Part 1: Risk and its numbers · Chapter 3 of 14 · 11 min

Losing streaks

Losses in a row are arithmetic, not bad luck: with 50% of trades winning, in 100 trades there is an 81% chance of at least 5 consecutive losses and 17% of at least 8. In our series the longest is 7 losses (levels bounce, 21-24 July 2026), 6 (swing break) and also 7 in the example with 60% of trades winning. Chance has no memory: after five losses the next trade is worth the same as any other.

Losing in a row is normal

Flip a coin 100 times: the longest run of consecutive heads is usually 6, and one test in six reaches 8 or more. It does not happen because the coin is rigged, but because 100 flips contain many chances to start from zero. Trades work the same way: if your system wins 50% of the time, losing 5 times in a row does not need bad luck, it is likely.

A losing streak is a stretch of consecutive trades that all lose. The number that matters to you is the longest streak you can expect in a given period, because that is the one that tests the account and you.

The real streaks

Let us count the losing streaks in the book's four series:

SeriesTradesLosingLosing streaksMean lengthLongestLongest winning
LA552281 (51%)1461.9279
LD367177 (48%)991.7968
SD5527 (49%)141.9344
SX (60% winning)552219 (40%)1301.6877

In LA, the 146 losing streaks have these lengths: 73 of a single loss, 42 of two, 13 of three, 9 of four, 6 of five, 2 of six, 1 of seven. Half last one trade; only 9 in 146 (6%) reach five or more. But that 6% did show up: the 7-loss streak of LA is from 21-24 July 2026 and cost 7.1 R. Even SX, with 60% of trades winning, had a 7-loss streak: a true edge does not prevent long streaks.

Figure 1. Cumulative R of series LA (bars) and LD (pink), with the two longest losing streaks highlighted: LD, 6 losses between 13 and 16 March 2026 (grey); LA, 7 losses between 21 and 24 July 2026 (blue).Data: gold, XAUUSDT perpetual future (real candles) · daily (D1) · from 15 Dec 2025 to 5 Oct 2026.
Read the chart in words

Daily gold chart with, below, two curves of cumulative R and two bands marking the days of the longest losing streaks.

WhenWhatPrice
12 Mar 2026 → 17 Mar 2026LD: 6 losses in a row
20 Jul 2026 → 25 Jul 2026LA: 7 losses in a row

Times in UTC.

Figure 2. The seven LA trades that hit the stop one after another (21-24 July 2026, UTC hours): 1 and 2 short, 3 to 6 long, 7 short. The numbers sit on the entry candles. The risk per trade went from 11.5 $ to 30.2 $; each loss cost between 1.01 and 1.03 R.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 20 Jul 2026 20:00 UTC to 24 Jul 2026 18:00 UTC.
Read the chart in words

Hourly gold chart of 20-24 July 2026 with seven numbers on the entry candles of the seven consecutive losing trades.

WhenWhatPrice
21 Jul 2026 02:00 UTCPoint 14,049.25
21 Jul 2026 15:00 UTCPoint 24,087.40
22 Jul 2026 16:00 UTCPoint 34,162.00
22 Jul 2026 22:00 UTCPoint 44,135.62
23 Jul 2026 11:00 UTCPoint 54,097.48
24 Jul 2026 00:00 UTCPoint 64,055.91
24 Jul 2026 06:00 UTCPoint 74,043.73

Times in UTC.

How long can it last

If results are independent, the probability of a streak of at least m losses in n trades can be computed exactly. For 100 trades:

Winning trades≥ 5 losses≥ 8 losses≥ 10 losses≥ 12 losses
30%99.9%85.8%58.0%32.9%
40%97.6%49.0%20.5%7.7%
50%81.0%17.0%4.4%1.1%
60%45.9%3.6%0.6%0.1%

With 500 trades the situation gets worse, and by no small amount:

Winning trades≥ 8 losses≥ 10 losses≥ 12 losses
30%100%99.0%88.4%
40%97.1%70.5%35.1%
50%62.5%21.5%5.8%
60%17.7%3.0%0.5%

A system that wins 40% of the time (with wins larger than losses, for example trend following) has a 70% chance of seeing 10 losses in a row in 500 trades. A system with a 30% success rate sees them almost for certain.

At a glance, the typical longest streak (median of 4,000 simulations, with the 5th to 95th percentile range in brackets):

Winning tradesIn 100 tradesIn 500In 1,000
30%10 (6-17)15 (11-22)16 (12-24)
40%7 (5-12)11 (8-16)12 (9-17)
50%6 (4-9)8 (6-12)9 (7-13)
60%4 (3-7)6 (5-9)7 (5-10)

A quick rule: the longest streak is about the number k for which you expect to find such a streak just once, that is n × p × (1 − p)^k ≈ 1, with p the share of winning trades. With p = 0.4 and n = 500, k ≈ log(500 × 0.4) / log(1/0.6) ≈ 10-11: the table gives 11. The approximation is handy for estimating in your head, but it holds if results are independent.

