Copy Score
See who is actually worth copying.
Profit shows who won. Copy Score shows whether following them has historically left anything after the cost of copying. One number, about the trader, free on every surface.
What you see on a trader
Copy Score
Copying them has worked on their finished trades.
A read on the trader, not on this trade.
Based on their finished trades to date, not a forecast. Assumes every one of their trades was copied exactly, so your filters, budget and timing will change it.
The problem with leaderboards
Two traders, both up $10 million, and only one of them is copyable
Imagine two traders. Both made $10 million last quarter. On any leaderboard in this industry they sit side by side.
Looks identical. You cannot copy this.
The first made it on $1 billion of volume: thousands of trades, a fraction of a cent of edge on each, most of it earned by posting orders and waiting rather than taking them. Follow them and you pay the spread they collected, pay a fee on every one of those thousands of trades, and arrive a moment too late on all of them. Their edge was in being the market.
Looks identical. This one you can follow.
The second made it on $20 million of volume: fewer, larger, more considered positions. Follow them and you are buying the same thing they bought, at close to the same price.
Identical on a leaderboard. Opposite in your account. Ranking by P&L is not a small approximation. It is the wrong question asked confidently.
Rank by profit vs rank by copyability
Illustrative of the pattern we measure, not a live board. Traders with the largest headline profits frequently sit mid-table once the cost of following them is charged. Quieter traders rise.
Ranked by profit
- 1High-frequency market maker
- 2Arbitrage desk
- 3Large-volume scalper
- 4Event specialist
- 5Patient position trader
- 6Sports handicapper
- 7Occasional big-ticket
- 8Momentum follower
- 9Long-shot punter
- 10Broad diversifier
Ranked by copyability
- 1Event specialist
- 2Patient position trader
- 3Sports handicapper
- 4Occasional big-ticket
- 5Broad diversifier
- 6Momentum follower
- 7Large-volume scalper
- 8Arbitrage desk
- 9High-frequency market maker
- 10Long-shot punter
What it is
One number: what following this trader is worth, after what it costs
Not how much they made. What is left for you after the realities of copying them.
The score starts from a trader's actual settled results (real positions, real outcomes, no marked-to-market guesses) and then subtracts the things that stand between their result and yours. What comes out is an estimate of the return a follower could expect per copied trade, after costs.
Positive means the trader's edge has historically survived the journey to a follower's account. Negative means it has not, no matter how good their P&L looks.
We ask three questions. We do not grade the click in front of you.
- Have they finished enough trades for this to mean anything?
- After worse fills, venue fees, and our 0.5%, is there historically anything left on their tape?
- Do they have a record in this kind of market (NFL, politics, crypto, and so on)?
Win rate and dollar P&L stay on the profile. Those are how they did. Copy Score is whether following them has been worth the toll.
The toll
Why copying is not free
Three costs sit between their P&L and yours. A fourth, our 0.5%, applies when you use Auto Copy.
The price you pay, not the price they paid
You arrive after them. On a fast-moving market that difference is small. On a thin one it is most of the edge. We measure it from our own fills rather than assuming it.
The exchange's fee, which is not flat
Prediction-market fees peak near even odds and fall away toward certainty. Copying a 5¢ long-shot can cost several times what the same trade costs at 90¢.
How they actually make money
A trader earning from rebates, spreads, or arbitrage has an edge that exists because they are on the other side of the trade. Copy them and you take the side they were paid to avoid.
Our fee on Auto Copy
Auto Copy includes Polycopy's 0.5%. Manual copies do not. The score still has to clear the venue costs either way.
What it costs to follow a trade, by the price of the share
As a percentage of the money committed. The curve is a property of how prediction-market fees and a fixed tick of slippage sit on cheap shares. A cost model that ignores price is wrong in both directions at once.
18.2% of your money, at 5¢. 1.7% at 90¢. A trader who specialises in cheap long-shots has to be several times better just to break even for you. Copy Score prices each trader against the curve they actually trade on.
