Glossary
Losing streaks
Last reviewed: 14 September 2026·Tradelyze
A losing streak is a run of consecutive losing trades with no winning trade in between. Long streaks are normal over many trades. At a 50% win rate, this page's calculation gives a 17.0% chance of at least one run of 8 losses in 100 trades, and 62.5% in 500. One streak rarely proves a strategy is broken.
In plain English
Flip a fair coin 500 times. By this page's calculation, there is an 86% chance that heads comes up at least 7 times in a row somewhere. Nothing is wrong with the coin; that is simply what hundreds of flips look like. Trades behave the same way. The more trades you take, the longer the worst run of losses tends to be. The real question is whether your streak is longer than your strategy's own numbers allow.
New to this? Start with win rate and expectancy, then risk of ruin.
What is a losing streak in trading, and why does it matter?
A losing streak is a run of consecutive losing trades with no winning trade in between. Its length is counted in trades: three losses followed by a win is a streak of 3. A break-even trade needs a rule of its own; this page counts only trades that lost money. Win rate is the share of closed trades that made money, and the loss rate is the share that lost.
Losing streaks matter for money in two ways. The first is arithmetic: a string of full losses eats an account's loss allowance faster than any other sequence of trades. That is what ends prop firm challenges, where a firm closes the account at a fixed loss. The second is psychological. A trader in the middle of a streak often stops following the strategy, or changes its settings, just before the trades that would have recovered the loss.
Both problems come from the same mistake: expecting shorter streaks than the numbers produce. People are poor judges of randomness. Mark Schilling describes a classroom experiment, which he credits to Révész, in his 1990 paper in The College Mathematics Journal. Students who invented 200 “coin flips” could be told apart from students who really flipped, around 85% of the time in Schilling's own courses. The giveaway was that the invented sequences had runs that were too short. Schilling's own summary is that “human beings make rather poor randomization devices.”
How long a losing streak should you expect at your win rate?
The losing streak to expect depends on two numbers: the win rate and how many trades are taken. The table gives both views. The typical longest streak is the median: the longest losing streak is that long or shorter at least half the time. The other columns give the chance of at least one streak of 5, 8 or 10 losses anywhere in the sample. A streak of 8 here means 8 or more losses in a row.
| Win rate | Trades | Typical longest streak | 5+ in a row | 8+ in a row | 10+ in a row |
|---|---|---|---|---|---|
| 40% | 100 | 7 | 97.6% | 49.0% | 20.5% |
| 40% | 250 | 9 | 99.99% | 82.5% | 45.2% |
| 40% | 500 | 11 | >99.99% | 97.1% | 70.5% |
| 50% | 100 | 6 | 81.0% | 17.0% | 4.4% |
| 50% | 250 | 7 | 98.6% | 38.4% | 11.2% |
| 50% | 500 | 8 | 99.98% | 62.5% | 21.5% |
| 60% | 100 | 4 | 45.9% | 3.6% | 0.6% |
| 60% | 250 | 5 | 79.2% | 9.2% | 1.5% |
| 60% | 500 | 6 | 95.8% | 17.7% | 3.0% |
Three readings stand out. First, win rate moves streak risk sharply. Over 250 trades, a streak of 8 has an 82.5% chance at a 40% win rate and a 9.2% chance at 60%. Second, trade count matters almost as much. At a 50% win rate, the chance of a streak of 10 rises from 4.4% in 100 trades to 21.5% in 500. Third, a streak of 5 is close to certain for most strategies that take a few hundred trades.
What the table assumes
Every figure assumes independent trades: the result of one trade does not change the odds of the next, and the win rate stays the same throughout. Real losses can bunch together when market conditions change. When they do, long streaks usually come more often than the table shows. Treat the table as the optimistic case for a strategy with that win rate.
A lower win rate is not automatically worse. Take a strategy that wins 40% of its trades and makes twice its loss on each win. It earns as much per trade as one that wins 60% at even money. Both average a profit of 0.2 times the amount risked per trade. What the lower win rate does bring is longer streaks, so it needs more room in the account. Expectancy explains how win rate and payoff combine. Risk of ruin by risk per trade shows how those longer streaks raise the chance of hitting a loss limit.
How do you calculate the probability of consecutive losing trades?
The probability of consecutive losing trades has two versions, and they give very different answers. The chance that a particular set of trades all lose is the loss rate multiplied by itself once per trade. At a 50% win rate, the next 8 trades all lose with a chance of 0.5 to the 8th power, which is 0.39%, or about 1 in 256. The chance of at least one such streak somewhere in a long sample is much higher, and it needs a different calculation.
