A losing streak is a sequence of consecutive losses. A drawdown is a decline from a previous equity high. A challenge failure is a breach of a contractual boundary. These events are related, but they are not interchangeable. Calling the probability of five consecutive losses the risk of ruin can materially misstate what a strategy may do.
Streak calculations are still useful. They provide an understandable first check of whether a proposed position size leaves room for ordinary adverse sequences. The important step is to ask the correct question: a streak starting at one specific trade, a streak appearing anywhere in a longer sample, or the first crossing of an account loss limit. Each requires a different calculation.
One specified streak is the easy calculation
Assume independent trades with a constant 55% probability of winning and therefore a 45% probability of losing. The probability that the next four trades all lose is 0.45 to the fourth power, or 4.100625%. The probability that the next five all lose is 1.84528125%. These are exact under the assumptions.
The word next matters. Those figures describe one fixed starting location. They do not describe the probability that a streak occurs somewhere during 50 or 100 trades. Longer histories offer many overlapping places for a streak to begin. Ignoring those opportunities makes normal losing sequences appear far more surprising than they are.
| Consecutive losses | Probability from one fixed starting point |
|---|---|
| 3 | 9.11% |
| 4 | 4.10% |
| 5 | 1.85% |
| 6 | 0.83% |
| 7 | 0.37% |
| 8 | 0.17% |
None of these probabilities says that a winning trade is due after several losses. Under independence, the probability of losing the next trade remains 45%. If losses are dependent, that conditional probability may change, but it must be estimated from evidence about the process rather than inferred from a desire for the sequence to balance out.
Calculate a streak anywhere in the sample
For a run of at least four losses anywhere in 20 trades, overlapping candidate sequences prevent a simple multiplication by 17 starting positions. A sequence of five losses contains two overlapping runs of four. Counting each starting location as an independent event double-counts shared outcomes.
An exact recurrence avoids the problem. Keep probabilities for surviving histories ending with zero, one, two or three consecutive losses. Initially the zero-loss state has probability one. A win sends all surviving mass to state zero. A loss moves each state to the next. Mass leaving state three on a loss has achieved the target streak and is removed from the surviving states.
After the required number of trades, one minus the surviving probability is the chance that at least one target streak appeared. This method extends to any streak length by storing more states. It does not need Monte Carlo noise, and it gives a useful reference for checking a simulation.
| Trades observed | At least 4 losses in a row | At least 5 | At least 6 |
|---|---|---|---|
| 20 | 35.84% | 16.42% | 7.14% |
| 50 | 69.94% | 39.49% | 19.37% |
| 100 | 91.50% | 64.68% | 36.29% |
The same 55% win probability produces nearly a 70% chance of at least one four-loss run in 50 trades. That is compatible with a positive trading edge. The table uses fixed-length histories that continue for the full stated number of trades. If a challenge stops early at a target or loss limit, the distribution of streaks actually observed before stopping is a different problem.
Translate position size into remaining room
Imagine a static loss allowance of $10,000 from a $100,000 starting balance. Fixed $2,000 losses reach the boundary after five consecutive losses from the start. Fixed $1,000 losses require ten. Fixed $500 losses require twenty. If touching the limit is failure, the boundary-reaching loss is not survived.
This arithmetic assumes no earlier gains or losses, no costs beyond the stated loss, and no open-position overshoot. After the account has already lost $3,000, the remaining buffer is $7,000. At $2,000 per full loss, three losses leave $1,000 and a fourth crosses the floor. The current buffer matters more than the original advertised allowance.
For fractional current-equity risk, use compounding. After k losses at fraction f, equity is initial equity multiplied by (1 − f) to the power k. Five 2% losses leave about 90.3921% of the initial balance, so the account has not yet reached a 10% static loss floor. Six leave about 88.5842%, which has crossed it. This differs from fixed 2% of initial balance.
A streak probability is not a failure probability
Consider a strategy whose wins are +0.2R and losses −1R. The sequence loss, win, loss, win, loss, win, loss totals −3.4R without ever containing two consecutive losses. A 3R loss boundary is breached even though a long losing streak never occurs. Unequal payoff sizes and accumulated small losses make streak-only analysis incomplete.
