Kelly Criterion on Polymarket 2026

Using the Kelly Criterion on Polymarket — prediction market trading dashboard — Photo via Picsum

Polymarket is where the Kelly Criterion finally feels at home. Sportsbooks pay out in awkward decimal odds, casinos give you negative expected value to start with, and stock markets don't have clean win/loss outcomes. Polymarket gives you a binary YES/NO market priced between 0 and 1 — exactly the textbook setup Kelly was written for. This post walks through how to use the Kelly Criterion on Polymarket, end-to-end, in 2026.

Step 1: Find a market where you actually have an edge

Kelly only works if you're right more often than the market. "Edge" means your estimate of the true probability differs from the market price by enough to matter — typically 5 percentage points or more after frictions.

Where edges actually come from on Polymarket:

Domain knowledge the market underweights — you watch a niche sport, follow a specific election cycle, or track a tech company's product cadence
Time-zone arbitrage — you read overnight news from a region the typical Polymarket user is asleep for
Resolution-source nuance — you've actually read the market rules and noticed the price doesn't match what UMA will resolve to

If your edge is "I just feel like it," stop. Kelly applied to a non-edge is a bankroll incinerator.

Step 2: Pin down your probability estimate

Write down a number. Not a feeling, a number. "I think YES is 65%." Force yourself to be specific because the math doesn't accept vibes.

Now apply some honesty: how confident are you in that number? A good calculator turns this into a credible interval using a Beta posterior. You give the system a confidence level (Low ≈ 10 observations of evidence, Medium ≈ 30, High ≈ 100) and it constructs a distribution around your point estimate.

The reason to do this is brutal: full Kelly sized off a wrong point estimate is the fastest way to lose money on Polymarket. Sizing off a conservative bound — say, the 25th percentile of your belief distribution — automatically protects you from the cases where you were just optimistic.

Step 3: Look at the order book, not just the price

The displayed YES price (the "ask") only fills the top of the order book. If you stake $1,000 in a market with $500 of size at the top of book, you fill half at the top, then walk into deeper levels and pay 1¢ to 2¢ more per unit.

This matters because your effective entry price is what Kelly cares about, not the screen price. A 5% edge at the screen evaporates fast if the average fill is 2 cents worse than displayed.

Practical workflow:

Open the market on Polymarket and check the order book
Note: top-of-book size, then 2-3 levels of additional depth (offset and dollar size)
Plug those into a Kelly calculator that supports walk-the-book
The solver iterates until it finds the bet size where the recommended Kelly fraction is consistent with the actual blended fill price

Step 4: Apply fractional Kelly

Take the Kelly fraction the formula spits out and multiply it by 0.25 (or 0.50 if you have very calibrated edges). This is your target bet size. Drawdowns will be a fraction of what full Kelly would give you, and your growth rate is most of the way to optimal.

For a worked example: $10,000 bankroll, 60% belief, displayed ask 50¢, with a textbook Kelly of 20% and a quarter Kelly factor → bet $500. That's the size. Not $2,000. Not the maximum you can afford to lose. The calculated number.

Step 5: Place the bet, log the outcome, repeat

This is where most people drop the ball. Kelly is a long-run discipline. You need to track every bet — your estimated probability, the market price, the size, the eventual outcome — so you can periodically check whether your edge is real or imaginary.

If your tracked Brier score is worse than the market's, you have negative edge and Kelly is sizing you into the ground. The fastest way to figure this out is to compare yourself against a baseline. Our public ProbBrain accuracy dashboard does this for our own signals — every published call, every resolution, the running accuracy.

Pros and cons of using Kelly on Polymarket

Pros

Polymarket's binary structure maps cleanly to textbook Kelly
You can stake whatever size you want — no minimum bet barriers
Resolution dates are explicit, so you can plan capital lockup
USDC settlement is fast — winnings are reusable within a week of resolution

Cons

Order books can be thin on niche markets — walk-the-book is real
Some markets have ambiguous resolution rules — read carefully before sizing up
Trading fees + spread eat into edge on small markets
Kelly assumes independent bets; correlated positions (e.g., multiple Iran-related markets) need adjustment

Frequently asked questions

How big should my first Kelly bet on Polymarket be?

Smaller than you think. Use quarter Kelly with a conservative belief estimate (Low confidence, 75% credible bound) until you've tracked 30-50 bets and confirmed your edge is real. Once your hit rate is calibrated, you can dial up the fraction.

Should I use Kelly for short-resolution Polymarket markets?

Yes, but be aware of friction: tight markets have small edges that are easily eroded by fees and slippage. The lower your edge, the more sensitive your sizing is to those frictions — a calculator that walks the book becomes essential.

What's the biggest mistake people make using Kelly on Polymarket?

Sizing off the screen ask without accounting for order book depth. The displayed price is for the top slice only. If you're sizing into thin liquidity, your real Kelly bet is smaller than your spreadsheet thinks.

Bottom line

Polymarket is the cleanest playground for Kelly Criterion bet sizing in 2026 — binary outcomes, transparent prices, USDC settlement. The wrinkle is that real-world frictions like order book depth and estimate noise need to be in your sizing math, not just the textbook formula. A serious Kelly Criterion calculator built for Polymarket handles those for you.

Size Your Polymarket Bets Like a Pro

Try our free Kelly Criterion calculator — Bayesian shrinkage, walk-the-book sizing, 10k-trial Monte Carlo. Built for serious Polymarket traders. Auto-fills from live ProbBrain signals.

🎯 Open the Kelly Calculator →

18+ only. Gamble responsibly.

Search This Blog

Crypto Bonus Hub