Position sizing and bankroll management for Polymarket copy trading
Position sizing for copy trading comes down to one rule: size from the drawdown, not the return. Decide the worst loss you can genuinely sit through, scale your copy allocation so the wallet's worst historical stretch stays inside it, and cap every layer below that — per trade, per market, per category. Everything else in this guide is the arithmetic that makes the rule concrete.
Why size from the drawdown?
Because losses cost more than the same-sized gains pay. The recovery math is brutally asymmetric:
| Drawdown | Gain required to get back to even |
|---|---|
| −5% | +5.3% |
| −10% | +11.1% |
| −20% | +25% |
| −30% | +42.9% |
| −40% | +66.7% |
| −50% | +100% |
A wallet whose worst stretch was −30% doesn't need to be avoided — it needs to be sized so that −30% of the copy allocation is a number you can watch happen without pulling the plug at the bottom. Quitting mid-drawdown is how copiers convert a temporary loss into a permanent one, and the drawdown you'll experience is not capped by the history: treat the historical worst as a floor, not a ceiling.
How much of your bankroll should one wallet get?
Three layers, each a hard cap the layer below can't override:
- The copy allocation. The slice of your total bankroll assigned to copy trading at all. This is the only money that should ever touch a bot, and the loss of all of it should be survivable — financially and psychologically.
- The per-wallet cap. Inside the allocation, scale each wallet so its historical max drawdown lands inside your tolerance. If you can stomach −15% of allocation on one wallet and the wallet's worst run was −30%, that wallet gets at most half the allocation. One wallet should never be the whole allocation: even the ~1.3% of wallets that pass every vetting test can decay, and farming findings can surface on a previously clean record.
- The per-trade / per-market cap. Proportional mirroring means the leader's conviction sizes your trades. Cap it: a fixed maximum per trade and per market protects you from the leader's worst idea — and from the oversized-winner pattern where one advertised lottery ticket carries a record.
A useful discipline borrowed from Kelly-style thinking, without the false precision: betting some fraction of the mathematically "optimal" size is robust to your edge estimate being wrong — and your edge estimate is always less certain than it feels. Estimated from someone else's history, through copying costs you can only model, it deserves a deep discount. Half-Kelly is aggressive here; most copiers should size far below it.
Size from the realistic number, never the headline
Whatever return figure you size against, make it the post-cost one. A copier typically keeps 20–40% less than the leader's headline after fees, slippage, and latency — and bot fees of 1% per side take their slice on top. Sizing against the headline systematically over-allocates to wallets whose copyable edge is thinner than advertised. The Copy Simulator exists for exactly this: it replays the wallet's actual fills at your capital with modelled costs, which makes the drawdown and return numbers you're sizing from honest ones.
Capital lockup is part of sizing too: positions stay committed until markets resolve, and resolution runs on the oracle's clock, not yours. An allocation that's fully deployed is also fully illiquid — keep slack for the drawdown you're planning to sit through.
Category budgets and stop rules
Budget by category. Wallets are rarely sharp everywhere — edge concentrates. If the wallet's profit lives in sports, a category cap stops its politics experiments from spending your allocation. Polymarket's own copy-trading guidance recommends category matching; we'd phrase it harder: copy the edge, not the whole person.
Write your stop rules before you start. Three triggers, decided in advance, in writing:
- A farming finding on the wallet — exit, not resize. The record itself is now suspect.
- A drawdown beyond the historical worst — the wallet is outside everything your sizing was based on.
- Score decay — a deteriorating Copy Score is the model's way of saying the original case is eroding. Alerts automate the trigger; the rule only works if it was written down before the money moved.
The honest footnote: sizing controls how much you lose when wrong — it cannot make a negative-edge wallet profitable, and the median active Polymarket wallet has one. Selection first, sizing second.
CopyGrade is analysis-only — it never executes trades or holds funds. Not financial advice.