 triples it.

Think of $0.3 \times 0.4$ as $\frac{3}{10} \times \frac{4}{10} = \frac{12}{100} = 0.12$. When you multiply decimals, you're working with fractions of 10, so the answer gets smaller—not bigger.

Not a straight line—it curves. Compound interest grows faster and faster.

Adding two whole numbers always gives a whole number—closed under addition.

Multiplication is repeated addition. Exponents are repeated multiplication.

Once you can handle integers and fractions separately, combine the skills: apply the sign rules you know from integers to fractions and decimals. $-\frac{2}{3} + \frac{1}{4}$ uses common denominators AND sign rules at the same time.

Without rules, $2 + 3 \times 4$ could mean 20 or 14. We agree to multiply first: 14.

Imagine voting for your favorite fruit by placing a sticker in a column. When you're done, the column with the most stickers is the winner—you can see the answer at a glance.

If 3 pizzas feed 12 people, how many feed 20? Think multiplication, not addition.

Adding 5 three times: $5+5+5 = 3 \times 5$. Multiplying 2 four times: $2 \times 2 \times 2 \times 2 = 2^4$.

If $3^2 = 9$, then $\sqrt{9} = 3$. The root asks: 'What number squared gives 9?'

Patterns are like the beat of a song—clap-snap-clap-snap repeats over and over. Once you hear the rhythm, you can predict what comes next without looking.

Skip counting is like hopping along a number line instead of walking step by step. Counting by 5s is like hopping over 4 numbers each time: $5, 10, 15, 20, \ldots$

$\sqrt{25}$ asks: what number times itself equals 25? Answer: 5.

$5^2 = 25$ is a $5 \times 5$ square's area. $5^3 = 125$ is a $5 \times 5 \times 5$ cube's volume.

If you have 5 cookies and eat 2, how many are left? You take away to find out.

How much taller is a 6-foot person than a 4-foot person? The difference is 2 feet.

$3 + 5 = 5 + 3$ shows addition is symmetric. $3 - 5 \neq 5 - 3$ shows subtraction isn't.

Tally marks are like keeping score with your fingers—every fifth mark crosses the group, making it easy to count by 5s. It's faster than writing numbers while things are happening in real time.

A clock is like a race track with two runners—the short hand (hours) moves slowly, the long hand (minutes) moves fast. When the long hand points to 12, it's exactly on the hour, like the start of a new lap.

'60 miles per hour' tells you the distance in one hour—easy to compare.

If you know the total and one group size, division tells you how many groups — the missing factor is the answer to that division.

A balance scale is like a seesaw—the heavier side goes down. To find out how heavy something is, add known weights to the other side until the scale balances perfectly.

You are decoding a story into variables, equations, and constraints.

Abstraction is the move from "three apples, three chairs, three ideas" to the concept of "three" — stripping away what varies to reveal what is shared.

Ambiguity is a fork in the road with no sign — different readers take different paths and arrive at different answers, each thinking they are right.

This is like that, so maybe what works there will work here.

What are we assuming to be true? Everything follows from these starting points.

'$P$ if and only if $Q$'—they're equivalent, true together or false together.

Cardinality answers "how many?" — count each distinct element once and you have the cardinality.

If the universal set is all students in your school and set $A$ is students who wear glasses, then the complement of $A$ is every student who does NOT wear glasses. It is everything outside the circle in a Venn diagram—the NOT operator applied to a set.

A complete system has no hidden truths that are provably beyond reach — there are no true statements you cannot prove from the axioms.

Math concepts don't exist in isolation—they're all connected.

Gateway concepts—get these and everything else becomes easier.

Once you truly understand a concept, you stop thinking through all its parts and just "see" it as one thing — like reading words instead of individual letters.

You cannot truly understand limits without understanding functions; you cannot understand derivatives without limits. Concepts form a dependency graph.

A promise or rule: if the condition holds, the consequence follows.

To enter a theme park ride, you must be tall enough AND have a valid ticket—both conditions must hold. If you are tall enough but lost your ticket, you cannot ride. A conjunction $P \wedge Q$ works the same way: it is true only when every single part is true, and false the moment any part fails.

