"One percent per EA" sounds disciplined, and it is the default most algorithmic traders reach for. But fixed risk per strategy quietly assumes those strategies are independent — and when they are not, the number you feel safe with can be several times the drawdown you actually carry. That gap is where funded challenges die.
The hidden assumption
Assigning each EA the same fixed risk feels neutral and fair. What it silently assumes is that the EAs' losses do not line up — that when one is bleeding, the others are roughly flat, so the account only ever feels one loss at a time. Under that assumption, three EAs at 1% each behave like a well-spread book. The assumption is almost never true for retail algo portfolios, which cluster on the same majors, the same sessions and the same broad market regimes.
Fixed risk per EA is safe only if the EAs are independent. Correlation, not the per-EA number, decides your real aggregate drawdown.
Why risk does not add up the way you hope
When strategies are independent, their combined worst-case loss grows with roughly the square root of their number, not the number itself. Five independent EAs at 1% do not routinely produce a 5% day; their typical bad day is closer to the square root of five — about 2.2% — because it is unlikely all five lose at once. This is the diversification benefit fixed sizing implicitly banks on.
The trap is the other direction. When strategies are correlated, that square-root cushion collapses. Fully correlated EAs behave as one big position: five of them at 1% is a 5% day whenever the shared driver moves against you. Real portfolios sit between these poles, but crucially they drift toward the correlated end exactly when it hurts most — in a sharp risk-off move, a central-bank surprise or a liquidity gap, correlations everyone measured as "low" snap toward one, and every EA loses together.
A worked example
Take a $100,000 account with a 5% daily drawdown ($5,000). You run five EAs at a fixed 1% each, and you sized them believing the book's worst day is around 2.2% — comfortably inside the limit, with room to spare. On a calm day that holds. Then a US inflation print lands off-consensus. All five EAs are effectively long the dollar-risk factor without your ever intending a concentrated bet. They stop out together: five times $1,000 is $5,000, the entire daily budget, in one release. The "2.2% book" just delivered a 5% day and breached the limit — not because any single EA misbehaved, but because the independence that justified the sizing never existed under stress.
What actually goes wrong
- Correlation is treated as static. The low correlation you measured in calm markets is not the correlation you get in a crash; it rises exactly when it can hurt you.
- Same-factor bets counted as diversification. Several EAs on correlated pairs, or all long-volatility, are one bet wearing several names.
- Concurrency is ignored. An EA holding multiple positions, or a grid that adds on drawdown, spends more than its nominal 1% at the moment of stress.
- The equity rule bites first. Correlated floating losses hit an equity-based daily limit simultaneously, so the breach can fire before a single stop closes.
- Adding EAs feels free. Each new "1% EA" seems to add negligible risk, but if it is correlated with the book it adds close to a full 1% to the worst case, not a diversified sliver.
Sizing that respects correlation
The fix is not to abandon fixed risk but to set the fixed number from the aggregate, not the other way round. Decide the worst combined day you can tolerate, estimate how correlated the EAs really are — honestly, and assuming correlation rises under stress — and back out the per-EA risk that keeps the combined worst case inside the limit. Practical guards:
- Size the per-EA number down to the correlation you would see in a bad week, not an average one.
- Treat EAs on the same instrument or factor as a single line for budgeting.
- Cap total simultaneous exposure across the book, not just risk per trade.
- Re-derive every EA's slice whenever you add or remove a strategy.
- Keep margin: never let the modelled worst case equal the hard limit exactly.
Seeing the gap
The Per-Strategy Risk Allocator below makes this concrete. Enter your account balance, the daily drawdown budget, the number of strategies and a correlation estimate, and it shows the combined worst case at that correlation against your budget — so you can watch a "safe" fixed 1% per EA turn into a full-budget day as correlation climbs from independent toward one. It also reports the risk you could safely use at a given correlation, which is the honest version of "how much per EA." Where that risk budget comes from, and how to simulate whether the resulting system passes, are covered in the companion guides.
Takeaway
Equal fixed risk per EA is a bet on independence you rarely hold. Size from the correlated worst case down, assume correlation spikes under stress, and the per-EA number that survives a bad day is smaller than the one that feels safe.
