The thesis is the easy part
Most prediction-market traders spend 90% of their effort on edge ("Trump is +220 but should be +180, that's value") and 10% on execution ("how do I size this and when do I get out"). For serious capital, those weights are backwards.
The fastest way to wreck a Polymarket bankroll isn't being wrong on a thesis. It's being right but oversized, or right but with no exit plan, or right but holding through a 60% drawdown because "the market'll come back." All three failure modes are about risk management, not about edge. Sort risk management first; thesis-work returns are bounded by your survival.
Three rules that do most of the work
These three rules — if you actually follow them — get you 80% of the way to professional-grade risk management. The remaining 20% is detail and personal preference.
Rule 1: No single position is more than 5% of your bankroll.
If you have $10k on Polymarket, no individual position is bigger than $500. This bounds your maximum loss on any one event at 5% of total capital (because a binary you bought at $0.50 can resolve at $0.00). With a stop-loss at, say, $0.20, your effective single-trade risk drops to ~3%.
The 5% number isn't sacred. Some traders use 2% (very conservative), some use 10% (aggressive). The principle that matters: pick a number in advance, write it down, and don't override it because you're "really sure" about a specific trade. The day you talk yourself into a 15% position is the day you lose 15% on a binary that should have been mispriced and wasn't.
Rule 2: Every position has a stop, set at entry.
Before you buy the position, decide the price at which you'll exit if you're wrong. Set the stop immediately, not later. Tools like UnusualBets' TP/SL bracket let you arm the stop the moment you take the position, so the worst-case outcome is bounded the moment the trade is on.
A stop you "remember to set later" is a stop you don't have. A stop you'd "manually exit if it gets that low" is a stop you'll second-guess at the moment of truth ("let me just give it another day"). The only stop that actually protects you is the one that fires without your involvement.
Rule 3: No more than 25% of your bankroll on correlated bets.
If you have a thesis like "Republicans will sweep the 2026 midterms" and you express it across the House, Senate, and Governor markets, those are one thesis dressed up as three positions. Size the combined exposure at no more than 25% of your bankroll. The single-position 5% rule doesn't protect you when ten positions all move together on the same political-news headline.
The hardest version of this rule is recognizing correlation that isn't obvious. "Will Bitcoin hit $200k by end of year" and "Will the S&P 500 hit ATH" share more risk than they look. Your job is to map the underlying risks, not the surface labels.
Concrete numbers, for a $10k Polymarket bankroll
Let's make the abstraction concrete. With $10k on Polymarket and the rules above:
- Maximum single-position size: $500.
- Maximum loss per single position: ~$300 (with a stop at 60% of entry).
- Maximum loss per thesis cluster: ~$1,500 (with stops on each position).
- Maximum total open exposure: $5,000 (50% of bankroll, leaving half in reserve).
What you can do under those constraints:
- 10 simultaneous independent positions of $500 each.
- Or 4 thesis clusters of $1,250 (3 positions of ~$400 inside each).
- Or some mix.
What you can't do:
- One $2,000 conviction trade because you're "sure this time."
- Three $1,000 positions all expressing the same trade.
- Fifteen $500 positions because the dashboard is showing you a lot of mispriced markets and you can't help yourself.
These constraints feel restrictive when you start. After two months of trading inside them, they feel obviously correct.
Kelly is for textbook problems
Some traders advocate Kelly sizing — pick your size based on the edge and the odds. The textbook version is not the right approach on Polymarket. Three reasons.
First, your estimate of edge is noisier than you think. Kelly assumes you know the probabilities. You don't. You have a model that estimates them with significant uncertainty. Sizing as if you know the probabilities exactly overstates your edge and oversizes your positions. The standard fix — "fractional Kelly", typically 1/2 or 1/4 — is just Kelly with a fudge factor. The fudge factor is the admission that the model is wrong.
Second, resolution risk is real. Polymarket markets resolve based on real-world events. Edge cases happen: the market resolves ambiguously, the resolution date slips, the criteria are interpreted unexpectedly. Kelly doesn't account for this; flat sizing does.
Third, sleep matters. A position size that maximizes your expected log-wealth but keeps you up at night will cause you to intervene at exactly the wrong time. Sizing under what you can "live with" beats sizing at the math-optimal point.
Use the 5% rule. Use a stop. Move on.
What stop-loss prices actually look like
A common mistake is setting stops too tight. A YES share at $0.45 in a market that moves 10¢ on noise is going to oscillate. A stop at $0.40 will trigger on a normal day. The stop should be set at the point where your thesis is wrong, not at the point where the market is moving against you.
Concrete heuristic: in a typical liquid Polymarket binary, the daily range is 3-8%. Set your stop outside the noise band. For a $0.45 entry, a stop at $0.30 (~33% adverse move) is reasonable; $0.40 is too tight.
The TP is more flexible. Set it where you'd be happy to take the trade off — usually 50-70% of the way from entry to $1.00 for a buy-low thesis. Tighter TPs win more often but smaller; wider TPs win less often but bigger. Match this to your overall trade frequency: fast traders should set tighter TPs and harvest small wins; position traders should set wider TPs and accept lower hit rates.
What an automated worker doesn't fix
Two things to be honest about.
TP/SL is best-effort. Even with sub-second triggers and post-fill polling, a stop can miss. The market can resolve before the worker can fill. The book can gap below your stop. You should size each position assuming the stop might fail — which is to say, you should size each position with the same 5% rule you'd use without a stop at all.
Risk management is not a substitute for edge. The 5%-position rule keeps you in the game. It doesn't make you profitable. If your thesis-work has no edge, the best risk-management framework on the planet will produce a slow, controlled loss. Risk management is the foundation; edge is the building. You need both.
Putting it together
- Decide your bankroll. Put it on Polymarket. Don't add to it more often than every 90 days.
- Set the 5% per-position cap. Refuse to override it.
- For every position, set the TP and SL at the moment of entry. Use a bracket tool so the stop fires whether or not you're online.
- Track your win rate and average win/loss size weekly. If you don't track, you're not trading — you're gambling.
- Sleep.
Set your first bracket — it's the first practical step.