Skip to content
Menu
Finabase Learn

Trading · Risk management

Losing streaks and drawdown recovery: what a trader should expect

How long a losing streak to expect from a win rate, how deep a profitable system falls at each risk per trade, and how many trades it takes to climb back.

On this page
  1. How long a losing streak to expect
  2. From streaks to money
  3. Drawdowns are deeper than streaks
  4. The way back
  5. Deciding in advance what is normal
  6. What the numbers assume
  7. How Finabase does it
  8. Frequently asked questions
  9. Summary
  10. Check your understanding

A system that wins half the time will, in the course of a hundred trades, almost certainly lose five in a row and has a better than even chance of losing six. That is not a sign that it has stopped working. It is what a 50% win rate looks like.

Most traders abandon a method during a losing streak, and most of them do it without knowing whether the streak was unusual. This lesson gives the numbers to check against. It works out how long a streak to expect from a win rate, how deep a profitable system's drawdowns run at different risk levels, and how long the climb back takes. The aim is to know in advance what ordinary bad luck looks like, so that it can be told apart from a broken method.

How long a losing streak to expect

Assume each trade wins with the same probability and that one result does not affect the next. Then the chance of a streak of a given length anywhere in a series can be calculated exactly. The figure does so for a series of 100 trades.

Chance of at least one losing streak of a given length in 100 trades 0% 20% 40% 60% 80% 100% 2 3 4 5 6 7 8 9 10 11 12 13 14 15 win rate 35% win rate 45% win rate 55% win rate 65% Length of the streak (consecutive losing trades) Probability, assuming independent trades
Probability of at least one run of consecutive losses of each length within 100 independent trades, by win rate. Exact calculation by Finabase.
Win rate 5 or more in a row 6 or more 7 or more 8 or more 10 or more Expected longest streak
35% 100% 96% 85% 69% 37% 9.1
40% 98% 87% 69% 49% 20% 7.9
45% 92% 73% 50% 31% 10% 6.9
50% 81% 55% 32% 17% 4% 6.0
55% 65% 36% 18% 8% 2% 5.3
60% 46% 21% 9% 4% 1% 4.6
65% 29% 11% 4% 1% 0% 4.0

The table should change how a streak is read. A trend-following method that wins 40% of its trades and makes its money on large winners will show a streak of seven losses in about two out of three hundred-trade stretches. A trader who has decided to stop after five straight losses will stop a 40% system in nearly every such stretch.

More trades, longer streaks

The longest streak grows with the length of the record, though slowly. A rule of thumb from the mathematics of runs is that the longest losing streak in n trades is about the logarithm of n × (win rate) taken to the base 1 ÷ (loss rate). Exact values:

Win rate Expected longest streak in 100 trades in 250 trades in 1,000 trades
40% 7.9 9.7 12.4
50% 6.0 7.3 9.3
60% 4.6 5.6 7.1

Ten times as many trades adds about three losses to the expected worst streak at a 50% win rate. An active trader who places a thousand trades in a year should expect, at some point in that year, a streak that a hundred-trade backtest never showed.

What a streak does not mean

After six straight losses the chance that the next trade wins is still the win rate. The market does not owe a win, and the feeling that one is "due" is the error Tversky and Kahneman called belief in the law of small numbers: expecting a short run to look like the long-run average. Raising the bet because a win must be coming is the same error with money attached.

The opposite error is just as common: reading the streak as proof the method is dead. A streak is weak evidence either way. Stronger evidence is in the measures that use every trade, the expectancy and its sampling error (Edge and expectancy: the average trade, worked on real trade lists).

From streaks to money

A streak of losses is a drawdown: a fall from the account's highest value. The depth depends on the risk per trade. With a fixed percentage risked each time (Position sizing: how many shares to buy for a set risk), k losses in a row cost 1 − (1 − risk)^k of the account.

Risk per trade After 5 losses After 8 losses After 10 losses After 15 losses
0.5% 2.5% 3.9% 4.9% 7.2%
1% 4.9% 7.7% 9.6% 14.0%
2% 9.6% 14.9% 18.3% 26.1%
3% 14.1% 21.6% 26.3% 36.7%
5% 22.6% 33.7% 40.1% 53.7%

Put the two tables together. At a 40% win rate a streak of eight losses has about an even chance of appearing in 100 trades. At 1% risk it costs 7.7% of the account. At 5% risk it costs a third.

Drawdowns are deeper than streaks

The worst drawdown of a system is usually not one unbroken streak. It is a longer stretch in which losses outnumber wins: lose four, win one, lose three, win one, lose five. The account keeps falling even though no streak is longer than five.