This page computed the “somewhere” version exactly with a run-length recursion. The idea is to walk through the trades one at a time and track how many losses in a row the sequence currently ends with. A win resets that count to zero. A loss adds one to it. Any sequence whose count reaches the streak length is set aside. Whatever probability is left after the last trade is the chance of never having that streak.
St(j) = chance that after t trades there is no streak of k yet, and the last j trades lost
start: S0(0) = 1
after a win: St(0) = p × [St−1(0) + … + St−1(k−1)]
after a loss: St(j) = q × St−1(j−1), for j = 1 to k−1
chance of at least one streak of k = 1 − [Sn(0) + … + Sn(k−1)]
Mark Schilling's 1990 paper derives the same exact distribution with a counting recursion. It covers a fair coin and a coin that lands heads more or less than half the time. For the 50% rows, this page checked the recursion above against Schilling's fair-coin formula, and every figure matched exactly. All the table's figures were also checked against a simulation of 100,000 random trade sequences, which agreed to within a few tenths of a percentage point.
Schilling also gives a quick estimate for the typical longest run. Translated to trades, the typical longest losing streak is about the logarithm of trades × win rate, taken in base 1 ÷ loss rate. In a spreadsheet, that is:
Worked example. A strategy wins 50% of its trades and takes 500 trades. The estimate is the base-2 logarithm of 500 × 0.5 = 250, which is about 8.0. The exact median in the table is also 8. For a fair coin, Schilling adds a rule of thumb: the longest run is very likely within three either way of that estimate, rounded to the nearest whole number. He puts the chance at 95.3% for 200 tosses. For 500 trades at a 50% win rate, this page's calculation puts a longest streak of 5 to 11 at 94.2%. The same formula shows that, at a 50% win rate, doubling the number of trades adds only about one trade to the typical longest streak.
Why is a long streak over hundreds of trades far more likely than on any single run?
A long losing streak is far more likely over hundreds of trades than on any single run. The reason is that a long sample holds hundreds of places where a streak could start. At a 50% win rate, any one stretch of 8 trades all loses with a chance of only 0.39%. But a 500-trade backtest contains 493 overlapping stretches of 8 trades, and each one is another chance. By this page's exact calculation, at least one of them is a full losing streak 62.5% of the time.
This matters when you judge a streak. A trader with a 50% win rate who has taken 400 trades and then loses 8 in a row often reads it as a 1-in-256 event. It is not. The right comparison is the chance of such a streak somewhere across all the trades taken, and that chance is large. The same logic applies to a backtest: over a long test, the longest losing streak on record is expected to be long.
How do losing streaks turn into daily loss limit and trailing drawdown breaches?
Losing streaks turn into prop firm breaches because a firm's loss allowance is a fixed number of dollars, and each full losing trade uses a fixed slice of it. Dividing the allowance by the loss on one full losing trade gives the number of losses in a row the account can take. A drawdown is a fall from an earlier high in account value. A trailing drawdown limit sits a fixed dollar distance below the account's highest balance and moves up with it. A daily loss limit caps what the account may lose in one trading day.
Worked example. Take a constructed $50,000 challenge account with a $2,000 trailing allowance, traded by a strategy that wins 55% of its trades. At $500 per full loss, including costs, 4 losses in a row use up the allowance. By this page's calculation with independent trades, at least one streak of 4 comes up 69.9% of the time in 50 trades and 91.5% in 100. Halve the loss to $250 and the account can take 8 in a row. At least one streak of 8 then comes up 4.0% of the time in 50 trades, 8.4% in 100 and 20.3% in 250.
| Loss per full losing trade | Losses in a row to use it up | Chance in 50 trades | Chance in 100 trades | Chance in 250 trades |
|---|---|---|---|---|
| $500 | 4 | 69.9% | 91.5% | 99.8% |
| $250 | 8 | 4.0% | 8.4% | 20.3% |
The streak count is the fastest route to a breach, not the only one. At $500 per trade, the sequence loss, loss, win, loss, loss, loss adds up to −$2,000 with no streak longer than 3, if the win makes $500. A win that does not make a new high does not move a trailing floor back down. Slippage and gaps past the stop make single losses bigger than planned. So treat the count as the most losses in a row you can survive, and expect the real margin to be thinner. Risk of ruin covers the mixed sequences that the streak count leaves out.
