Kelly Calculator
Optimal position sizing from edge, bankroll and odds
The Kelly Calculator answers the single hardest question in trading: how much should I actually bet? You can have a perfect read on a market, but if you size positions by gut feel you will either leave money on the table or blow up your bankroll on a bad streak. The Kelly Criterion turns your edge into a mathematically optimal stake — and Predite wraps it in a calculator that runs live as you type, with conservative defaults and a hard fail-safe built in.
Find it at /dashboard/kelly, listed as Position Sizer (Kelly) under *Analysis* in the sidebar. It is also on the Tools hub and in the Cmd/Ctrl+K search.
What It Does
The calculator takes your view of a market, the price the market is offering, and your available capital, then tells you the dollar size that maximizes long-run bankroll growth. It is a standalone tool — nothing is stored, nothing is sent to an exchange, and every keystroke recomputes instantly. Use it to sanity-check a position before you place it manually, or to size trades you found in the EV Scanner.
Alongside the recommended size, it surfaces the numbers that actually matter for a decision:
- •Edge in percentage points (pp) — your probability minus the market price
- •EV per trade in dollars — your expected profit on this single bet
- •If you win / if you lose — the exact payout and the exact loss
- •Expected growth per trade — the geometric (compounded) growth rate, which is what Kelly is really optimizing
- •A what-if table comparing Quarter, Half, three-quarter, and Full Kelly side by side
The Four Inputs
- Bankroll ($) — the total capital you are willing to allocate to this market or category, *not* your entire net worth. If you have $50,000 in your account but only ever risk $10,000 on politics markets, enter $10,000. Kelly sizes relative to this number, so getting it right is the most important input.
- Your Probability (%) — your honest estimate of the true chance the YES outcome happens. This is the subjective heart of the calculation. Pull it from your own research, historical base rates, a model, or Predite's AI Probability estimate. If you genuinely have no view, Kelly has nothing to work with.
- Market Price (¢) — the current price of the YES contract, which on Polymarket and Kalshi is just the market's implied probability. A contract trading at 50¢ means the market thinks it's a coin flip. Your edge is the gap between your number and this one.
- Kelly Fraction — a slider from aggressive (Full Kelly) down to conservative. It defaults to ¼ (25%), the recommended setting for almost everyone. More on why below.
The Formula
The textbook Kelly formula is:
f\* = (bp − q) / b
where:
- •b = the payout odds (net winnings per dollar staked). Buy YES at price P and you risk P to win (1 − P), so b = (1 − P) / P.
- •p = your probability of winning (your subjective estimate).
- •q = your probability of losing = 1 − p.
For a binary prediction-market contract this collapses to a much cleaner form:
f\* = (your probability − market price) / (1 − market price)
That fraction is the percentage of your bankroll Full Kelly says to wager.
Worked Example
You think a market is 60% likely to resolve YES, but it's trading at 50¢, and your bankroll is $10,000.
- •Edge = 60% − 50% = 10pp
- •Full Kelly = (0.60 − 0.50) / (1 − 0.50) = 0.10 / 0.50 = 0.20 → 20% of bankroll = $2,000
- •At the default ¼ Kelly, recommended size = 0.20 × 0.25 = 0.05 → 5% = $500
The payout ratio at 50¢ is 1:1, so if you win you make $500 and if you lose you are down $500 — but because your probability beats the price, the EV per trade is positive: 0.60 × $500 − 0.40 × $500 = $100 on that $500 stake. (Stake the full Kelly $2,000 instead and the same arithmetic gives $400 — EV scales with the stake, which is exactly why the fraction is a risk choice and not a free lunch.) The calculator shows all of this without you doing any arithmetic.
A second example shows why price matters as much as edge. If your probability is 50% but the contract trades at just 20¢, Full Kelly is (0.50 − 0.20) / 0.80 = 0.375 — a much larger position, because the 4:1 payout rewards you heavily when you're right. Cheap contracts with real edge get sized up; expensive ones get sized down.
Why Fractional Kelly (¼) Is the Default
Full Kelly is *growth-optimal in theory*, but the theory assumes you know the true probability exactly. In the real world your probability estimate is, at best, a good guess — and Full Kelly is brutally unforgiving of overconfidence. It routinely produces drawdowns over 50%, and a single overestimate of your edge can push you toward ruin.
Fractional Kelly fixes this. By staking a *fraction* of the Full Kelly amount, you trade a little expected growth for a large reduction in volatility:
- •¼ Kelly keeps roughly 75% of Full Kelly's growth rate while cutting variance to about 1/16 of Full Kelly.