Imagine building with a set of rules: if one rule says 'the door must be open' and another says 'the door must be closed,' the system is inconsistent and no valid state exists. Consistency matters because from a single contradiction you can logically derive any statement at all (the principle of explosion), making the entire system meaningless.

The rules of the game. What must be true? What can't happen?

Flip and negate. Always has the same truth value as the original.

One case where it fails is enough to kill a 'for all' claim.

Divide and conquer: a hard problem of size $n$ becomes $n$ easy problems. Long division, partial fractions, and integration by parts all use decomposition.

Units must balance on both sides of any physical equation — if the units do not match, the formula is wrong regardless of the numbers.

Start from what you know (the hypotheses) and chain logical steps forward until you reach what you want to prove — no detours, no tricks, just forward reasoning.

At least one must be true. Logical OR is inclusive — "P or Q or both" — unlike the exclusive everyday "either/or."

What happens at the extremes? When $x = 0$? When $x \to \infty$? When inputs are unusual?

An element is simply one item inside the collection — either it is in, or it is out. There is no "partially in."

Think of an empty box that is still a valid box—it just holds nothing. The empty set plays the same role for sets that zero plays for numbers: it is the identity element for union ($A \cup \emptyset = A$) and the annihilator for intersection ($A \cap \emptyset = \emptyset$). It is also a subset of every set, which keeps logical statements about 'all elements of $\emptyset$' vacuously true.

Treating different things as equal because they share what matters.

Error analysis asks "how wrong could my answer be?" — not just "what is my answer?" — because every measurement and approximation carries uncertainty.

Derivation: here are the steps. Explanation: here's why it makes sense.

Does this pattern work more generally? Can we remove restrictions?

What's lurking behind the scenes that we forgot to account for?

Imagine a perfect world: frictionless surfaces, perfect circles, rational actors.

Picture two overlapping circles in a Venn diagram—the intersection is only the overlapping region where both circles cover. For example, if set $A$ is students who play soccer and set $B$ is students who play piano, then $A \cap B$ is students who do both. It is the AND gate of set theory: an element must satisfy both conditions to be included.

What stays the same when things change? That's often the key.

What happens when things get really big, really small, or reach boundaries?

A logical statement is any claim that can be judged definitively as true or false — questions, commands, and paradoxes are not statements.

A good solution should be understandable by someone else, not just by you.

When a proof or solution feels 'just right'—clean, inevitable, illuminating.

Like dominoes: first one falls, and each one knocks over the next.

Building a mathematical version of reality to understand and predict.

Every algebraic step must be a valid equivalence — adding the same to both sides, multiplying by a non-zero quantity, or applying a one-to-one function preserves meaning.

A mental model is your internal simulation of how something works — good mental models make predictions that match reality; wrong ones produce systematic errors.

Looking at the same thing from different angles reveals different truths.

Flipping true to false or false to true. 'It is NOT the case that...'

The same word meaning different things in different conversations — context tells you which meaning applies, but this can trip up a reader who is new to the context.

A chain of reasoning that convinces you something MUST be true.

Assume the opposite of what you want to prove, then follow the logic to a statement that is impossibly false — proving your assumption must have been wrong.

Choose the argument tool that matches the claim type and assumptions.

It is not guessing the answer; it is proving why the answer must be true.

$\forall$ means 'for all' (everyone). $\exists$ means 'there exists' (at least one).

Computation is following a recipe; reasoning is deciding which recipe to use and why. Most math mistakes come from computing when you should be reasoning first.

After decomposing a problem, you must reassemble the pieces correctly — like completing a jigsaw puzzle, the boundary conditions between parts must match.

The same idea can be shown in multiple ways—each reveals different aspects.

Is this answer fragile, or does it survive small errors and changes?

When you double the length of a cube, its volume grows by $2^3 = 8$. Scaling laws reveal how fast quantities grow — they often explain why small and large things behave so differently.

Is this result stable, or does a tiny change blow everything up?

