Theory
A losing streak is normal
See how long a losing run can last in a sound plan, and why one trade or ten trades prove nothing either way.
~9 min read · 9 exercises · 250 points
Before you start
Seven in a row
Back on your Pebble practice account, and this time Marc is the one with a rough week: seven losses in a row, though his journal says he followed the plan every time. Nora is sure the plan is broken.
You'll work out what a sound plan's normal losing runs look like, how long they can get over many trades, and how to tell a normal streak from a plan that has really stopped working.
A sound plan can lose most of its trades
Before judging a losing streak, measure the plan in risk units: one unit is the risk per trade. On a 10,000 account at 1%, one unit is 100 dollars of risk.
Marc's plan wins 40% of its trades. A win makes 2 units, a loss costs 1, and costs take 0.1 unit from every trade, so a win nets +1.9 and a loss −1.1. Its expectancy, the average result per trade after costs (from Expectancy: the average trade in module 4), is 0.4 × 1.9 − 0.6 × 1.1 = 0.76 − 0.66 = +0.10 unit, about +10 dollars a trade.
That is small but positive. The plan earns on average while losing 6 trades in 10. So losing often is built into it: a sound plan with a 40% win rate is supposed to produce long runs of losses. The rest of this lesson asks how long those runs get.
- Do"40% winners at +1.9 / −1.1 units after costs: +0.10 unit a trade. Losing 6 in 10 is expected."
- Don't"It loses more often than it wins, so it must be a bad plan."
Win rate alone says nothing without the size of wins and losses.
How long a normal losing streak is
A losing streak is a run of losing trades in a row. Intuition says a sound plan rarely loses many times in a row. The numbers say otherwise.
Pebble's simulator played Marc's 40% plan 1,000 times, 100 trades each, and recorded the longest losing run in every run. The most common longest run was 7. A run of 6 or more losses happened in 86.5% of the runs, 8 or more in 47.9%, and 10 or more in 19.5%. Only 13.5% of runs got away with at most 4 or 5 in a row.
Dataset · 13 rows
Longest losing run in 1,000 simulated runs of 100 trades
Plan: 40% winners, +1.9 / −1.1 risk units after costs. `runs` = how many of the 1,000 had that longest losing run; `share_pct` = runs ÷ 10.
First 5 of 13 rows
So for this plan, Marc's seven-loss week is ordinary, not a sign of a broken plan. The lower the win rate, the longer the normal streaks; a plan that wins 40% will meet runs of 8 to 10 over a few hundred trades.
A useful habit follows: write down in the plan the streak you expect before it arrives, so the answer is on paper when it comes.
- Do"Expected: losing runs of 8 to 10 in a row. Review only after 30 more trades."
- Don't"Seven losses in a row? Then it's broken. Obviously."
Seven was the most common longest run for a sound plan.
A streak doesn't make a win due
After a long losing run, it feels as if a win must be "due". It isn't. The trades of a plan don't remember each other: the next trade of Marc's 40% plan wins with 40%, after seven losses just as after seven wins. Why independent events behave like this is covered in the statistics course; here we only use the result.
Two mistakes come from the "due" feeling. One is raising the size to win the streak back, which turns the next ordinary loss into a big one. The other is the reverse superstition, "it's cursed, skip the next one", which drops trades for no reason the plan knows about.
What a streak does tell you is about yourself: it is exactly the moment when people double the size or throw the plan away, the revenge-trading pull from Fear, greed, hope and revenge trading. The protection is a size that doesn't change with the streak.
- Do"Seven losses, all by the plan. Next trade: same 1% risk as always."
- Don't"A win is due, so this one goes in at double size."
The odds haven't changed; only the loss has grown.
The streak is the price of the plan
When a streak arrives, the useful question is not "is the plan broken?" but "what did the trades look like?". Open the journal first. If the trades were taken by the plan, the streak is simply the price of running it: a plan that wins 40% pays for its wins with runs of losses.
The sensible reply to Marc's seven-loss week has three parts. Name the streak as normal for this plan (seven was its most common longest run). Keep the risk per trade where it was, small, so the next normal loss stays small too. And leave the judgment of the plan to a proper sample, 30 trades or more, the review line of the one-page plan from A one-page trading plan.
Dataset: Longest losing run in 1,000 simulated runs of 100 trades · shown above
The tempting replies are the expensive ones: switching plans on the spot, doubling the size, or betting that a win is due.
