Log in

Our Polymarket Trader Ranking Methodology

Most leaderboards rank by raw win rate or a single eye-catching return. That is exactly how you end up copying a wallet that went 1-for-1 on one lucky bet. The Polymarket trader ranking methodology behind Polyspect is built to do the opposite: surface traders whose results are likely to repeat. Here is precisely how Polyspect scores traders, with no black box.

Step 1: Public data, raw and reproducible

Everything starts with Polymarket's public APIs. We pull the markets in scope and download the raw individual trades for each one. We store those trades unmodified, then reconstruct each wallet's net position per market. Because we keep the raw fills, the entire score can be re-run from scratch and lands byte-for-byte identical every time. No hand-tuning, no cherry-picking.

We settle every resolved position against the actual oracle outcome: the winning side pays 1, the losing side pays 0. That gives us a wallet's realized PnL the same way the market itself settled it.

Step 2: Gate the cohort first

Before anything is ranked, a wallet has to clear three thresholds. This filtering happens up front so noise never reaches the leaderboard:

A trader who clears the gate is genuinely active. One who does not never competes for a ranking, no matter how flattering their win rate looks. The same gate drives our trader leaderboard, and you can read the full rationale in plain language on what makes a trader profitable.

Step 3: Percentile-rank each metric

For the gated cohort, we compute four core metrics per wallet and then percentile-rank each one against everyone else, rather than scaling by raw magnitude:

  1. Wilson-adjusted win rate — win rate with a small-sample penalty (more below).
  2. ROI — return on the capital actually deployed.
  3. Realized PnL — dollars actually won, after settlement.
  4. Turnover — a capital term, so consistent size is rewarded over tiny lucky bets.

Percentile ranking depends on order, not size. A wallet sitting at the 90th percentile for ROI scores the same whether it returned 3x or 40x. That single design choice is why one freakish 44x outlier cannot bend the whole scale and dominate the board.

Why Wilson, and why percentile rank

The Wilson lower bound discounts win rates measured over very few markets. A 1-for-1 record does not read as a perfect 100% — it reads closer to 21%, because one observation tells you almost nothing. A 25-of-30 record holds up far better. This is what stops a tiny lucky streak from masquerading as elite skill.

Percentile ranking handles the opposite failure: extreme magnitudes. Together they neutralize both ends — the lucky 1-for-1 and the lottery-ticket ROI — so the score rewards traders who are good and consistent.

Step 4: Combine into one weighted score

The four ranks are blended with a fifth drawdown-safety term (a wallet's worst peak-to-trough loss, ranked so smaller drawdowns score higher). Production weights are: win rate 0.27, ROI 0.23, realized PnL 0.23, turnover 0.17, and drawdown-safety 0.10. The drawdown term exists so a trader who racks up twelve small wins and one account-ending blow-up cannot ride win rate alone to the top.

What the score deliberately excludes

The composite score is realized-only. Unrealized, mark-to-market PnL on open positions is shown next to a trader, along with extra risk metrics like profit factor, reward/risk and max drawdown — but it is never folded into the core score. Open positions can swing either way, and we will not let paper gains inflate a ranking. You see those numbers; they just do not buy a higher rank.

See it for yourself

The honest way to test any ranking is to watch it play out. Put a top-ranked wallet on virtual paper copy trading and follow the equity curve in real time before risking a cent, or line a few up on compare traders side by side. When you are ready, create a free account and start with the politics category at no cost. Prediction-market trading carries real risk of loss, and past performance does not guarantee future results.