Win rate 99%: a warning sign, not good news

Traders with a 95–100% trade win rate look like a dream: almost every trade profitable. In practice that percentage of losing trades is almost never skill - it is a construction with one large bill attached. Here is how it works, and why the rating penalises it.

What win rate actually is

Win rate is the share of winning trades, not of winning days. Those are different things, and confusing them is the source of almost every wrong conclusion:

• a 70% win rate per trade can correspond to 55% of profitable days;
• the reverse also holds: 90% per trade with 45% profitable days is the classic grid signature.

On a trader page we show both: the exchange's per-trade win rate (ProfitWinRateE4) and our own daily win rate over 180 days. If they differ by more than 15 percentage points, a few large trades are carrying the result - which is a completely different style.

What a grid looks like in the statistics

A grid (or martingale) is a sequence of trades with increasing size after a loss. Its identifying features:

• win rate 95–100%. A losing streak in a grid ends with a profitable exit, so losing trades are rare by construction.
• Very many trades in one pair. Profit concentrates in a single symbol - often 80–100% of the total.
• Short holding times. Positions live minutes, not days.
• Drawdown that is deep but narrow. Few days with a 90%+ drop, and those are exactly the ones that happen.
• Follower PnL is often negative. While the trader closes the grid in profit, followers whose own limits and margin differ can end up on the losing side.

We do not call this fraud: a grid is a legitimate strategy. It simply moves the risk from many small trades to one large one. On the equity curve it looks like a staircase up with occasional vertical drops.

Why it is dangerous specifically for a follower

Copying does not reproduce the trader's account:

1. Your leverage is your own. Bybit does not copy the master's leverage by default, and if you do copy it, your margin may not be enough.
2. Volume filters. A trade the trader opens at 5% of capital may not open for you: a minimum order size or a daily limit will block it.
3. Slippage. On a run of entries into one position your average price is worse than the master's.
4. Auto-unsubscribe. 20 failed copies in a row, or 100 in 24 hours, and the platform unsubscribes you - usually right at the reversal that closes the grid.

Net result: you can get the worst part of the strategy - all the entries and none of the exit.

How the site checks it

• Rating Breakdown → "Exchange win rate". We use the median of the 30/90/180 day and all-time windows, which is robust to one outlier period, unlike an exchange number for a single window. The scale is steep: 85% is nearly free, 90% is noticeable, 95%+ costs 12 points, 100% costs 20.
• The return bonus is capped. If the 180-day exchange win rate is ≥ 90%, the ROI bonus is limited to +10 points: a grid should not buy the top of the leaderboard.
• Profit concentration. The review mentions the share of profit in one symbol when it exceeds 50%.
• "Days with a drop of 90% or more". For a grid this is 1–3 days costing 18–30 points - the most expensive penalty in the formula.

What to do if the numbers are good but the construction looks wrong

1. Look at holding time. Under an hour with a high trade count is almost certainly a grid.
2. Look at the number of pairs. One pair means grid with high probability.
3. Look at follower PnL. Positive all-time and negative over 90 days means the entry is bad right now, whatever the rating says.
4. If you copy anyway, keep the deposit share small and set a PCSL: leaving by a pre-set rule is better than deciding while it hurts.

A high win rate is not a reason to be pleased - it is a reason to ask: what did it cost to produce, and who pays when the next losing streak is longer than usual? The rating answers that question in advance.