A market maker can print a beautiful P&L that is the spread you would pay. Their result is real. Mirroring them is often taking the other side of the fill they are selling. Copy Score is there to flag that, not to hide it. Some traders also carry a hard-to-mirror flag (market maker, arbitrage, or 90+ orders a day). The flag is why mirroring fails. Copy Score is whether copying has paid.
How to read it
Start with the band, not the decimal
When we can score a trader, Copy Score is a signed whole number with no percent sign. +6 means: on their finished trades, after those copy costs, the leftover was on the order of six. −4 means the leftover was negative. We round to a whole number on purpose. Tenths would pretend at precision we do not have.
That number is not a win probability, not a 0 to 100 rating, and not a letter grade. The bands are what we are confident in. The precise decimal is not, and we deliberately do not put tenths in front of you.
Before you read the bands
Most traders score negative. The typical graded trader sits near -2.05 — so a negative number is the norm, not a sign that copying cannot work. Copying anyone costs something. What changes is whether their edge clears it.
Based on their finished trades to date, not a forecast. Assumes every one of their trades was copied exactly, so your filters, budget and timing will change it.
| You see | What it means |
|---|---|
| Clears(and a number) | Copying them has worked on their finished trades. |
| Ahead(and a number) | They finish ahead of what copying costs — but only just. |
| Caution(and a number) | They make money. It doesn't survive being copied. |
| Losing(and a number) | They've been losing on their own trades, before copying costs even start. |
| Not comparable | They close out early, so we can warn about them but never rank them alongside hold-to-the-end traders. |
| Thin history | Too few finished trades to stand behind. A big return either way can still be one bet that came in. |
The number appears on the first four. The last two are the absence of a score, not a middling one. In the product, a trader with no number shows a dash, and the words above are in the hover.
Not scored is a third case: the classifier has not graded this wallet. That is not a middling score; it is the absence of one.
Using it
Require both signals to agree
Strong past results and a strong Copy Score. Our own testing shows that either one alone is worth little.
Start with the band, then the P&L
Skip or be careful with Caution, Losing, Not comparable, and Thin history. Then look at people you would copy on volume and category anyway. Do not sort a list of strangers by the number and call the top row the best to copy. We measured that ranking on our own fills. It did not produce a profitable book.
Treat a dash as information
Some traders show no number because they have too few settled positions to judge, or because they close out early — we can tell you when one of those looks bad, but we will not rank them against traders who hold to the result, because the part of their return that comes from selling well is not a part a copier receives. That is not a gap in our data. It is us refusing to guess. A trader with three resolved bets and a spectacular return is not a signal.
Match the trader to your budget
A trader placing hundreds of small trades a day and a trader placing three considered ones a week demand completely different amounts of capital to follow properly. Copying a high-frequency trader on a small budget means catching a thin, arbitrary slice of what they do.
Expect to keep less than they make
You arrive after them and you pay fees they may not. A trader's edge has to clear that gap before any of it reaches you. That is the gap Copy Score exists to measure, so you can pick traders whose edge is big enough to survive it.
We do not publish the exact formula, the thresholds, or the weightings. A published scoring rule on a public order book is a rule that gets traded against. What we do publish is the method, the tests, and the results, on Copy Score performance.
Category record
A record in this kind of market is a separate fact
A trader can clear copy costs overall and still be ordinary (or worse) in the market on this card.
The category mark is not “this kickoff is a lock.” It is “if you copy them, copy them here, not in a category they do not have a record in.” On our copies, following the same people in a market type they already had a record in beat following them in a market type they did not.
Where you will see it
Where Copy Score shows up
Copy Score is free on every surface. There is nothing to unlock.
Discover
Copy Score is a way to look at traders. The board is still ranked by period P&L. Copy Score is a filter on who is even in the conversation, not the sort.
Trader profile
One box for the trader, plus where they have a category record. Same meaning as Discover.
Auto Copy
Used to decide who you are allowed to follow, not which of their clicks to skip. Your configs (price, size, horizon) still decide how a copy is placed.
Feed cards
Same trader, same category, same Copy Score. That is the point. We are not re-grading every bet.
Help Center walkthrough: Understanding Copy Score.
