How much to risk per trade. The best size is a band, not a maximum.

Ask how much to risk per trade and the usual answer is a constant — one or two percent. Under an evaluation’s rules the honest answer is not a constant but a curve, because two of the account’s clauses pull in opposite directions. Risk too little and the account cannot reach its profit target inside the time window; risk too much and ordinary losing runs breach the drawdown. Somewhere between those two walls the probability of passing is highest.

For a trader with a real but modest edge, the model’s probability of passing peaked at 36.1% (±0.2) near 1.1% of balance per trade under a trailing drawdown, and at 50.1% near 1.4% under a static one. Move away from that size in either direction and the probability falls.

The curve below rises from zero, because at a tiny size the account cannot cover the 8% target in 20 trading days. It reaches a top. Then it bends back down, because past that size the account is large enough to trip the 6% drawdown before it finishes. The peak is broad, not a spike — being roughly right matters far more than hitting an exact number.

The peak is not a coincidence of these numbers; it is the meeting point of two mechanisms every evaluation shares.

The left wall is time. A profit target has to be reached inside a fixed number of trading days. A small edge compounded at a tiny size barely moves the account, so most attempts reach the deadline short of the target. At 0.5% risk — half the trailing optimum — only 16.6% of attempts passed, and most of the rest ran out of time rather than breaching the drawdown. Under-sizing is not gently sub-optimal; it is a cliff.

The right wall is the drawdown. Larger positions mean larger swings, and a run of ordinary losses that a small account would survive is, at a big size, enough to hit the floor. Doubling the size to 2.0% did not double anything good: the probability of passing was down to 26.5%, with 73.5% of runs ending against the drawdown. But note the asymmetry — past the peak the odds do not keep sliding so much as scatter across a luck-dominated plateau, while under-sizing collapses them outright.

The best size is simply where these two walls are furthest apart: large enough to reach the target in time, small enough to survive a normal losing streak. That is why the answer is a band, and why it depends on the trader’s edge and the account’s rules rather than on a rule of thumb.

The most useful thing the model shows is not that failure rises on both sides of the peak, but that on each side it fails for a different reason. Track why attempts end as size increases and the dominant cause switches from “ran out of time” to “hit the drawdown.”

For the trailing archetype, the drawdown overtook time as the larger cause of failure between 0.5% and 0.6% risk per trade. Below that band, an account mostly dies of the clock; above it, mostly of the floor.

This matters for a diagnosis. A trader who keeps timing out is under-sized and should read the left wall; a trader who keeps hitting the drawdown is over-sized and should read the right. The same symptom — “I keep failing” — has two opposite cures, and the crossover is the line between them.

The two curves in the first figure came from the same trader and the same 6% drawdown — they differ only in how the floor is measured. A static floor sits at the starting balance for the whole evaluation. A trailing floor follows the account’s high-water mark upward and never comes back down.

That difference is not cosmetic. Under a trailing rule, a good early run does not bank a cushion; it drags the breach line up behind the account, so the very same profits that bring the target closer also bring the floor closer. The account is chased. The whole curve shifts down and to the left: the best size is smaller, and the best odds are lower.

• Static floor · size 1.4%50.1%
• Trailing floor · size 1.1%36.1%

Same edge, same target, same drawdown depth — the trailing measurement alone cost this trader about 14 percentage points of peak probability and pushed the best size from 1.4% down to 1.1%. Even at the same size (1.1%), the static account passed 47.4% against the trailing account’s 36.1% — a 11.3-point gap from the rule mechanic alone. The static-vs-trailing study maps this mechanic across the rule corpus.

The number is not transferable. This study used one synthetic archetype and one illustrative trader; a real evaluation may add a daily loss limit, a consistency rule or a minimum number of trading days, each of which only tightens the picture. It also fixes one sizing convention: each trade risks a set fraction of the current balance, so the account compounds as it grows — a trader who instead risks a fixed dollar amount, or sizes off the starting balance, would trace a different curve. And the whole curve is anchored to the reference edge: raise the win rate or the reward-to-risk and the peak lifts and widens — and, less obviously, its best size moves down, because a stronger edge reaches the target without having to court the drawdown, so it wants a smaller position, not a larger one; weaken the edge and the curve sinks and flattens, with the best size forced the other way — up toward the top of the range — as the account gambles on size to beat the clock. The value here is the shape, not the coordinate.

Three things do carry across, and they are the point:

• There is a best size, and it is a band with two different walls — running out of time on one side, the drawdown on the other.
• The size that is best under a static floor is too large under a trailing one; the trailing mechanic moves the whole trade-off down and to the left.
• Sizing well converts an edge efficiently, but it cannot invent one — at the best size in the model, 63.9% of attempts still failed.

Finding the band for a specific set of numbers is a computation, not a guess. The PropSurvival simulator runs exactly this sweep on a trader’s own win rate, average win and loss, cost and trade frequency against a chosen rule set, and reports the same curve and the same failure-cause split shown here. The position-sizing methods guide covers how to turn a chosen risk percentage into a contract count, and risk of ruin covers the survival side of the same trade-off.

What is the best risk per trade for a prop-firm evaluation?

There is no single number that transfers between traders. The probability of passing is a curve with a peak: too small and most attempts run out of time, too large and most breach the drawdown. For the illustrative profile here the peak sat near 1.1% of balance under a trailing drawdown and near 1.4% under a static one.

Why does risking more reduce the probability of passing?

Past the peak, larger positions swing equity far enough to breach the drawdown before the target is reached. The cause of failure flips from running out of time to hitting the floor, crossing over between 0.5% and 0.6% risk in the model.

Why is a trailing drawdown harder than a static one at the same size?

A static floor is fixed at the starting balance; a trailing floor follows the high-water mark up and never falls. Profits that would build a cushion instead ratchet the breach line up behind the account. In the model this alone lowered the peak by about 14 points.

Does this mean a trader should size at exactly 1.1%?

No. Every figure is model-derived for one synthetic archetype and one illustrative trader. It shows the shape of the trade-off, not a number to copy. The simulator computes the same sweep on a trader’s own inputs.

Is sizing more important than edge?

Sizing optimally cannot manufacture an edge. Even at the best size in the model, 63.9% of attempts still failed under the trailing archetype. Size decides how efficiently an edge is converted into a pass; it does not replace it.