A skilled trader wins again and again: over many closed trades, in separate time windows, with wins that cover the losses. A lucky trader shows one strong streak. Judge by profit in dollars, the number of trades, the size of wins and losses, and repetition. Not by ROI, and not by one month of profit.
The question a broker should ask
For the B-book, the question is not "who has the highest return?"
It is: does this account take real money from the book, again and again, with skill? That question has two parts, size and skill. ROI measures neither.
Trader A: $200 deposit, $1,000 profit. ROI 500%.
Trader B: $500,000 account, $75,000 profit. ROI 15%.
ROI says watch A. For the book, A is noise. B is the account that costs money, month after month.
Why a profit rule fails too
Many desks use a simple rule: profit over $1,000 in 30 days, so the client goes to A-book.
We ran that rule on 90 days of live flow. It moved 143 clients. Against keeping everyone in the book, it cost about $145,000. Most of yesterday's winners were lucky. The rule hedged them right when their luck ended.
On the same flow, our risk system moved 72 clients, only the ones with a repeatable edge. Result: about +$33,000.
The difference is not the threshold. It is who you move.
Five checks you can run on your own book
- 01Measure the threat in dollars. Net profit in dollars shows what the account takes from the book. A percentage hides it.
- 02Wait for enough closed trades. A few trades show luck, not skill. Until there are enough, the honest answer is "not enough data yet".
- 03Read the win rate together with the size of wins and losses. One client had 130 wins and 3 losses. It looks like skill. But his average loss was equal to 26 average wins, so one bad day takes back a month. He has no edge. He stays in the book.
- 04Look for repetition. Split the history into two separate windows. A real edge shows in both. A streak shows in one.
- 05Remove abuse first. Profit from price gaps, feed errors or copied accounts is not skill. Check for it before you judge skill, or the abuser will look like your best trader.
Where a sudden return still matters
An outsized day from an account that normally trades quietly is a signal. The signal is the break from its own pattern, not the percentage.
It is a reason to look at the account, not a reason to route it. In our system, it raises a review for the dealer. The route stays on the 90-day class.
How we do it
Our risk system runs these checks, and 789 measures in total, on every active trader, every minute.
Every account gets one of four verdicts: repeatable edge, no edge, abuse, or not enough data yet.
Our desk acts on the verdict automatically. The dealer supervises.