What changed
We retired the per-trade grade
We used to publish a Copy Score on each trade: a 0 to 100 number and a letter grade that claimed to estimate whether that trade would beat the market.
That number mostly tracked the market price. On our Auto Copy fills, the trades the old score liked most lost more, not less.
We killed that grade. We kept the name because it still answers the copy question. The question changed: not “will this trade win?” but “is this person worth copying, especially in this market?”
The old backtest tables described the retired score. They are not how Copy Score works now. The evidence page is Copy Score performance.
Straight answers
What the Copy Score does not do
Anything sold as a way to beat a market deserves scepticism, so here is the other side, stated plainly. These come from our own testing, not from a compliance page.
| It does not | Because |
|---|---|
| Predict any individual trade | It is a statement about a trader over many positions, not about the next one. |
| Guarantee a positive return | A +0.26 tendency for skill to persist is real and useful. It is not certainty. |
| Replace your own judgement about risk | It measures whether an edge survives being copied. It does not size your position or tell you what you can afford to lose. |
| Work equally well on every trader | Traders who sell most of what they buy cannot be judged on settled positions alone. We say so rather than scoring them anyway. |
| Stay right forever | Fees change, markets change, and traders change. The score is rebuilt continuously and re-tested against fresh out-of-sample data. |
FAQ
Common questions
What is Copy Score?
Copy Score tells you whether a Polymarket trader is worth copying after the cost of following them. It is a read on the trader's finished history, not a prediction that this specific trade will win.
How is Copy Score different from win rate?
Win rate and P&L describe how the trader did. They do not tell you whether you can capture that result. You fill worse than they do and you pay fees. Copy Score subtracts those copy costs and asks what is left.
What does the number mean?
When we can score someone, it is a signed whole number with no percent sign. Positive means that on their tape, leftover after copy costs was positive. Negative means it was not. It is not a win chance and not a 0 to 100 rating.
Why isn't the most profitable trader the best one to copy?
Profit tells you who won. It tells you almost nothing about whether you can win by following them. A market maker can print a large P&L that is the spread you would pay. Ranking by P&L and ranking by copyability produce different lists.
Is Copy Score a trading recommendation?
No. It is an indicator. It helps you avoid traders who are hard to copy. It does not tell you to buy or sell. Past copyability is not a guarantee.
Why did Copy Score change?
The old score graded each trade. That grade tracked the market price and did not pick winners on our copies. Copy Score now asks whether following the person has historically left anything after costs.
Copy the traders whose edge survives the trip
Not the biggest number on a leaderboard. Use Copy Score as a warning, then copy people in the categories they already have a record in.
Keep reading
Performance
Sealed scores on 1,796 traders, and what our copies showed
Help: Understanding Copy Score
The states, and how to act on them
Help: Trader types we flag
Market makers, arbitrage, very high frequency
Trader Data
How our trader data is built, and every indicator in it
Copy Trading
How copying works on Polymarket
All Trading Bots
Every bot type on Polymarket
Kalshi vs Polymarket
Platform comparison for bots
Risk disclaimer
Copy trading does not guarantee profits
A trader who clears copy costs on their own tape can still lose in the next month, and our fills can differ from theirs. Copy Score is a decision-support indicator, not financial advice.
Figures on this page come from measurements on settled positions and real executed fills. Out-of-sample tests cover 1,796 traders across a fixed cut-off. Cost figures are measured from executed trades, not modelled as a flat percentage.
The cost curve above is a model of what copying costs at each price, not a quote. The exchange sets its own fees and has changed them; we re-measure ours against real charged fees and update the model, so the same trader can move bands when costs move.
Copy Score is calculated as though every one of a trader’s trades were copied exactly. No real bot does that — filters, budget, timing and unfilled orders all subtract from it. Treat the score as a ceiling on what following someone could return, not as an expected outcome.
Markets change. Trader behavior changes. No score removes risk, and past copyability is not a guarantee of future copyability. Past performance is not a guarantee of future results. Copy trading carries risk, including the loss of the amount you commit.
Polycopy does not tell you to place a trade.
A Strategy Bot runs this filter for you, and runs it again every night.
See the strategies