A daily loss limit adds a second count: how many full losses fit inside one session. A $1,000 daily allowance at $250 per full loss fits 4. A strategy that never takes more than 3 trades a session cannot lose 4 full stops in one day. Its daily risk comes from losses bigger than the stop. A strategy that takes 10 trades a session can. Firms also differ on whether open losses count and when the day resets. Which backtest statistic predicts a daily loss limit breach? shows what to measure in a backtest. How a profitable account fails a trailing drawdown shows why a streak late in a challenge can still end it.
Check this before you pay a challenge fee
Work out how many losses in a row your allowance takes at the size you plan to trade. Find the row in the streak table closest to your win rate. Read the chance of a streak that long over the number of trades the challenge will need. If it is high, size down before you change the strategy. Position sizing for prop firm challenges turns a dollar allowance into a trade size.
Does Monte Carlo simulation capture streak risk better than the original backtest?
Monte Carlo simulation can capture streak risk better than the original backtest, but only when it keeps losing runs together. A backtest shows one order in which the trades happened, so it shows one longest streak and one maximum drawdown. Monte Carlo simulation rebuilds the trade list many times in different orders and mixes, then reports the range of drawdowns those versions produce. That range shows how much deeper the worst stretch could have been with less luck.
The method decides whether streaks survive. Shuffling trades one at a time, as if each trade had no connection to the next, keeps the number of losses but scatters clusters of them. If the real losses bunched together, a one-at-a-time shuffle understates streak risk. A block bootstrap instead redraws short runs of consecutive trades, so a losing streak stays a streak. Because it draws those runs with replacement, some runs repeat and others drop out. Two losing runs can land next to each other and form a longer streak than the backtest contained. Does trade shuffling understate risk? works through the difference.
Tradelyze's Monte Carlo check uses a stationary block bootstrap, 1,000 draws by default. When there are enough trades to measure it, the typical run length is set from how strongly your trades cluster, and the drawdown figures use longer runs still. The robustness card reports two numbers that reflect streak risk:
- MC Max DD Real→P95. The first number is the maximum drawdown of your trades in their real order. The second is the 95th-percentile drawdown: the depth that 95 of every 100 simulated versions stayed within.
- Ruin Probability. The share of simulated versions whose drawdown went past the maximum total drawdown limit of the prop firm or custom rule set you selected. It must be below 20% for the Monte Carlo check to pass, a line Tradelyze chose rather than a published standard. It is blank when the rule set has no total drawdown limit.
Neither number is a streak length. Both measure the drawdown that streaks and mixed losing runs produce, which is what a firm's limit reacts to. The simulated versions carry no calendar, so daily loss limits are not simulated; Tradelyze checks them on the real trade order in each firm's Rule Results table. The trades come from the chosen settings over the same history the optimizer searched, so treat both figures as optimistic. Ruin Probability and MC Max DD Real→P95 explains both in full.
How do you tell a normal losing streak from a strategy that has stopped working?
Tell a normal losing streak from a broken strategy by comparing the streak with what the strategy's own numbers allow, using limits you wrote down before trading. A losing streak on its own is weak evidence. This page sets no cutoff for “broken”, because no primary source sets one. Tradelyze does not set one either. The steps below give the comparisons; the cutoffs are yours to decide in advance, because mid-streak every loss looks like proof.
- Before going live, record the backtest's own figures. Note its win rate, its trade count, its longest losing streak and its maximum drawdown. If you ran a Monte Carlo simulation, note the 95th-percentile drawdown too. That is the second number in Tradelyze's MC Max DD Real→P95.
- Rule out execution first. Run the backtest over the live dates and compare its trades with the ones you took. If the trades differ, the problem is fills, costs or code, not a dead edge. Is it overfitting or execution? sets out the signs of each.
- Compare the streak with the numbers, not with the backtest's longest streak. Use the win rate and the number of live trades taken so far, as in the example below.
- Compare the drawdown with the Monte Carlo range. A live drawdown still inside the 95th-percentile depth is within what the trades allowed. A drawdown well past it, or past your firm's limit, is the point to stop and investigate.
- Do not re-tune because of the streak. Settings tuned until a few bad weeks turn profitable fit recent noise. Why re-optimizing until it passes is a trap explains the problem.
Worked example. A constructed backtest has 300 trades, a 55% win rate and a longest losing streak of 6. After 80 live trades, the strategy loses 7 in a row, a streak longer than anything in the backtest. It feels like proof, but the numbers say otherwise. At a true 55% win rate, this page's calculation gives a 14.5% chance of at least one streak of 7 in 80 trades. If the win rate had fallen to 45%, the chance would be 41.9%. The streak is nearly three times as likely under the worse win rate, yet still common under the original one.