The reverse also holds. A static-floor account may accumulate a large profit cushion and then experience four consecutive losses without failing. A trailing rule may preserve much less cushion. The same losing run has a different contractual effect depending on where it occurs and how the floor moves.
First-passage failure analysis tracks account state through time. For a simple fixed-step process, state includes cumulative profit and whether a terminal boundary has been hit. Daily limits add the daily reference and reset schedule. Trailing limits add a high-water mark. Real portfolios add open equity and concurrent exposure. A streak counter alone does not contain that information.
Distinguish static loss from peak drawdown
Suppose the account rises from $100,000 to $108,000 and falls to $101,000. The drawdown from the peak is $7,000, or approximately 6.48% of the peak. The account remains $1,000 above its start. A static floor at $90,000 is far away, while a $5,000 trailing floor based on the peak would be $103,000 and would already have been crossed.
Do not divide every drawdown by initial capital unless that is explicitly the metric you intend to show. Cash drawdown, percentage of peak equity and percentage of initial account are different units. A report comparing risk with a contractual cash allowance should use the actual cash floor, while a performance report can additionally show peak-relative drawdown.
Intraday peaks can also matter. If a trailing rule follows open equity, an unrealized gain can lift the floor before the trade closes. A sequence of profitable closing trades does not prove that the account survived. You need the intervening equity path or an appropriately conservative representation of it.
Check dependence before trusting the independent model
The streak table assumes the probability of a loss stays at 45% regardless of recent results. Strategies often violate that assumption. A mean-reversion strategy may struggle repeatedly during a strong trend, and several same-day trades may share the same signal error. Those relationships can lengthen losing clusters.
Examine outcomes by session and market regime. Compare the loss probability after a loss with the overall loss probability, but do not overinterpret small conditional samples. A few observed clusters are not enough to estimate a stable transition model. Keep uncertainty visible and compare independent resampling with session or block resampling.
Dependence can also come from the trader rather than the market. Increasing size after losses, skipping signals after a drawdown or entering correlated positions changes the process. A backtest using constant independent outcomes cannot evaluate those behaviors unless the policy is explicitly implemented. Test the policy you will execute, not a more disciplined imaginary one.
Use stress tests that target the failure mechanism
If the account is mainly exposed to consecutive full-stop losses, the streak table is a useful sizing screen. If many small wins are punctuated by rare large losses, stress the tail losses and their timing. If daily limits dominate, retain same-session clusters and overlapping trades. The test should reflect the mechanism that can actually terminate the account.
Increase costs carefully. A trade previously classified as a small win may become a loss, which changes both the streak distribution and cumulative equity. If costs are already included in the historical net results, apply only the extra stress amount. Keep break-even trades as their own category when their treatment affects streak definitions.
Decide whether a zero result interrupts a losing run. For a strict sequence of negative trades, it does. For a broader period without a positive recovery, it may not. Neither definition is inherently wrong, but the label must make the distinction clear. Otherwise two reports can disagree because they measure different events rather than because one calculation is broken.
A practical sequence for your own strategy
- Measure net trade outcomes and define what counts as a loss.
- Calculate fixed-location and whole-sample streak probabilities separately.
- Compare planned cash risk with current daily and total buffers.
- Run a full first-passage account simulation with exact rules.
- Repeat using session or block sampling to examine clustering.
- Evaluate the policy on a separate period instead of choosing risk from the best historical streak.
When a live streak arrives, a good analysis helps you distinguish an expected adverse sequence from a possible change in the strategy. It cannot decide that question from streak length alone. Review execution, market conditions and the size of the original evidence before changing the system. A predeclared response is more reproducible than improvising a new rule after every uncomfortable run.
The useful conclusion is specific: this position size can tolerate this amount of adverse movement under these assumptions. A high win rate does not remove losing sequences, and a streak probability does not replace an account-level failure model. Use the simple calculation as an entry point, then measure the full path that the contract will judge.