Position Sizing on Polymarket: How Much to Put on a Window

Most trading advice stops at finding an edge, as if the rest is bookkeeping. It is not. Edge decides whether to trade at all; size decides whether you are still trading in three months. Two people with the identical edge and the identical entries can finish a year one up and one broke, purely on how much they put on each window. Here is how sizing works on short binary markets, where the swings are fast and the temptation is loud.

Why binary outcomes punish big size

An Up/Down window pays $1 or nothing. There is no partial outcome, no stop-loss that saves you mid-window - each position is fully resolved, win or lose. That makes the variance brutal even when you are right on average: a genuine 55% edge still produces runs of six or eight consecutive losses over a long enough sample. If each of those positions is 20% of your bankroll, the run ends you before the edge ever pays. If each is 2%, it is an unpleasant week.

This is the part that surprises people: ruin is not caused by being wrong, it is caused by being wrong while oversized. The streaks data makes the point concretely - a fair coin clusters, and so does a slightly favourable one.

Fixed fraction: boring and correct

The default that works is a small constant percentage of bankroll on every position - commonly 1-2% on fast windows. It scales down automatically after losses and up after gains, and it removes the decision you are worst at making: how much to risk while emotional. The precise figure matters far less than holding it steady.

What about sizing up when a trade looks especially good? Only if "especially good" is a measured number. If the gap between price and the base rate on the probabilities page is twice as wide as usual, a larger size is defensible. If it is a feeling, sizing by it puts your biggest money behind your least examined judgement - the exact opposite of what you want.

What Kelly actually says

For a share bought at price p when you believe the true probability is q, the Kelly-optimal fraction is (q - p) / (1 - p). Buy at 50¢ believing 55%, and the formula says 10% of bankroll. Almost nobody should trade that size, and the reason is important: the formula assumes you know q. You do not - you have an estimate from a finite sample, and if your true edge is smaller than you think, full Kelly is catastrophically aggressive. Serious traders use a fraction of it, often a quarter, which on that example lands right back at about 2.5%.

Liquidity is the real cap

There is a ceiling your bankroll does not control. On a thin book your own order eats through the available offers and pushes the price against you, so a large size arrives at a worse average entry than the screen promised - and as reading the odds shows, a worse entry means a higher win rate required. Combined with the fees on short windows, detailed in fees, deposits and withdrawals, oversizing can erase an edge through execution alone. Check the depth on the live windows before deciding size, not after.

Rules worth keeping

1. Fixed small fraction per position, held constant through winning and losing runs.
2. Never size up to recover - doubling after losses is the fastest documented route to zero.
3. Cap your day. A hard daily loss limit stops a bad session becoming a bad month.
4. Size to the book, not the bankroll, whenever depth is thin.
5. Audit yourself - your own address is public, so use the wallet check method on it and see what your sizing actually produced.

If you mirror someone else, the same rules apply to your size, not theirs - copy trading covers why. And the obvious but necessary point: trade only money you can lose entirely without it changing your life. Sizing is the discipline that keeps a real edge from being wasted, but no sizing rule turns a negative edge positive.