Think of a set as a bag that can hold anything — numbers, names, shapes — but with two strict rules: no duplicates allowed and the order in which items sit inside the bag does not matter.

The art of knowing what to throw away. Good simplification keeps the behavior that matters while discarding noise.

What does this general statement say about MY specific situation?

Seeing 'Oh, this is really a quadratic' or 'This has the same structure as...'

Every single thing in $A$ can also be found inside $B$. Think of $A$ as fitting entirely within $B$, like a small circle inside a big one.

Looks the same from different perspectives or after certain changes.

Seeing that the same mathematical structure appears in two apparently different contexts — then using what you know about one to solve the other.

List every possible combination of T/F for inputs, and compute the output.

Pour both sets into one container and remove duplicates. Everything from either pile ends up in the union — this is the OR operation for sets.

Each circle represents a set; overlapping regions show shared elements; the rectangle border is the universal set.

Going from individual values to totals, averages, or other summaries.

Start with a prior belief, then reweight it by how likely the evidence is under each hypothesis.

Same as combination count, but now viewed as a coefficient in algebraic expansions.

Flip a biased coin $n$ times—how many heads? The binomial distribution gives the probability of each count.

A summary of spread and center in one picture. Box shows the middle $50\%$.

$X$ causes $Y$ means changing $X$ will change $Y$. Not just 'they move together.'

Where is the data located? How spread out is it around that location?

Roll a single die and the outcomes are flat (uniform). But average the rolls of 30 dice and the result looks like a bell curve every time. No matter how weird the original data looks—skewed, bimodal, flat—the averages of large enough samples always settle into a normal shape. It's one of the most surprising facts in all of mathematics.

When multiple outcomes are possible and we can't control which occurs.

You expect a die to land on each face about $\frac{1}{6}$ of the time. You roll it 600 times and compare what you observed to what you expected. If the differences are small, the die is probably fair. If they're large, something is off. The chi-square statistic measures 'how far off are the observed counts from what we expected?'

Total variation in $y$ has two parts: what the regression line explains and what's left over (residual variation). If $r^2 = 0.85$, the regression line accounts for $85\%$ of why $y$ values differ from each other, and $15\%$ is unexplained. Think of $r^2$ as a report card for how well $x$ predicts $y$.

How many ways to choose a group? $\{A, B, C\} = \{C, B, A\}$.

Is A bigger/better/different than B? By how much? Is the difference real?

Single-event probability asks about one thing happening. Compound probability asks about combinations: 'What's the chance of rolling a 6 AND flipping heads?' or 'What's the chance of drawing a heart OR a face card?' The word 'and' usually means multiply; the word 'or' usually means add (but subtract the overlap).

If I know $B$ happened, what's the chance of $A$? Updates probability with new info.

You can't know the exact average height of all Americans, but after measuring 200 people you can say: 'I'm $95\%$ confident the true average is between 167 cm and 173 cm.' It's like casting a net—wider nets catch the true value more often, but narrower nets are more useful. A $95\%$ confidence level means that if you repeated this process 100 times, about 95 of those nets would contain the true value.

Do two things go up and down together? $r = +1$ means perfectly together, $r = -1$ means perfectly opposite.

Data is raw material for understanding—numbers, words, or categories we collect to answer questions.

A picture is worth a thousand numbers. Graphs reveal patterns we'd miss in tables.

The rational strategy under uncertainty is not always to pick the option with the best single outcome but the one with the best expected outcome weighted by its probability.

Knowing $A$ happened tells you something about $B$—they're connected.

If you took many measurements, where would most values fall? What's the shape?

An event is a question like 'Did I roll higher than 3?' that has yes/no answer.

Expected value is what you would "expect" on average after very many trials — not the most likely single outcome, but the center of the distribution.

You want to know if a fertilizer helps plants grow. You can't just give it to some plants and hope for the best—you need a plan: a group that gets the fertilizer, a group that doesn't (control), random assignment so the groups are fair, enough plants so one weird result doesn't fool you (replication), and ideally the person measuring growth doesn't know which group is which (blinding).

