Blog · Prediction markets
Expected Value (EV): The Only Formula You Need Before Predicting
You can forget almost all the jargon in prediction markets, but one formula is non-negotiable. Expected value (EV) is the number that tells you whether a bet makes mathematical sense before you place it. It doesn't predict the next outcome (nothing does) but it answers the only question that matters in the long run: if I made this exact bet a thousand times, would I end up richer or poorer?
Whoever trades without computing EV is gambling blind, no matter how much "analysis" they think they are doing. Whoever computes it has, at minimum, a compass. And the formula fits on one line.
The formula, in one line
The expected value of a bet is the sum of each possible outcome multiplied by its probability:
EV = (probability of winning × what you win) − (probability of losing × what you lose)
If the result is positive, the bet makes money on average. If it is negative, it loses money. That's it. The difficulty is never the formula, it is estimating that probability well, and we will come back to that.
Example 1: the coin flip with a prize
Start with the cleanest possible case. I offer you this deal: you pay $0.60 to play; heads pays you $1, tails pays nothing. The coin is fair: 50% each side.
- EV = (0.50 × $1) − $0.60 = $0.50 − $0.60 = −$0.10
Every round costs you, on average, 10 cents. You might win three in a row (luck exists) but play a thousand rounds and you will lose about $100. Now the opposite deal: same game, but you pay $0.40 to play.
- EV = (0.50 × $1) − $0.40 = +$0.10
Same coin, same randomness, yet one version bankrupts you and the other enriches you over time. The difference was never luck: it was the price. That is the entire lesson of EV, told in two coins.
Example 2: a prediction market
On Polymarket the arithmetic gets even friendlier, because every winning share pays exactly $1 and the share price (between $0 and $1) is the probability the market assigns to the event, its implied probability. That collapses the formula into a subtraction:
EV per share = your estimated probability − share price
Take an election market. After studying polls, base rates and context, you estimate the candidate has a 60% real chance. The YES share trades at $0.45:
- EV = 0.60 − 0.45 = +$0.15 per share. Every dollar invested returns, on average, 33% (0.15/0.45). A value bet.
- If your estimate were 40%: EV = 0.40 − 0.45 = −$0.05. The market is more optimistic than you: don't buy, however compelling the narrative feels.
Notice the nuance that changes everything: the decision doesn't depend on whether you think the event will happen, it depends on whether your probability beats the price. You can buy YES on something that probably won't happen (you estimate 30%, it trades at $0.18) and it can be an excellent purchase. Beginners bet on outcomes; survivors bet on mispriced probabilities.
Recorded straight off the map, no sound.
Positive EV does not mean winning this time
This is where most accounts die: confusing the average with the next play. An EV of +$0.15 promises nothing about this market; it promises something about a hundred markets like it. In between lies variance, perfectly normal losing streaks that break anyone who bet too much per play.
That is why EV never travels alone; its inseparable partner is bankroll management: deciding how much to risk per position so you survive the variance long enough for the average to assert itself. The Kelly criterion is the classic framework for sizing positions according to your edge. And don't forget to subtract fees from the margin: they have changed over time (check the official page), but as a general rule limit orders pay less than market orders, on thin edges, that can be the difference between positive and negative EV.
The real problem: where does your probability come from?
The formula is trivial; your estimate is the whole business. If your 60% is really a 45% dressed up as intuition, the EV you computed is fiction. Getting better at estimating is a craft in its own right (superforecasting practically turns it into a sport) built on primary data, base rates and measured calibration, not headlines.
It is also built on tools. For events anchored to a place (elections, conflicts, weather, crises) the geolocated predictions on Argos Atlas pin each Polymarket market to the exact location of the event, alongside the ARGOS signal, which compares the market's price against an independent probability estimate built from the map's own data. In this article's terms: the signal hunts for exactly what the formula describes, events where price and estimated probability diverge, i.e. positive-EV candidates, already flagged on the map.
With the estimate made and the EV computed, the rest is execution: open the market on Polymarket, look at the live price with its spread, and decide whether the edge survives the frictions.
The discipline the formula forces on you
The most valuable thing about EV is not the number, it is the habit. Computing it before every position forces you to state your probability explicitly (so you can audit it when you're wrong), forbids buying "because it's obviously going to happen" without looking at the price, and gives you a cold criterion for walking past 95% of markets, which is what anyone who lasts in this game does. The formula fits on one line. The discipline of applying it fills a career.
This content is informational and does not constitute financial advice or an investment recommendation.
Frequently asked questions
What does it mean for a bet to have positive EV?
That, repeated many times, the bet makes money on average. You compute it by multiplying each possible outcome by its probability and adding them up. In a prediction market it is very direct: if your estimated probability is higher than the share price, the EV is positive.
Does positive EV guarantee I will win?
No, and this is the hardest part to internalize. EV describes the long-run average, not the next outcome. You can make an excellent bet and lose it, or a terrible one and win it. That is why EV always travels with bankroll management: you need to survive the streaks long enough for the average to show up.
How do I calculate EV on Polymarket?
With a subtraction: your estimated probability minus the share price. Since every winning share pays exactly $1, if you estimate 60% and the share costs $0.45, your EV is +$0.15 per share. If you estimate 40% at that same price, it is −$0.05: don't buy.
Does EV account for fees?
The basic formula doesn't, and you should factor them in because they shrink the real margin. Fees have changed over time, so check the platform's official page; as a rule of thumb, limit orders usually pay lower fees than market orders.
- The 10 Beginner Mistakes in Prediction Markets
- Limit Orders on Polymarket: The Quiet Way to Pay Less
- How Many People Actually Make Money on Polymarket? The Real Data
- Selling Before Resolution: Taking Profits in Prediction Markets
- Weather Prediction Markets: Trading the Degrees
- Crypto Prediction Markets: Price Exposure Without Buying the Asset