Polymarket Bankroll Management: How Much to Bet Per Trade

Ask ten Polymarket traders how much to bet per position and you'll get ten vibes-based answers. Here's the actual rule: risk 1–5% of your bankroll per position, with an absolute cap of 10% no matter how confident you feel. On a $100 bankroll that's $2–5 positions, occasionally stretching to $10 on your single highest-conviction trade. Everything below explains why that narrow range isn't conservative paranoia — it's the difference between a trader who compounds and one who eventually blows up.

The core rule: 1–5% per trade, 10% hard cap

Position size scales with bankroll, not with confidence. A useful reference table:

• $100 bankroll: $1–5 typical position, $10 absolute maximum
• $500 bankroll: $5–25 typical position, $50 absolute maximum
• $2,000 bankroll: $20–100 typical position, $200 absolute maximum

If you're still deciding how much capital to put in at all, see our guide on how much money you need to start on Polymarket. This article picks up from there: once the money is in your account, sizing is the single decision that determines whether you're still trading in six months.

Why sizing beats picking

Picture two traders. Trader A picks winners 60% of the time — genuinely skilled. But Trader A bets 40% of the bankroll per position, confident it's a sure thing. Trader B is a more modest 55% win rate, sizing a disciplined 2% per trade. Over a long enough run, Trader B ends up richer and Trader A goes broke. That's not a quirky exception — it's streak math.

Trader A loses 40% of the time. Three losses in a row have a 0.4³ = 6.4% probability — not rare across dozens of trades, and it will recur over a career. At 40% sizing, three losses multiply the bankroll by 0.6³ = 0.216: down 78.4%, needing a 363% return to recover. One more ordinary losing streak and the account is functionally dead.

Trader B, at 55% win rate and 2% sizing, faces a five-loss streak with probability 0.45⁵ ≈ 1.8% — rare, but it will surface occasionally across hundreds of trades. At 2% sizing, five losses cost about 1 − 0.98⁵ ≈ 9.6% of the bankroll: painful, fully recoverable. The same five-loss streak sized at 20% instead costs 1 − 0.8⁵ = 67.2% — requiring a 205% return to recover, from a trader who is still, on paper, better than random.

Fractional Kelly for prediction markets

The Kelly criterion gives the bankroll fraction that maximizes long-run growth given a real, known edge. For a binary prediction market where you buy a share at price q (paying q cents on the dollar, collecting $1.00 if it resolves YES) and you believe the true probability is p, the formula simplifies to:

Worked example: you think a market is really 70% likely, but it's trading at 60¢. Full Kelly: f* = (0.70 − 0.60) ÷ (1 − 0.60) = 0.10 ÷ 0.40 = 25% of bankroll. On a $500 bankroll, that's a $125 position — on a single market, based entirely on your own probability estimate, which is exactly the problem.

Full Kelly assumes your probability estimate is exact. It never is — prediction-market edges come from research and judgment, not measured statistics, so your 70% could easily be a true 62% or 76%. Overestimate your edge even slightly and full Kelly overbets badly, because the formula punishes input error without mercy. That's why disciplined bettors run quarter-Kelly (25% of the full Kelly stake) instead of full or even half-Kelly: less theoretical growth rate, far more forgiveness for being wrong about your own edge.

Quarter-Kelly on the example above: 25% ÷ 4 = 6.25% of bankroll — on $500, a $31 position. Notice that lands above the normal 1–5% range but still under the 10% absolute cap: treat that as the ceiling reserved for your highest-conviction trades, not the default.

Market price	Your estimate	Full Kelly	Quarter-Kelly stake
50¢	55%	10.0%	2.5%
60¢	65%	12.5%	3.1%
60¢	70%	25.0%	6.3%
30¢	40%	14.3%	3.6%
80¢	85%	25.0%	6.3%
90¢	92%	20.0%	5.0%

Two patterns worth noticing. A small percentage-point edge produces a much larger Kelly stake near the extremes (90¢ vs. your 92%) than in the middle of the range, because the denominator (1 − q) shrinks as price rises. And when your estimate equals the market price — no edge — the formula correctly outputs 0%. If you can't articulate why your probability differs from the market's, Kelly is telling you not to trade at all.