The backtest's own streak was never a ceiling either. By the same calculation, a strategy that wins 55% of 300 trades has a 46.0% chance of at least one streak of 7. A longest streak of 6 in that backtest was simply typical, and Schilling's estimate for 300 trades at 55% comes out at about 6.4.
A live win rate is noisy too
Eighty trades is a small sample, so the live win rate itself moves a lot from luck. How many trades before a win rate means anything shows how wide that noise is. A strategy can also stop working for real, as markets change or other traders crowd the same trade. Retiring a strategy that sits clearly outside its own range is a sound decision. When to abandon a strategy instead of tuning it covers that decision.
How can you re-check a strategy in Tradelyze?
Tradelyze offers two ways to re-check a strategy, Re-Evaluate and Verify Trial, and neither reads a live trading account. Tradelyze does not report a longest-losing-streak figure, so start by counting the streak yourself. In the Top Trials table, a trial's Download Trades button saves that trial's trades as a CSV file with each trade's profit and loss. Keep the trades in order and count the longest run of losing trades. Compare that count with the streak table on this page.
Re-Evaluate is a button on a strategy whose run has completed or failed. Its tooltip reads “Re-run this strategy through the current pipeline. Costs 1 credit.” It re-runs the Pine Script, price data and trade list you already uploaded, repeats the previous run's parameter settings and replaces the earlier results. Its dialog opens with the last run's prop firms already ticked. It adds no new trades. It is a fresh check of the same history, not a test of the weeks that produced a live losing streak.
Verify Trial, in the Top Trials table, tests one trial's settings in TradingView. Its dialog says: “We'll create a copy of your Pine script with these trial parameters as defaults. Run it in TradingView, then upload the trade and market data CSVs.” Submitting that copy costs 1 credit, and the run checks whether Tradelyze's re-run matches your TradingView trades. What Verify Trial does has the details.
To test the period that produced a live streak, export the TradingView trade list and price data covering those newer dates, and submit them as a strategy. Then compare the new report's drawdown figures with the ones you recorded before going live. Credit costs are kept current in how credits are used. To judge the whole report, not one tile, use the pre-trade checklist, and for account questions see the Learn FAQ.
Check it on your own strategy
In a Tradelyze report, streak risk shows up in the robustness card's MC Max DD Real→P95 and Ruin Probability rows, and in each firm's Rule Results table. Tradelyze re-runs an uploaded TradingView Pine Script strategy on your price data and checks it against your exported trade list. It then runs parameter optimization, walk-forward analysis, a five-check robustness score and prop firm rule checks. It does not place trades, give financial advice or guarantee a challenge pass, and it is in beta.
Create an account. Already a user? Open your strategies.
Frequently asked questions about losing streaks
How many losing trades in a row is normal?
It depends on win rate and trade count. By this page's calculation with independent trades, the typical longest losing streak at a 50% win rate is 6 losses in 100 trades, 7 in 250 and 8 in 500. At a 40% win rate it is 7, 9 and 11; at 60% it is 4, 5 and 6. Losses that bunch together in bad market conditions can make streaks longer still.
What is the probability of 5 losing trades in a row?
That depends on the question. The chance that the next five trades all lose is the loss rate multiplied by itself five times: 3.1% at a 50% win rate. The chance of at least one run of five losses somewhere in 100 trades is 81.0% at the same win rate, by this page's calculation with independent trades. Over many trades, a run of five is expected.
How do I calculate the longest losing streak to expect?
Mark Schilling's 1990 College Mathematics Journal paper gives the approximation: the logarithm, in base 1 divided by the loss rate, of the number of trades times the win rate. In a spreadsheet, enter =LOG(trades*winrate, 1/(1-winrate)). At a 50% win rate over 500 trades it gives about 8. It assumes independent trades, so treat the answer as a typical value, not a maximum.
Does a long losing streak mean my strategy stopped working?
Not on its own. At a 55% win rate, this page's calculation gives a 14.5% chance of at least one streak of 7 losses in 80 trades. Compare the streak with what the backtest's win rate and trade count allow. Check live fills against the backtest over the same dates. Watch drawdown against a Monte Carlo 95th percentile you noted before going live.
Why is my live losing streak longer than anything in my backtest?
A backtest is one sequence of trades, so its longest streak is a single draw, not a ceiling. By this page's calculation, a strategy that wins 55% of 300 independent trades has a 46.0% chance of at least one 7-loss streak. A backtest whose longest streak was 6 was therefore typical. Live costs, worse fills and changed markets can also lower the win rate.