To measure that, the figure simulates a system with a real edge: 45% of trades win 1.5R, the rest lose 1R, an expectancy of +0.125R per trade. Each run is 250 trades, roughly a year of one trade per session, and there are 5,000 runs.

Deepest drawdown in 250 trades of a profitable system, by risk per trade 0.5% risk: typical 6.4% 0.5% risk: 1 in 10 10.3% 1.0% risk: typical 12.5% 1.0% risk: 1 in 10 19.9% 2.0% risk: typical 23.9% 2.0% risk: 1 in 10 36.4% 3.0% risk: typical 34.3% 3.0% risk: 1 in 10 50.3% 5,000 simulated runs: 45% winners at 1.5R, losers 1R (expectancy +0.125R). "1 in 10" is the 90th percentile.
Deepest fall from a peak within 250 trades of a system that wins 45% of trades at 1.5R (expectancy +0.125R), by risk per trade. "Typical" is the median of 5,000 simulated runs and "1 in 10" the 90th percentile. Simulation by Finabase.
Risk per trade Typical worst drawdown One run in ten was worse than One run in a hundred was worse than
0.5% 6.4% 10.3% 14.7%
1% 12.5% 19.9% 27.5%
2% 23.9% 36.4% 48.4%
3% 34.3% 50.3% 64.0%

Measured in units of risk with a constant bet size, the typical worst drawdown was 13R, one run in ten exceeded 21.5R and one in a hundred exceeded 31.5R. The expected longest losing streak for this system over 250 trades is about eight. The worst drawdown was half as deep again as the worst streak, in a system that makes money.

The run that sits at the one-in-ten mark is not a failed system. It has the same edge as every other run. This is the central difficulty of judging a method from its own results: a profitable system and an unprofitable one can look the same for a long time.

Time under water

Depth is half of the experience. The other half is duration. In the same simulation the longest stretch between one equity high and the next had a median of 73 trades, and in one run out of ten it exceeded 159 trades. At one trade a day, a typical year with this profitable system includes about three and a half months without a new high, and an unlucky year includes more than seven.

The way back

Two pieces of arithmetic govern recovery.

The percentage needed is larger than the percentage lost. After a fall of d, the gain needed to return to the peak is d ÷ (1 − d). A 10% drawdown needs 11.1%; 20% needs 25%; 50% needs 100%. Maximum drawdown (MDD): the worst fall from a peak, and the way back sets out the full table.

The expected time is the hole divided by the edge. A system that earns an average of E per trade climbs out of a drawdown of D in about D ÷ E trades.

Cutting size in a drawdown

A common rule is to reduce risk per trade after the account has fallen by a set amount, for example to halve it after a 10% drawdown. The arithmetic is a trade.

  • What it buys. If the fall continues, each further loss costs half as much. With percentage sizing this happens automatically to a small degree; the rule makes it much stronger. If the method really has stopped working, the damage is limited while that becomes clear.
  • What it costs. The recovery is made with smaller bets than the decline. Halving the size doubles the number of trades needed to regain the peak, about 160 instead of 80 in the example above.

The rule is therefore insurance against the possibility that the edge is gone, paid for by a slower recovery when it is not. It is reasonable for the same reason that a fraction of the Kelly bet is reasonable: the edge is an estimate, and a deep drawdown is evidence, weak but real, that the estimate was too high.

Increasing size to recover faster is the opposite trade and has the opposite result: it speeds recovery when luck turns and deepens the hole when it does not, which is the path to the outcomes described in Risk of ruin: the chance a trading account does not survive.

Deciding in advance what is normal

The practical use of these tables is to write limits before they are needed.

  1. Know the system's expected streak. From the win rate, read the streak that has an even chance of appearing in the next hundred trades.
  2. Know its expected drawdown. A backtest gives one maximum drawdown, which is a single sample. A simulation of the kind above gives the range. Expect to see the one-in-ten figure at some point.
  3. Set the stop-trading level beyond that range. A drawdown twice the simulated one-in-ten level is far stronger evidence of a broken method than five straight losses.
  4. Write down the trades. A journal that records whether each loss followed the rules separates bad luck from bad execution (Trading plan and trading journal: what to write down).

What the numbers assume

Each of these makes real streaks and drawdowns worse than the calculated ones, so the tables are best read as the mildest version of what to expect.