Correlated positions: the hidden bankroll killer

Sizing discipline breaks down silently when positions look diversified but aren't. Five separate markets — will State A go blue, will the Senate flip, will the incumbent's approval exceed X, will the debate move the polls, will the party win the popular vote — can each be sized at a careful 3% of bankroll. But if all five resolve substantially on the same underlying event, you don't have five 3% positions. You have one 15% position wearing five different tickets, and it moves as a single block the moment that one event breaks.

This is one of the more common versions of the mistakes covered in our beginner mistakes guide: the portfolio looks spread out on the positions tab, but the actual risk is concentrated in one storyline. Real diversification means spreading capital across genuinely independent categories — politics, sports, crypto, macro data releases, entertainment — and across events with different resolution dates and different underlying drivers. Before sizing a new position, ask: if this specific event breaks against me, which of my other open positions move with it? Add up anything correlated and size the group, not the individual line.

Drawdown rules that keep you in the game

Bankroll management isn't only about entry sizing; it needs exit rules for bad stretches too.

• Monthly stop-loss: if you're down 25% of bankroll in a calendar month, stop opening new positions for the rest of the month. Close what needs closing, then go back to your trade journal and find the pattern — was it sizing, a bad run of theses, or chasing losses?
• Scaling up: only increase per-trade sizing or add fresh capital after 2–3 consecutive profitable months, measured net of fees. One good month is noise; three in a row, net of costs, is a signal your process actually has an edge.
• Taking profit off the table: once the bankroll has grown meaningfully — many traders use 2× the original deposit as the trigger — withdraw the original stake and keep trading with profits only. It doesn't change your sizing percentages, but it changes what you're risking if things turn.

Fees change the math since 2026

Sizing rules interact with cost structure, and that structure changed in 2026. Taker fees now run roughly 0.75–1.8% depending on category (geopolitics remains fee-free; sports and crypto markets tend to sit at the higher end), while makers — traders whose limit orders sit in the book instead of crossing the spread — pay nothing. The regulated US platform, live since December 2025, runs a flatter ~0.30% taker fee. Settlement itself is unaffected either way: winning shares still pay exactly $1.00, losing shares exactly $0, and Polygon gas stays under a cent per transaction.

The sizing implication is direct: a strategy that round-trips capital often — entering and exiting the same market repeatedly — pays the taker fee on both legs each time it crosses the spread. At 1.8% per side, ten round trips on the same position size quietly costs close to a fifth of that stake in fees alone, meaning your edge has to clear that bar before sizing decisions even matter. Placing limit orders instead of market orders drops that drag close to zero and should be the default for anything but genuinely time-sensitive trades — see our trading strategies guide for how to set them. High-turnover approaches need a real, provable edge; positions held to resolution barely notice the fee at all.

The one tool that makes this real: a trade journal

None of the above matters without a record, because your memory will not preserve your actual sizing discipline accurately — it remembers the wins and rationalizes the losses. Keep a simple spreadsheet, one row per trade:

Column	What it captures
Date	When you entered
Market	Which contract
Side	YES or NO
Size	Dollar amount and % of bankroll
Entry	Price paid per share
Thesis	One sentence: why you think you have an edge
Exit	Price at close or resolution
P&L	Dollar and percentage result
Lesson	What you'd do differently

Review it against the drawdown rules above once a month. If you can't fill in the “thesis” column with something more specific than “seemed likely,” that's the sizing rule telling you to sit that one out.

Bottom line

Cap every position at 1–5% of bankroll, 10% absolute maximum, even for your best idea. Size real edges with quarter-Kelly rather than full Kelly. Treat correlated positions as one position. Set a monthly stop-loss and only scale up after several clean profitable months. Route orders through limits, not market fills, so 2026's fee structure works with you instead of against you. None of this requires being right more often than you already are — it just keeps you at the table long enough for your edge to show up.