Basics · Nº 005 · 16 October 2026 · 8 min read
Your good strategy will probably lose ten trades in a row. Plan for it.
We ran a profitable strategy 10,000 times over three years. 95% of the runs made money. Yet two-thirds of all runs also lost ten trades in a row, and the typical run went eleven months without a new high. A trader who quits at a 15% drawdown would have abandoned this working strategy four times out of five, and missed a median further gain of 30%.
It is month seven. Your strategy has just lost its eleventh trade in a row. The account is 18% below its peak, and it has not made a new high since February. Every instinct says the strategy has stopped working: the market has changed, the edge is gone, it is time to stop. Here is the uncomfortable part. For a perfectly healthy, profitable strategy, every one of those things is normal. The moment you feel most certain it is broken is a moment when the evidence says almost nothing, and quitting does not undo the losses already taken. It only gives up whatever edge was left.
"Ten losses in a row means the strategy is broken."
The experiment: a strategy we know works
We used the same strategy as in Nº 002. It wins 40% of the time; its winners make 2R and its losers lose 1R, where R is the amount risked on a trade (risk ₹5,000 per trade and 1R is ₹5,000). It risks 1% of the account per trade and pays 0.08% of the account in brokerage, taxes and slippage on every trade. Its edge is positive and known: about +0.12R per trade after costs.
A few terms used throughout. The edge is the average result per trade. A drawdown is a fall from the account's highest point to a later low, and a new equity high is the moment the account finally climbs above its previous best. The median is the middle result: half the runs did better, half worse.
We ran the strategy 10,000 times, 450 trades per run (about three years at 150 trades a year). This is a Monte Carlo simulation: many possible histories of the same strategy, generated at random, so we can see the full range of what it can do rather than one lucky or unlucky path. Because the edge is built in, we know with certainty that every losing streak and every drawdown in the results is just the strategy's normal variance, not a sign that anything has changed.
Over three years, the median run returned +63.5%, and 94.6% of runs were profitable. By any reasonable standard, this is a strategy worth trading. Now look at what trading it actually feels like.
Losing streaks: longer than anyone expects
| Losses in a row, at least once | Share of runs |
|---|---|
| 6 or more | 100% |
| 8 or more | 95.8% |
| 10 or more | 66.8% |
| 12 or more | 33.0% |
| 15 or more | 8.2% |
The median longest streak was 10 consecutive losses. One run in three saw twelve or more. One in twelve saw fifteen.
Why so long? Each individual trade has a 60% chance of losing, so the chance that any particular ten trades all lose is 0.610, about 0.6%. That sounds rare. But 450 trades contain hundreds of overlapping stretches of ten, and every one of them is another chance. Rare events become likely when you give them enough opportunities. "One in 165" sounds safe. "One in 165, with more than 400 overlapping chances" is a very different proposition, and intuition rarely makes that adjustment.
Drawdowns: deep, and long
| Drawdown at some point | Share of runs |
|---|---|
| 10% or worse | 99.2% |
| 15% or worse | 79.3% |
| 20% or worse | 44.7% |
| 25% or worse | 20.2% |
| 30% or worse | 8.3% |
The median worst drawdown was 19.2%; one run in ten fell 29% or more; one in a hundred, 39.9%. And these drawdowns were not brief dips. The typical run's longest stretch without a new equity high was 136 trades, about eleven months. For one run in ten it was 287 trades: nearly two years of trading a profitable strategy with nothing to show for it.
That is the reality of a strategy that makes 63.5% over three years: long periods of going nowhere, interrupted by the gains that pay for them. A smooth equity curve is the exception, not the rule.
The quitter's problem
Now add a human. Suppose a trader runs this strategy with a simple rule that feels prudent: "If I'm down 15% from my peak, something is wrong and I stop." We applied that rule, and a 20% version, to all 10,000 runs. (A 10% threshold is not worth a row: 99.2% of runs hit it, so almost everyone would quit.)
| Quits at a drawdown of | Traders who quit | Their account when they quit (median) | What the system made after they quit (median) | System rose after they quit |
|---|---|---|---|---|
| 15% | 79.3% | +7.4% | +30.4% | 87.2% |
| 20% | 44.7% | +2.6% | +23.3% | 84.6% |
With a 15% pain threshold, four traders in five abandon a strategy that works, typically while their account is still up overall. After they left, the median system went on to make another 30.4%, and in 87% of cases it ended higher than where they quit: not because drawdowns rebound, but because the edge simply kept working.
Be clear about what this does and does not mean. The drawdown does not make a rebound more likely: in this simulation every trade is independent. Quitting after the same number of trades for no reason at all, whatever the drawdown, would have cost exactly the same (a median of +30.4% forgone). What the quitter loses is simply the edge on every trade they no longer take, a median of 289 trades here. The drawdown did not tell them the strategy was broken. It just made giving up its edge feel justified.