Theoretical probability is what SHOULD happen in a perfect world: a fair coin should land heads $50\%$ of the time. Experimental probability is what ACTUALLY happens when you try it: flip a coin 20 times and you might get heads 12 times ($60\%$). The more times you flip, the closer your experimental result gets to $50\%$—that's the law of large numbers in action.

Factorial counts the number of ways to arrange $n$ distinct objects in a row — for 3 items, there are $3! = 6$ possible orderings.

How many times do you have to roll a die before you get a 6? The geometric distribution answers this kind of question. Each trial is independent, and you keep going until you succeed. Most of the time it doesn't take too long, but occasionally you have an unlucky streak—that's why the distribution has a long right tail.

Group data into bins and count how many fall in each. Shows the shape of data.

Think of a courtroom trial: the null hypothesis ($H_0$) is 'innocent until proven guilty.' You look at the evidence (data) and ask: 'Is this evidence so strong that it would be very unlikely if the defendant were truly innocent?' If yes, you reject the null hypothesis. If not, you don't have enough evidence to convict—but that doesn't prove innocence.

They don't 'know about' each other. One happening tells you nothing about the other.

You computed a sample regression line with slope $b = 2.3$. But is the true population slope actually different from zero? Maybe there's really no linear relationship and you just got a slope by chance. The regression t-test asks: 'Is my sample slope far enough from zero that it's unlikely to have occurred by random variation alone?'

The IQR ignores the extreme 25% on each end, capturing only the spread of the central bulk of data — making it robust when outliers inflate the regular range.

As the number of trials grows, the sample mean converges to the true expected value — randomness averages out over many trials, making the average predictable.

You have a scatter plot with points scattered around a general trend. The LSRL is the line that gets as close as possible to all the points simultaneously—it's the 'best' straight line through the cloud. 'Best' means it minimizes the total squared prediction error.

Like stacking coins above each number — taller stacks mean that value appeared more often in the data.

When a poll says 'the approval rating is $52\%$ with a margin of error of $\pm 3\%$,' it means the true value is likely between $49\%$ and $55\%$. The margin of error is the '$\pm$' part—it tells you how much wiggle room to give the estimate. Larger samples and less variability shrink the margin of error.

Imagine redistributing all the data equally — the mean is the value each person would get if everyone shared equally. It is the balance point of the data.

Standard deviation can feel abstract with its squaring and square roots. MAD is simpler: just ask 'on average, how far is each data point from the center?' If the mean test score is 80 and the MAD is 5, a typical student scored about 5 points away from 80—some above, some below.

To measure is to quantify—turning 'how much' or 'how many' into a number.

Half the values are below, half are above. The true 'middle.'

A graph can tell any story the creator wants by choosing which data to show, where to start the axis, and how to scale the bars — visual clarity requires honest design.

The mode is the "most popular" value — if you had to guess one number and wanted to be right as often as possible, pick the mode.

Does the model's predictions match reality? Good fit = close match.

The static on a radio—it's there, but it's not the music you want to hear.

The normal distribution describes data that clusters symmetrically around the mean with a characteristic bell shape — most values are near the mean, and extreme values become rapidly less likely.

Converting to a standard reference so you can compare apples to apples.

Observational: you watch people who already smoke and compare their lung cancer rates to non-smokers. Experimental: you randomly assign people to smoke or not (unethical, but illustrates the point). The observational study might find that smokers differ from non-smokers in many ways (diet, exercise, stress)—so you can't be sure smoking caused the cancer. The experiment controls for everything else.

The weird one that doesn't fit. Is it a mistake, or something interesting?

The model memorized the training data instead of learning the underlying pattern.

The p-value answers: 'If nothing special is going on ($H_0$ is true), how surprising is my data?' A tiny p-value means the data would be very rare under $H_0$, so maybe $H_0$ is wrong. Think of it like this: you flip a coin 100 times and get 92 heads. If the coin is fair, the chance of that happening is astronomically small (tiny p-value)—so you'd conclude the coin is probably not fair.

You want to know if a tutoring program improves math scores. Instead of comparing two separate groups, you test the SAME students before and after tutoring. Each student is their own control. By looking at the difference (after $-$ before) for each student, you eliminate individual variation and focus purely on the change.