- Do"Seven by the plan: normal for 40%. Same 1% risk; I'll judge it after 30 trades."
- Don't"New plan from tomorrow, and double size to get it back."
Two expensive reactions to one ordinary streak.
Two checks that can tell a broken plan
If a streak can't tell whether a plan is broken, what can? Two checks, and they work only together.
First: were the trades taken by the plan? The journal answers it (the habit from Practice account, starting small and the journal). If half the losing trades broke the rules, the streak is about execution, and judging the plan on them is pointless.
Second: after enough trades, is the average result after costs below zero? "Enough" means at least 30 trades, and even then the answer is a first estimate. If thirty trades taken by the plan average below zero after costs, the plan has a real problem.
Neither a long streak, a painful week nor a low win rate passes this test. Marc's plan loses 6 trades in 10 and still averages +0.10 unit; it would fail a "win more than you lose" test and pass the real one.
- Do"Journal: 30 trades, all by the plan, average −0.2 unit after costs. Time to stop and rethink."
- Don't"Six losses in a row, the plan is dead."
Neither check was done.
Ten trades can make one plan look brilliant or hopeless
Small samples swing in both directions, and that cuts against good starts as much as bad streaks.
The same simulator counted the wins in the first 10 trades of each of the 1,000 runs of Marc's plan, which truly wins 40%. In 17.8% of runs it showed 6 or more wins (11.6 + 5.7 + 0.5), roughly one run in six. In 17.2% it showed 2 wins or fewer (0.5 + 3.5 + 13.2). One plan, and after 10 trades it can look like a 60% plan or a 20% one.
Dataset · 11 rows
Wins in the first 10 trades, 1,000 runs of the same 40% plan
First 5 of 11 rows
Nora's 7 wins out of 10 fit this picture: it could be a good plan or plain luck, and ten trades can't tell the two apart. That's why a plan is judged on at least 30 trades, and even then with care. The statistics course explains why small samples are this noisy; here the simulation is enough.
- Do"7 of 10 so far. Too few to tell; same size, look again at 30."
- Don't"7 of my first 10 won, so I can triple my size."
A lucky start of a 40% plan looks exactly like this.
A first estimate needs 30 trades, not 10
A win rate measured on a handful of trades is mostly noise. The practical rule of this course: count at least 30 trades before you trust a win rate or an average result, and treat even 30 as a first estimate, not a verdict.
The rule works in both directions. A bad start doesn't prove the plan is broken; a good start doesn't prove it is good. Seven wins in ten trades are entirely possible for a plan that truly wins 40%: in the simulation, a 40% plan showed 6 or more wins in its first 10 trades in 17.8% of runs.
What to do meanwhile is boring and protective: keep logging every trade, keep the size small and the same, and look again when the count reaches 30. Raising the size after a lucky start is how a 40% plan with a good first week turns into a large loss in the second.
- Do"After 30 trades: 13 wins, 43%. First estimate; keep going."
- Don't"10 trades, 7 wins: my plan wins 70%."
One run in six of a 40% plan looks like this.
Size decides whether you survive a normal streak
Sound doesn't mean safe. In the simulation, 24% of the 1,000 runs of Marc's positive plan still ended 100 trades below zero: the average gain is small and luck is loud. And even a run that ends well can dig deep on the way. Run 6 started with 6 wins in 10, later lost 8 in a row, sat 16.8 risk units below its peak and still finished at +7.0 units.
Dataset · 100 rows
Run 6 of the simulation, trade by trade
`result_units` is after costs; `total_units` is the running total, both in risk units (one unit = 100 dollars of risk here).
First 5 of 100 rows
How much that fall hurts depends only on the size of a unit. At 1% risk per trade it is 16.8% of a 10,000 account (1,680 dollars); at 2%, 33.6%; at 5%, 84%. The streak is the same in all three; the risk per trade from module 3 decides whether it is a bad month or the end of the account.
How it fits together: losses come in normal streaks, small samples mislead both ways, and only a small fixed size lets you survive both. The next lesson turns this into rules that protect you from yourself: a checklist, a daily loss limit and breaks.
- Do"1% a trade: run 6's worst stretch is −16.8% of the account, painful but survivable."
- Don't"The plan is positive, so 5% a trade is fine."
The same normal stretch would take 84% of the account.
Ready to practise?
Answer the lesson's exercises and earn 250 points the first time you finish it.
Start the exercises