A profitable strategy with a 40% win rate will, more often than not, lose ten trades in a row, and typically falls about 20% from a peak and goes most of a year without a new high. Judged by feel, those are signs of a broken strategy. Judged against its own statistics, they are ordinary, and quitting because of them throws away a working edge.
Why quitting feels so right
Losses hurt more than equal gains feel good. In Tversky and Kahneman's 1992 work on prospect theory, they estimated that a loss weighs about 2.25 times as much as a gain of the same size. A drawdown is a slow accumulation of that pain. On top of it, recent events tend to feel more meaningful than older ones, so the last ten trades loom larger than the four hundred before them. At the bottom of a drawdown, the emotional evidence that "it has stopped working" is overwhelming, while the statistical evidence is close to zero.
The same instinct also drives the quieter version of quitting: skipping the next signal "until things settle down." Since nobody knows which trade ends a losing streak, skipping signals during a drawdown removes trades that are exactly as good as any others, including the winners that end the streak. That is one of the habits we priced in Nº 002.
Normal or broken? Decide before it happens
Strategies do stop working. Markets change, edges get crowded, costs rise. The answer is not to ignore every drawdown. It is to decide in advance, with numbers, what counts as abnormal.
There is a humbling reason to do this carefully. Small edges are hard to measure. Each trade in this strategy swings by about 1.5R around its +0.12R average, so even 450 trades pin the edge down only roughly: in our runs, the measured three-year average ranged from about −0.01R to +0.26R (the middle 95%), and 3.5% of runs measured an edge of zero or less, for a strategy that genuinely has one. Your backtest's edge is an estimate, not a fact, and any kill switch has to allow for that, in both directions.
- Know your strategy's normal range before going live. A proper backtest, followed by a Monte Carlo run like this one on its trade results, tells you the losing streaks and drawdowns to expect. Here: a 10-loss streak and a 20% drawdown are typical; 15 losses or a 30% drawdown are unusual but possible.
- Write down your kill switch while you are calm. For this particular strategy over this horizon, an illustrative rule might be: "Stop and review if the drawdown exceeds 40% (deeper than 99% of simulated runs), or if after 200 trades the average result is below −0.06R per trade (worse than 95% of simulated runs)." The thresholds come from the simulation, not from how a losing week feels. Be realistic about what such a rule can do: if this strategy's edge had quietly vanished (no edge before costs), the 200-trade test would catch it only 62% of the time, and checking it again and again raises the chance of a false alarm. No rule is sharp. A written rule is simply far better than a feeling.
- Size so the normal range is survivable. If a 20% drawdown would make you quit, a strategy that typically produces one is too big for you at that size. Halving the risk per trade roughly halves the drawdowns, provided costs shrink with position size, which flat per-order brokerage does not (see Nº 004 on position sizing).
- Review on a schedule, not on pain. Check the strategy against its expected statistics monthly or every 50 trades, not every time it loses.
- Take mood out of the execution. A drawdown is exactly when human execution degrades: skipped signals, smaller sizes, early exits. Automated execution keeps taking every trade the rules call for, so the rules, not your mood, decide what happens next, including when the kill switch fires.
The point of a backtest is not only to find out whether a strategy makes money. It is to find out what making that money will feel like, so that when the ten-trade losing streak arrives, as it did in two runs out of three, you recognise it as the price of the edge, not the end of it.
Methodology
- Runs
- 10,000
- Trades per run
- 450 (about 3 years at 150 a year)
- Strategy
- 40% wins at +2R, 60% losses at −1R
- Risk per trade
- 1% of current equity
- Costs
- 0.08% of equity per round trip
- Drawdown
- peak-to-trough on closed-trade equity
- Quit rule
- first time drawdown reaches the threshold
- RNG
- mulberry32, seed 20261018 (Nº 002 uses a different seed, so its figures differ slightly, e.g. 94.5% vs 94.6% profitable)
Trade outcomes are drawn independently, so every streak and drawdown here is pure variance around a known, positive edge. Real strategies can have clustered losses, which makes streaks and drawdowns longer than shown. Streaks and drawdowns also grow with the number of trades, so these figures are specific to a 450-trade horizon. "Longest stretch without a new high" counts an unrecovered drawdown at the end of a run. Costs of 0.08% of equity equal 0.08R at 1% risk. "Months" assumes 150 trades a year. Losses are exactly −1R, with no gaps through the stop. Every number, including the same-timing comparison, the edge-uncertainty range and the kill-switch check, comes from sim-streaks.js (plain Node.js: node sim-streaks.js).