Why a Winning Strategy Can Still Blow Up the Account

Scope: every number in this note is a compounding identity anyone can recompute — arithmetic, not market data. The simulated distributions the note specifies are not yet published.

A strategy with positive expectancy can still destroy the account that trades it, because expectancy and survival are different problems. The strategy determines the average outcome over many trades; position size determines whether the account lives long enough to collect it. Every number in this note is arithmetic — compounding identities anyone can recompute — not a market measurement. The arithmetic alone is enough to locate the danger.

If the strategy wins over time, what is left to go wrong?#

The sequence. A 55%-win-rate strategy with even payoffs is profitable on average, but an account does not experience averages — it experiences one specific ordering of wins and losses, and losing streaks of ordinary length are a statistical certainty, not a malfunction. The question that decides survival is not "does the edge exist?" but "what happens to the account while it waits for the edge to show up?"

That question has nothing to do with the strategy. It is set entirely by how much is risked per trade.

How fast does a normal losing streak become a hole?#

Compounding makes the damage nonlinear, and the recovery requirement grows faster than the loss. Five consecutive losses — unremarkable for any real strategy — at four different risk settings:

Pure compounding: remaining equity = (1 − r)⁵; required recovery = 1 / (1 − loss) − 1. No market data involved.

The last row is the whole story. Same strategy, same five losses — but at 5% risk the account is down 22.6% and now needs +29.2% to get back, earned with whatever confidence survived the streak. The asymmetry is mechanical: every loss shrinks the base the recovery must be earned on. Risk per trade does not scale the damage linearly; it bends the survival curve.

Is a losing streak evidence the strategy broke?#

Usually not — and acting as if it is causes the second, quieter blow-up. When size is too large, ordinary variance feels like strategy failure: a routine three-loss run at 5% risk is a −14% event the trader experiences emotionally, and the common response is to change the system right when nothing was wrong with it. The trader should have changed the size. Oversizing does not just threaten the equity; it corrupts the feedback loop the trader uses to evaluate their own method.

The degenerate case is escalation — raising size after each loss to "make it back". Doubling down converts the mathematics above from a survival curve into a cliff: variance increases exactly when the account can least absorb it, and the sequence loss → bigger position → bigger loss needs only a few steps to reach any drawdown threshold you care to name.

What is the actual research question?#

Not "what is the best percentage to risk?" — that has no universal answer. The measurable question is:

At what risk-per-trade does a strategy's normal variance become an account-threatening event?

Which is answerable by simulation, without any market data: hold the strategy's win rate and payoff distribution fixed, vary only risk per trade, and run the same trade sequence through each setting many times. The outputs that matter are distributions, not averages — maximum drawdown, probability of hitting a 10% / 20% / 30% drawdown threshold, longest streak survived, time to recover, share of runs that survive at all. This note states the design; the simulated distributions are not yet published, and nothing above depends on them — the compounding table is already sufficient to show the shape.

What changes tomorrow#

Before the next losing streak arrives — it will — decide which drawdown is survivable, then work backward from that number to a risk per trade, using the table's logic. The order matters: size chosen from a survivable drawdown protects the account and the judgment; size chosen from a profit target protects neither.

Related reading#

• The market does not know your P&L — the behavioral half of the same problem: what account state does to the decisions themselves.
• Before you believe a reversal story — the same discipline of separating what is measured from what is remembered, applied to chart narratives.