AI Hints
Statistics & Math for ML · Lesson 38 of 138

Expected Value

The long-run average.

⏱ 2 min read▶ Video at 43:16🎓 Free
Watch this lesson in the Statistics & Math for ML courseWatch on YouTube ↗Full Statistics & Math for ML course

If you rolled a die forever, what would the average roll be? It's not a value you can actually roll. It's three and a half. That's the expected value. The long run average of a random variable.

Expected Value
From the video: Expected Value
  • Roll a die forever
  • Average roll?
  • 3.5, the expected value

Weighted by probability

The expected value is each value times its probability, all added up. It's a weighted mean, with probabilities as weights. Let's check it by rolling a million dice. Run it. The formula says three point five, and the simulation agrees.

expected.pypython
import numpy as np

faces = np.arange(1, 7)
probs = np.full(6, 1 / 6)
print("expected:", (faces * probs).sum())

rng = np.random.default_rng(0)
print("simulated:", rng.integers(1, 7, 1_000_000).mean().round(3))
Output
expected: 3.5
simulated: 3.502

Is this game worth playing?

A game costs one dollar to play. You roll a die, and if it's a six, you win four dollars. Should you play? The expected winnings are four dollars times a sixth, minus the dollar you paid. Run it. On average, you lose about thirty three cents every game.

game.pypython
cost = 1
prize = 4
p_win = 1 / 6

expected = prize * p_win - cost
print(f"expected value: {expected:.2f} dollars per game")
Output
expected value: -0.33 dollars per game

Summary

The expected value is the probability weighted average, what you'd get on average in the long run. Next, the binomial distribution.

Key takeaways
  • E[X] = Σ value × prob
  • A weighted mean
  • Negative EV: don't play
Next up: The Binomial Distribution — counting successes