How many losing trades in a row can a prop firm account take?
Divide the dollar allowance by the loss on one full losing trade. A $2,000 trailing allowance takes 4 losses of $500, or 8 losses of $250. A win that does not make a new high does not reset a trailing drawdown, and slippage makes losses larger, so the true count can be lower. Compare the count with the streak table on this page.
What is the chance of a losing streak of 8 at a 60% win rate?
At a 60% win rate, at least one streak of 8 or more losses has a 3.6% chance over 100 trades, 9.2% over 250 and 17.7% over 500. That is by this page's exact calculation with independent trades. At a 40% win rate the same streak has a 49.0%, 82.5% and 97.1% chance. Win rate changes streak risk sharply.
Does Monte Carlo simulation account for losing streaks?
It depends on the method. Shuffling trades one at a time breaks up clusters of losses, so it can understate streak risk. A block bootstrap redraws runs of consecutive trades, so losing streaks stay together. Tradelyze's Monte Carlo check uses a stationary block bootstrap with 1,000 draws by default and reports MC Max DD Real→P95 and Ruin Probability.
Are trades really independent, as the streak table assumes?
Often not exactly. Losses can cluster when volatility or trend conditions change, and one market event can hit several trades at once. Clustering usually makes long streaks more likely than an independent-trades table shows, so read the table as the optimistic case for a strategy whose losses bunch together. A block bootstrap on your own trades keeps that clustering.
Does Tradelyze show the longest losing streak?
No. Tradelyze reports no longest-losing-streak figure. To count it, use a trial's Download Trades button in the Top Trials table. Keep the trades in order and find the longest run of trades with a negative profit and loss. The robustness card's MC Max DD Real→P95 and Ruin Probability show the drawdown that streaks and mixed losing runs produce.
What does Re-Evaluate do in Tradelyze?
Re-Evaluate re-runs a strategy whose run completed or failed through Tradelyze's current pipeline. It uses the Pine Script, price data and trade list already uploaded, repeats the previous run's parameter settings and replaces the earlier results. It costs 1 credit. It adds no new trades, so to test the dates of a live losing streak, submit exports that cover those dates.
Should I re-optimize a strategy after a losing streak?
Not because of the streak alone. Tuning settings until a few bad weeks turn profitable fits the most recent noise, on a smaller sample than the original backtest. First decide whether the streak falls inside what the backtest's win rate and Monte Carlo drawdown range allow. If it clearly does not, investigate the cause before changing any setting.
Sources
- Mark F. Schilling, “The Longest Run of Heads”, The College Mathematics Journal 21(3), May 1990, pages 196–207, Mathematical Association of America. doi:10.1080/07468342.1990.11973306; also JSTOR 2686886 — exact recursion formulas for the distribution of the longest run of heads, for a fair coin and for a coin with any probability of heads; the typical longest head run of about log1/p(nq), where p is the probability of heads and q of tails, which this page applies with losses in place of heads; the fair-coin rule of thumb that the longest run is very likely within three either way of the integer nearest log2(n/2), which he calculates at 95.3% for 200 tosses; the remark that “human beings make rather poor randomization devices”; and the classroom experiment, credited to Révész, in which invented coin sequences were identified around 85% of the time in Schilling's courses because their runs were too short.
- Tradelyze implementation, reviewed for this page — the Monte Carlo robustness check (a stationary block bootstrap with 1,000 draws by default, a run length read from how strongly the trades cluster when there are enough trades, and lengthened for the drawdown figures); Ruin Probability counted against the selected rule set's maximum total drawdown percentage, left blank when the rule set has none, and required to be below 20% for the check to pass; no longest-losing-streak figure in the report; each trade's profit and loss in a trial's Download Trades CSV; Re-Evaluate, offered when a run has completed or failed, re-running the stored Pine Script, price data and trade list with the previous run's parameter settings for 1 credit, opening with the last run's prop firms ticked and replacing the earlier results; and Verify Trial, which creates a draft copy of the Pine Script with the trial's parameters as input defaults, costing 1 credit on submission.
- Constructed calculations — the streak probability table, the typical longest streaks, Figure 1, the $2,000 allowance table and every worked example on this page were computed exactly for this page from an independent-trades model with the run-length recursion shown above. They were checked against Schilling's fair-coin recursion, with identical results, and against a simulation of 100,000 random trade sequences. They are not measured data from any strategy or firm.