Show My Trades
Metrics· Glossary

What Is a Z-Score in Trading? Streaks vs Chance

A trading z-score compares the streakiness of wins and losses against chance. How our Z-Score (Probability) is computed, and why 171 trades is a thin sample.

A z-score in trading measures whether wins and losses arrive in streaks or take turns, by counting the runs in the actual sequence and comparing that count with what chance would produce from the same wins and losses. It says nothing about whether the account made money, only whether one trade's outcome carried information about the next.

How it works

A run is an unbroken block of same-sign results. The sequence W W L W L L has four runs. The statistic is the Wald–Wolfowitz runs test:

Expected runs   R  = 2WL / N + 1
Variance        σ² = 2WL(2WL − N) / ( N² (N − 1) )
Z                  = (Runs − R) / σ

W = winning trades, L = losing trades, N = W + L

Trades are ordered by close time, and a trade counts as a win on its net result, with commission and swap already applied. A position that finished gross-positive and net-negative after costs is a loss here, and it breaks a winning run.

The sign is the reading:

  • Negative z: fewer runs than chance. Results cluster. Wins follow wins, losses follow losses.
  • Positive z: more runs than chance. Results alternate more than a coin would.
  • Near zero: indistinguishable from independent trades.

On the account page the score is followed by a percentage in brackets: the same score squashed onto a 0–100 scale, with 50% meaning exactly as random as a coin. Below 50% is streakier than chance, above 50% more alternating; a z of −2.41 shows as 8.24%, a z of +1.00 as 73.11%. It is a readable restatement, not a p-value, so 8.24% is not a significance level.

Why it matters

Strategy statistics almost always assume trades are independent. When the z-score says otherwise, other numbers on the page quietly stop meaning what they appear to mean. Losses that cluster produce a deeper drawdown than an equivalent win rate suggests, because drawdown depends on how the losses were arranged, not just how many there were.

A strongly negative score has a familiar source. Grid, averaging-down and basket systems open many positions around one idea and close them together, so a single decision emits a long block of same-sign results. Anything that scales into a losing position does the same, and the account that produces it usually also shows a wide gap between equity and balance drawdown.

A strongly positive score usually means the account is not running one sequence at all. Hedged pairs book a winner and a loser at the same moment, and two unrelated strategies interleaved on one login produce an artificial alternation belonging to neither. Where trades carry magic numbers, reading each stream separately is the only way to get an honest score.

The practical use is narrow and real: a negative score is the statistical case for rules about consecutive losses, and a score near zero is the case against them.

What the data shows

The figures below describe accounts published on ShowMyTrades, not traders in general. The median published account has 171 closed trades and a 68.8% win rate, which puts about 118 wins against 53 losses. Chance predicts roughly 74 runs from that mix, and one standard deviation is about 5.6 runs.

So on a median account the sequence has to collapse to about 63 runs, or stretch past 85, before the score reaches ±2. Anything between those bounds is consistent with a coin. That is a real constraint on the weight the number carries, and it tightens with history: the same pattern on 1,700 trades is a finding, on 171 it is a hint. It matters too that 53.9% of these accounts are more than 90% automated and 42.2% are under 10% — clustering is far more common in the first group, because a robot closing a basket writes several identical outcomes into the sequence at once.

Where you see it on ShowMyTrades

The metric appears in the Advanced Statistics module on every published account page as Z-Score (Probability), the score to two decimals with the mapped percentage beside it, in the Performance Metrics column alongside Profit Factor, Standard Deviation, Sharpe Ratio, Expectancy, AHPR and GHPR. The value is computed by our statistics service when the account syncs, not in your browser.

It also drives a line in the Smart Insights panel on the same page, where a score inside ±2 on an account with at least 30 closed trades is reported as a consistent trading pattern.

Common misunderstandings

  • "A negative z-score is bad." It is not a quality measure. It describes dependence between trades and is silent on profitability. Plenty of good trend systems score negative.
  • "It predicts the next trade." It summarises a sequence that already happened. Nothing in the runs test is forward-looking.
  • "The percentage is the chance the system works." It is the score on a different scale, nothing more. Fifty per cent is the random midpoint, not a pass mark.
  • "Ordering does not matter." It is the only thing that matters. Trades closing in the same second have no meaningful internal order, so basket-closing systems can move their own run count for reasons unrelated to the strategy.

For how many trades a statistic like this needs before it means anything, see how much history a track record needs.