Risk-adjusted return: ROI minus 1.5 x drawdown

Sort a leaderboard of 7,000 traders by return and the top of it is dominated by accounts that made 400% and then had a day that cost them almost everything. That is not a prediction - it is what the numbers look like when you select on one variable and ignore its shadow.

This page exists to show that, and then to offer the fix.

Why return alone is a bad ranking

Three effects push the money-sorted top to accounts you should not copy:

1. Survivorship. Accounts that were liquidated are gone from the list, so the survivors look better than they should.
2. A single lucky trade. A trader with 3% of trades per month can post an enormous monthly ROI with one position. It shows up in the average; the median follower is not in it.
3. Sequence. The same trader can show +300% over 180 days with a 90% drawdown somewhere inside. The return number does not care in which order it happened; a follower cares about all of them.

Our own calibration makes the point: on 7,173 traders, 180-day ROI has an AUC of 0.471 for predicting a liquidation. Daily volatility has 0.772. Return is not weak evidence - it is nearly no evidence.

The alternative: return minus 1.5 x drawdown

Score = ROI (compound, from our daily series) − 1.5 × max drawdown

Why subtract drawdown at all when return is already a percentage? Because two accounts with the same return are not equally copyable, and the difference is what happened on the way.

Why the coefficient 1.5? A drawdown costs more than its face value for a follower: it is where the copy friction bites, where subscribers quit, and where the auto-unsubscribe rules cut you off. At 1.5, a −20% drawdown cancels +30% of return, and a −40% drawdown cancels +60%. A trader who earned 100% while underwater 30% ranks clearly below one who earned 60% with a 12% drawdown.

Our three ranking pages

Page	Ranking	What it shows
/	the composite rating 0–120	the intended default: risk first, return as a bonus
/low-drawdown/	return, filtered to low risk	"what does a safe account look like"
/risk-adjusted/	return − 1.5 × drawdown, filtered to low risk	the ranking with the drawdown term made explicit
/no-risk-filter/	return − 1.5 × drawdown, no filter	a control: what pure money-sorting looks like

/no-risk-filter/ is deliberately included as a counterexample. If you only want one page to convince yourself of the point, it is that one.

How to read the score column

• High score, low rating: the account earned a lot and risked a lot. It may be worth a small size with a hard stop.
• High rating, modest score: the account is boring and durable. This is what a follower with real money usually wants.
• Negative score: return has been outweighed by the drawdown on the way. The account has been a poor bet on a risk-adjusted basis even if it is green.

Why this is not the rating

The rating also counts liquidation days, volatility, win-rate quality, holding time, all-time ROI, tenure and follower PnL. The score on /risk-adjusted/ is one deliberately crude statistic - useful precisely because it is easy to disagree with, which is how you check whether the 15-component rating is doing real work.

Use it as a second opinion, not as the ranking.