- •That asymmetry — almost all the upside, a fraction of the swings — is why ¼ Kelly is the standard for serious bankroll management.
The slider lets you move toward Half Kelly if you have strong, well-calibrated estimates and a higher risk tolerance, but we recommend most traders never go above Half. The what-if table makes the trade-off concrete: you can see the dollar size and growth rate at each fraction in one glance, with your current setting highlighted.
The 25% Bankroll Cap (Fail-Safe)
On top of fractional Kelly, the calculator enforces a hard ceiling: no recommendation will ever exceed 25% of your bankroll, regardless of what the math says.
Why? Because when your estimated edge is huge — say you mark something at 90% that's trading at 30¢ — raw Full Kelly would tell you to bet over 85% of your bankroll. That is exactly the situation where you are most likely to be *wrong*, because edges that large usually mean you've made an error, the market knows something you don't, or you're fooling yourself. Sizing 85% into a single market is how accounts get wiped out.
When the cap kicks in, you'll see an amber "capped" warning and a note explaining that the raw Kelly number was clamped to 25% as a defensive measure. Treat that warning as a prompt to double-check your probability — if Kelly wants to bet the farm, your inputs deserve a second look.
The No-Edge Guard
If your probability is equal to or below the market price, there is no edge, and the calculator refuses to recommend a bet. Instead of a size, you'll get a clear "No edge — don't bet" message.
This is the discipline Kelly enforces and the lesson most traders learn too late: betting without an informational advantage is negative-EV. If you think a market is 45% and it's priced at 50¢, the market is offering you a fair-or-worse price. Walk away. The tool will not let you talk yourself into it.
Step-by-Step: Sizing a Trade
- Open /dashboard/kelly.
- Enter your bankroll for this market or category — the real number you'd risk here, not your account total.
- Type your probability as a percentage. Be honest and, ideally, calibrated.
- Enter the current YES price in cents, straight from Polymarket or Kalshi.
- Read the recommended size — it updates live. The default ¼ Kelly is already applied.
- Check the edge, EV, and growth stats to confirm the trade is worth taking.
- If you see the capped warning, revisit your probability before proceeding.
- Glance at the what-if table to understand how much more (or less) risk other fractions imply.
- Place the trade manually on the platform — or, on the Bot plan, route it through live trading or a bot.
Tips and Gotchas
- •Calibrate before you trust your inputs. Kelly is only as good as your probability. Use Paper Trading to track whether your "70%" calls actually win 70% of the time. Until you're calibrated, lean toward smaller fractions.
- •**Bankroll means *risk* capital, per arena.** If you trade politics and sports separately, run Kelly per category with separate bankrolls so one segment can't dominate your sizing.
- •Re-run when the price moves. Your edge shrinks as the market converges toward your estimate. A trade worth 5% at 50¢ may be worth nothing at 58¢.
- •Round down, not up. If Kelly says $512, betting $500 costs you almost nothing in growth and buys you a margin of safety.
- •Mind liquidity and slippage. Kelly assumes you can fill at the quoted price. On thin order books, a large stake moves the price against you — pair the calculator with the Slippage Simulator before committing real size.
Limitations
- •Estimate accuracy is everything. Kelly *amplifies* errors. Overstate your edge and you'll systematically oversize; the cap and fractional default exist precisely because human estimates drift optimistic.
- •It ignores correlation. The formula sizes one bet in isolation. If you hold five positions that all move together (e.g., five markets tied to the same election), your *true* combined exposure is far larger than any single Kelly number suggests. Diversify across uncorrelated markets and scale each position down accordingly.
- •No outcome is guaranteed. Even a string of genuinely +EV trades can lose — variance is real, and losing runs happen. Kelly optimizes the long run, not any single bet.
- •This is not financial advice. Prediction-market positions carry full risk of loss. The calculator is a sizing aid, not a recommendation to trade.
Plan Requirements
The Kelly Calculator is included on the Pro ($59/mo) and Bot ($99/mo) plans. Starter ($29/mo) does not include it — upgrade to Pro to unlock the calculator along with the whale tracker, arbitrage scanner, backtesting engine, and AI tools. Live order execution (placing the sized trade automatically via Polymarket CLOB) is a separate Bot plan feature; on Pro you size with Kelly and place the order yourself.
How It Connects to Other Features
Kelly sits in the middle of a complete workflow. Find an opportunity in the EV Scanner, pull a probability from AI Probability, size it here, pressure-test the fill with the Slippage Simulator, place it (manually on Pro, automated on Bot), and log the result in your Trade Journal so your future probability estimates get sharper. Bot-plan traders can wire Kelly-style sizing directly into automated strategies.