How Finabase does it

  • Backtest. The Backtest tool lists every trade of a signal on one symbol in a transactions table, with wins, losses and the win rate. Reading down that table in date order shows the actual losing streaks. Each signal trades a fixed $10,000 with no costs.
  • Paper trading. The paper account keeps a value history and a transaction history (Practise with Finabase paper trading on the web). Its value chart is a drawdown record of your own decisions, which a simulation cannot provide.
  • Symbol pages. A symbol page shows how far the price is below its 52-week high, the drawdown of the asset itself.

Finabase has no streak or drawdown simulator. The figures on this page come from the script named in the sources.

Frequently asked questions

How many losing trades in a row is normal?

It depends on the win rate and the number of trades. For a 50% win rate, a streak of five or six losses is expected in every hundred trades, and seven or more has about a one-in-three chance. At a 40% win rate, eight in a row has about an even chance.

Does a long losing streak mean my strategy stopped working?

Not by itself. Compare it with the streak the win rate predicts. A streak inside the expected range says little. A drawdown far beyond what the system's simulated range allows, or a win rate that stays well below the tested one over many trades, is better evidence.

Should I reduce position size during a drawdown?

Reducing size limits further losses if the method has failed and slows the recovery if it has not. It is a form of insurance. Raising size to recover faster increases the chance of a much deeper loss.

How long does it take to recover from a drawdown?

On average, the size of the drawdown divided by the system's expectancy per trade. A 10R drawdown takes about 80 trades at +0.125R per trade. In percentage terms the gain needed is larger than the loss: 25% after a 20% fall.

Summary

Losing streaks follow from the win rate. In 100 trades a 50% system should expect to lose six in a row and a 40% system eight, and the worst streak lengthens slowly as the record grows. Drawdowns run deeper than streaks, because losses cluster without being consecutive: a profitable system simulated here typically fell 13R from a peak within 250 trades and 21.5R in one run of ten. Risk per trade turns those units into a share of the account, and the expectancy sets the time to recover. Knowing these ranges before they arrive is the only reliable way to tell an ordinary bad stretch from a method that has failed.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. A system wins 40% of its trades. Over the next 100 trades it loses seven in a row. What is the most reasonable reading?

    1. The system has stopped working, because seven straight losses is extremely rare
    2. This is within the expected range: a 40% system has about a 69% chance of a streak of seven or more in 100 trades
    3. A win is now due, so the next trade should be larger
    Show the answer

    B. This is within the expected range: a 40% system has about a 69% chance of a streak of seven or more in 100 trades For a 40% win rate the expected longest losing streak in 100 trades is about eight. A streak inside that range is weak evidence about the system, and the next trade still wins with the same 40% probability.

  2. A trader risks 2% of the current balance per trade and loses 10 trades in a row. About how far is the account below its peak?

    1. 20%
    2. 18.3%
    3. 10%
    Show the answer

    B. 18.3% With percentage risk each loss takes 2% of what is left: 1 − 0.98^10 = 18.3%. It is a little less than 10 × 2% because the bets shrink as the account falls.

  3. Why is a system’s worst drawdown usually deeper than its longest losing streak?

    1. Because drawdowns include trading costs and streaks do not
    2. Because a drawdown can span several losing runs separated by a few wins, without ever returning to the previous peak
    3. Because drawdowns are measured in percent and streaks in trades
    Show the answer

    B. Because a drawdown can span several losing runs separated by a few wins, without ever returning to the previous peak Lose four, win one, lose three, win one, lose five: no streak exceeds five, yet the account keeps falling. In the lesson’s simulation the typical worst drawdown was 13R while the expected longest streak was about eight losses.

  4. A system has an expectancy of +0.2R per trade and is in a 12R drawdown. Roughly how many trades should recovery take on average, at an unchanged bet size?

    1. 12
    2. 24
    3. 60
    Show the answer

    C. 60 Expected recovery time is the drawdown divided by the expectancy: 12 ÷ 0.2 = 60 trades. Halving the bet size during the drawdown would double that.

Sources and further reading

  1. Mark F. Schilling, “The Longest Run of Heads,” The College Mathematics Journal 21(3), 1990
  2. William Feller, An Introduction to Probability Theory and Its Applications, vol. 1, 3rd ed. (Wiley, 1968), ch. XIII — success runs
  3. Amos Tversky and Daniel Kahneman, “Belief in the Law of Small Numbers,” Psychological Bulletin 76(2), 1971
  4. Finabase calculation and simulation — exact streak probabilities; 5,000 runs of 250 trades, fixed seed (scripts/learn/figures/trading-risk-depth.ts)