With permutations, order matters — first place and second place are different. Think of ranking students: ABC and BAC are different orderings.

You cannot taste every cookie in the bakery to check quality — you taste a few (sample) and draw conclusions about the whole batch (population).

Power is your test's ability to detect a real effect when one exists. A test with high power is like a sensitive metal detector—it won't miss a coin buried in the sand. A test with low power is like searching with your eyes—you'll miss things that are actually there. You want power to be high (typically $0.80$ or above).

Every prediction uses patterns from the past to extrapolate forward — good predictions come with explicit uncertainty bounds, not false precision.

Instead of 'Will X happen?' ask 'How likely is X?' and plan for multiple outcomes.

How confident you should be that something will happen. 0 = impossible, 1 = certain.

$P(\text{heads}) = 0.5$ means if you flip many times, about half will be heads.

Raw counts can mislead when groups differ in size — saying "100 people in City A vs. 100 in City B have a disease" ignores that City A may be ten times larger.

Q1 = 25th percentile, Q2 = median (50th), Q3 = 75th percentile.

Truly random means you can't predict the next outcome, even with complete information.

The range answers "how spread out is the data from end to end?" — it captures the total span but ignores everything in between.

A representative sample is a miniature version of the population — every relevant group is included in the right proportions so the sample mirrors the whole.

A residual is how much the model got wrong for a specific data point. Positive residual means the actual value was higher than predicted; negative means it was lower. If you plot all residuals, the pattern (or lack thereof) tells you whether the model is appropriate.

What could go wrong, how likely is it, and how bad would it be?

Before you can calculate any probability, you need the complete menu of possibilities. The sample space is that menu—like listing every face of a die or every possible hand in a card game. Missing even one outcome throws off every probability you calculate, because all probabilities must add up to exactly 1 over the full sample space.

A biased sample gives you a skewed picture of the population — like judging average student height by only surveying the basketball team.

Imagine you survey 50 random people about their height, compute the average, then repeat with a different group of 50, again and again. Each group gives a slightly different average. The pattern of all those averages forms the sampling distribution. It's like taking the temperature of a city by sending out 100 different thermometers—each reads slightly differently, but together they cluster around the truth.

You want to know the average GPA of 10,000 students. You can't ask everyone, so you pick a sample. How you pick matters enormously: grab the first 50 students you see in the cafeteria (convenience—biased), or give every student a number and use a random number generator to pick 50 (SRS—unbiased). Stratified sampling is like making sure you get proportional numbers from each grade level. Cluster sampling picks entire groups (like randomly selecting 5 classrooms and surveying everyone in them).

Zoom in on tiny differences to make them look huge, or zoom out to hide them.

Each dot is one observation — as you scan left to right, the up/down pattern of dots reveals whether the variables tend to increase or decrease together.

Is this pattern real or just coincidence? The fundamental question of data analysis.

The typical distance from the average. Low SD = clustered. High SD = spread out.

You have two separate groups—say, students taught with Method A vs Method B—and want to know if there's a real difference. Unlike paired tests where the same subjects appear in both groups, here the groups are completely independent. You compare the two sample statistics and ask: 'Is the gap between these groups larger than what random variation alone would produce?'

Imagine surveying students about their favorite sport AND their grade level. A two-way table is like a grid: grades go down the side, sports go across the top, and each cell tells you how many students are in that specific combination. The totals on the edges (margins) tell you the overall counts for each category.

Think of a medical test. Type I error: the test says you have a disease when you don't (false alarm). Type II error: the test says you're healthy when you actually have the disease (missed detection). A smoke alarm that goes off when there's no fire is a Type I error; one that stays silent during a real fire is a Type II error. You can't eliminate both—reducing one tends to increase the other.

We don't know what will happen—statistics helps us reason under this condition.

The model misses important structure—it's not learning enough.

How spread out or bunched up the data is. No variability = everyone is the same.

Another spread measure—variance $= \text{SD}^2$. Same idea, different scale.

A universal measuring stick—$z = 2$ means '2 SDs above average.'