Low-win-rate systems create a particular kind of doubt. Losses arrive repeatedly, the latest drawdown feels unprecedented, and the next signal looks less credible than the last. Probability says many of these frightening sequences are ordinary.
One specified block is the easy calculation
Suppose each trade wins with probability p = 0.30 and loses with probability q = 0.70. If outcomes are independent, the probability that the next k trades are all losses is:
| Specified run | Calculation | Probability |
|---|---|---|
| 5 losses | 0.705 | 16.81% |
| 8 losses | 0.708 | 5.76% |
| 10 losses | 0.7010 | 2.82% |
| 12 losses | 0.7012 | 1.38% |
A 2.82% chance for the next ten trades sounds small. But a 100-trade sequence contains many possible places for a run to begin.
Somewhere in 100 trades is a different problem
It is tempting to multiply qk by the number of possible starting positions. That can exceed 100% and double-count overlapping runs. A run of eleven losses contains two overlapping runs of ten, for example.
The exact solution tracks k states. State j is the probability that the current sequence has avoided k straight losses and presently ends with j consecutive losses, where j ranges from 0 through k − 1. On each new trade:
- A win moves every surviving state back to state 0.
- A loss moves state j to state j + 1.
- A loss from state k − 1 reaches the forbidden run and leaves the surviving states.
After n trades, add the surviving state probabilities and subtract from one:
The exact 100-trade probabilities
| At least one run of | Probability at 30% wins | Plain-language reading |
|---|---|---|
| 5 consecutive losses | 99.94% | Effectively certain |
| 8 consecutive losses | 85.75% | Normal expectation |
| 10 consecutive losses | 58.01% | More likely than not |
| 12 consecutive losses | 32.91% | One chance in three |
| 15 consecutive losses | 12.05% | Uncommon, still plausible |
A trader who expects the 30% hit rate but sizes as though five losses is extreme has misunderstood the strategy. Eight losses are likely. Ten are more likely than not. Fifteen do not require a once-in-a-lifetime event.
Translate the streak into drawdown
If each loss equals −1R and each trade risks fraction f of current equity, k losses leave:
| Risk each trade | After 8 losses | After 10 losses | After 15 losses |
|---|---|---|---|
| 0.5% | −3.9% | −4.9% | −7.2% |
| 1.0% | −7.7% | −9.6% | −14.0% |
| 2.0% | −14.9% | −18.3% | −26.1% |
| 5.0% | −33.7% | −40.1% | −53.7% |
At 5% risk, a ten-loss streak that is more likely than not cuts equity by about 40%. Recovery then requires roughly 67%. The expectancy equation never changed. The sizing decision converted an ordinary path into a practical crisis.
When does a streak become evidence?
A long streak can be compatible with the model and still contain useful information. The correct question is not "Have I lost too many in a row?" but "How likely is the observed evidence under the model, including more than the run length?"
- Compare total win rate and payoff distribution with their historical intervals.
- Check whether losses cluster in a new market regime or one instrument group.
- Separate signal failure from execution drift, costs, and skipped winners.
- Use rolling and out-of-sample evidence rather than a single dramatic sequence.
- Predefine review thresholds before the drawdown changes your judgment.
Run length alone throws away information. A sequence with ten tiny losses and two +12R winners differs materially from a sequence with the same ten losses and no right-tail outcomes over a much larger sample.
The behavioral cost is part of the model
A tail-edge process requires participation in the next valid trade because that trade might be the rare outlier carrying the year. After ten losses, however, the next signal feels uniquely optional. Oversizing makes mathematically expected doubt financially and psychologically unmanageable.
Good sizing does not make the streak pleasant. It keeps the streak from taking away the ability to execute. A system that can be followed only on its average path is not an executable system.