A fair die pays you, in dollars, whatever number it shows. Roll it 600 times and each face comes up about 100 times, so you collect about 100×(1+2+3+4+5+6)=2100 dollars, or 6002100=3.5 dollars per roll. That 3.5 is the expected value. You can skip the 600 rolls: E=1⋅61+2⋅61+⋯+6⋅61=3.5, each value weighted by how often it happens.
The weighting matters when outcomes are not equally likely. A prize wheel pays $1 with probability 0.7, $10 with probability 0.2 and $50 with probability 0.1. A plain average of the prizes, 31+10+50≈20.33, treats the rare $50 as if it came up a third of the time. The weighted sum, 1(0.7)+10(0.2)+50(0.1)=0.70+2+5=7.70, is the true long-run average per spin.
The expected value is neither the most likely outcome nor what happens on one play. On that wheel the most likely result is $1, which comes up 70% of the time, and no spin ever pays exactly $7.70. If a spin costs $8, you lose 30 cents per spin on average, so over 1,000 spins expect to be down about $300, even though some spins pay $50.
Think of it like this. The expected value is the fair price of a bet: the ticket price at which neither the player nor the house comes out ahead over many plays. Pay more than the expected value and you lose in the long run; pay less and you win. The picture says nothing about one play, which never pays exactly its average.
Core idea. Expected value is each outcome's value weighted by its probability, summed — the average you'd settle on over many, many trials.