In perpetual futures a trader can have a high win rate and still be losing money, for years. That is not a paradox. It is something anyone can do on purpose: close the winners, hold the losers. As long as a position stays open, a paper loss never becomes "a losing trade."
Exchange APIs report closed P&L — a trade enters the statistics only when you press close. So "win rate" really measures what share of the trades you were willing to realise turned out positive, not what share of your calls were right. Those two can differ enormously.
An open position sitting at −90% simply does not exist in a win rate.
We ran the check across 3,794 addresses with peak exposure above $1M, putting closed win rate and P&L recomputed at market prices side by side.
830 of them (21.9%) satisfy both:
· median closed win rate 50.2% — more than half their trades look like wins
· yet marked to market, their cumulative P&L is negative
One in five. What they share is a book of open, unrealised losses — the money is gone, it has just not been booked as "a losing trade" yet.
We ignore closed P&L and use mark-to-market: every day, every position is revalued at that day's market price, so unrealised gains and losses count exactly like realised ones. Whether a position is still open makes no difference.
Change the measure and the leaderboard turns over.
The top address by mark-to-market P&L is 0x5b5d…c060, at $172.7M. In our own earlier scoring — which was based on closed P&L — the same address was labelled "market maker / wash trader", win rate 11.6%, profit factor 0.01: a tag that would get him filtered out of any leaderboard.
One address, two measures, two opposite conclusions. That is the entire reason we rebuilt the mark-to-market ledger from scratch.
Changing the measure is not the end of it. Put the top two side by side:
| 0x5b5d…c060 (#1) | 0xecb6…2b00 (#2) | |
|---|---|---|
| Mark-to-market P&L | $172.7M | $117.4M |
| Of which banked | $83.4M (48.3%) | $85.1M (72.5%) |
| Peak exposure | $1.05B | $313.6M |
| Max drawdown | $160.4M | $47.0M |
| Trading days / days listed | 363 / 423 | 423 / 423 |
#1 made more, but most of his money is still on the book, not banked, and he got there with roughly 3.3x the exposure and 3.4x the drawdown. #2 made less, has nearly all of it in hand, and was in the market every single day.
That is why our pages show realised share and risk used next to the headline number instead of ranking on the total alone — "how much" and "in exchange for what" are two different questions.
2 addresses in our data are flagged separately: 68.7% of their volume comes from liquidation flow, with a median closed win rate of 99.2%.
That near-100% figure is not a read on direction. They receive other people's liquidated positions at the liquidation price and close them into the market, earning the protocol's spread. There is no signal to copy here: they do not predict, they wait for someone to blow up. Rank a leaderboard by win rate and these float straight to the top.
Before trusting a number about a trader, ask which trades it counts. If the answer is "the ones he chose to end," then it is not only measuring judgement — it is also measuring his willingness to admit a loss.