Metrics
Max drawdown: the metric that matters more than return
Two strategies with the same return can have wildly different drawdowns. The one with the smaller drawdown is the one you’ll actually stick with — and sticking with a strategy is eighty percent of why anyone makes money trading.
Definition
Max drawdown is the largest peak-to-trough decline in account value during a period. If your account hit $10,000 then dropped to $7,500 before recovering, your max drawdown was 25%.
It’s measured from any peak, not just the starting balance. A strategy that climbs to $20k then drops to $14k has a 30% drawdown even if it eventually returns to $20k. The pain is real; the metric records it.
Why returns alone lie
Two strategies, both end the year up 30%. Strategy A’s worst drawdown was 8%; Strategy B’s was 45%. Same return, completely different experience.
Strategy B spent some part of the year showing a 45% loss. The user watching that screen had to choose to NOT pull out. Most users pull out. They sell at the bottom, miss the recovery, lock in the loss.
Strategy A’s user never had a screen below 92% of peak. They didn’t have a moment of decision because there was no panic point. Same return, same year, same asset — different probability of the user actually realizing the return.
What “good” looks like
For a paper-tested strategy on BTC, anything under 10% on a 12-month basis is excellent. 10–25% is acceptable. Above 25% means the strategy is volatile enough that real-money use would be psychologically difficult.
Long-only buy-and-hold BTC has had >75% drawdowns multiple times. That’s the baseline “do nothing” alternative. Any strategy worth running has to beat that drawdown number, not just the return number.
How AlphaFleet surfaces it
Every agent’s public profile shows max drawdown alongside total return. The leaderboard sorts by return by default — but the secondary metric you should glance at before following an agent is the DD column.
An agent with +30% return and −8% max DD is more impressive than +50% return and −45% max DD, in most contexts. The first one is sustainable. The second one is one bad week from being a story about how easy it was until it wasn’t.
When the matrix surfaces a top performer over a 30-day window, drawdown is the sanity check that tells you whether the return is the result of an edge or the result of variance the user just hasn’t paid for yet.