What it costs in the account

The cost of a losing streak depends on how much you risk per trade. With risk computed each time on the remaining capital, after k losses in a row the account is worth (1 − f)^k of the initial capital:

Risk per tradeAfter 5 lossesAfter 8After 10After 15After 20
1%95.1%92.3%90.4%86.0%81.8%
2%90.4%85.1%81.7%73.9%66.8%
5%77.4%66.3%59.9%46.3%35.8%

Let us put the two tables together. A system with 40% of trades winning and 2% risk: in 500 trades the typical longest streak is 11 (up to 16), and 11 losses at 2% leave about 80% of the account (−20%); 16 losses leave 72%. At 5% the same streak of 11 leaves 57%.

A nuance: counting losses is not enough, what they cost matters. In series SD the 4 losses in a row (30 June-30 July 2026) cost 1.1 R in all, because the losses of that system are small (0.35 R on average) and time exits are sometimes profitable. The number of losses is read together with their size in R.

Chance has no memory

After a streak of losses the temptation is strong to believe that "a winner must come now". If results are independent, the probability of winning after a streak of losses does not change. Let us check in the real series, looking at the share of winning trades right after one, two, three losses in a row (cases in brackets):

SeriesWinning share overallAfter 1 lossAfter 2After 3
LA49%52% (280)54% (134)49% (61)
LD52%56% (176)64% (77)63% (27)
SX60%59% (219)66% (89)60% (30)

For LA and SX the differences are small and do not go in a single direction. For LD after two losses the share rises to 64% against 52% overall: with 77 cases the error on a share is about ±11 points, so 12 points more sits at the limit. Out of nine comparisons, one at the limit is no surprise, and the correlation between one result and the next (−0.053 for LA, −0.074 for LD, +0.015 for SX, with an error of ±0.08-0.10) includes zero. The number of consecutive stretches with the same sign is also what is expected for independent results: 292 observed against 277 ± 23 for LA, 198 against 184 ± 19 for LD.

The practical conclusion: after five losses the system has the same probabilities as before. Raising the size because "a winner is due now" is a bet on a memory that the data do not show. For systems with filters on market conditions, a losing streak can be a change of conditions: in that case what to check is the system, not luck, and chapter 13 explains how to decide in advance.

What to do with it

  1. Compute the longest streak you can expect with your share of winning trades and the number of trades in a year.
  2. Choose the risk per trade so that streak is bearable: 2% with 11 losses costs 20%, 5% costs 43%.
  3. Do not change size after a loss, neither to recover nor to quit. Size is decided beforehand, for all trades.
  4. Keep count of losses in R, not only in number.
  5. If the streak exceeds the 95th percentile of the table, stop to check the system, not the portfolio.

Procedure in four steps

  1. Estimate the share of winning trades from your series, with its error.
  2. Read from the tables the typical losing streak for the number of trades in a year, and the 95th percentile.
  3. Compute what that streak costs in the account, at the risk you chose.
  4. If the cost exceeds what you know you can bear (decide it beforehand, with a cool head), reduce the risk per trade, not the number of trades.

Exercises

Chapter exercises

1. The chart marks seven LA trades, all stopped out, between 21 and 24 July 2026. If each risked 1% of the current capital of a 10,000 $ account, what was left after the seventh?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 20 Jul 2026 20:00 UTC to 24 Jul 2026 18:00 UTC.
Read the chart in words
WhenWhatPrice
21 Jul 2026 02:00 UTCPoint 14,049.25
21 Jul 2026 15:00 UTCPoint 24,087.40
22 Jul 2026 16:00 UTCPoint 34,162.00
22 Jul 2026 22:00 UTCPoint 44,135.62
23 Jul 2026 11:00 UTCPoint 54,097.48
24 Jul 2026 00:00 UTCPoint 64,055.91
24 Jul 2026 06:00 UTCPoint 74,043.73

Times in UTC.

2. A system wins 50% of trades. In 100 trades, how likely is it to find at least 8 losses in a row?

3. A system with independent results lost five times in a row. What is the probability that the next trade is a winner?

4. A system wins 40% of trades. Over 500 trades, which losing streak is typical (median)?

Have a question or want to share your exercise?

Post in the community, or join the free signals room on Telegram.

Educational content, not financial advice. Trading involves risk.