What is a Z-score in trading?
In systematic trading, the Z-score is a runs test that tells you whether a strategy’s sequence of wins and losses is genuinely random or streaky. A Z-score beyond ±1.96 means consecutive trades are dependent at the 95% confidence level; a Z-score near 0 means winning and losing streaks are just random noise.
(In general statistics, the Z-score measures how many standard deviations a value sits from the mean. In trading we apply that same idea to the run of wins and losses, using the runs test below.)
| Z-score | Confidence of dependency | Streak meaning | Money-management implication |
|---|---|---|---|
| 0 to ±1.0 | Below 68% | Random noise | Trade normally; don’t chase streaks |
| Around ±1.65 | ~90% | Weak dependency | Monitor only |
| Below −1.96 | 95%+ | Clustering — wins follow wins, losses follow losses | Consider cutting size or pausing during a losing run |
| Above +1.96 | 95%+ | Alternating — a loss tends to follow a win | Can add size after a loss |
The Z-score (runs test) formula
Z = ( N × (R − 0.5) − X ) / √[ X × (X − N) / (N − 1) ], where X = 2 × W × L.
Here N = total number of trades, R = the number of runs (unbroken streaks of wins or losses), W = number of winning trades and L = number of losing trades. A negative Z-score means fewer runs than random (clustering); a positive Z-score means more runs than random (alternating).
Why the Z-score matters for a trading system
The Z-score does not measure whether a strategy is profitable — it measures whether its wins and losses are dependent. If losses cluster (Z below −1.96), a losing trade makes the next loss more likely, so reducing size during drawdowns can help. If results alternate (Z above +1.96), a loss tends to be followed by a win, which can justify a different money-management rule. Between −1.96 and +1.96, streaks are statistically just luck and should not drive sizing decisions.
Frequently asked questions
What is a good Z-score for a trading strategy?
There is no “good” or “bad” Z-score in the profit sense — it measures dependency, not quality. A Z-score between −1.96 and +1.96 means your winning and losing streaks are random; readings beyond that reveal exploitable structure you can manage with position sizing.
Does a high Z-score mean my strategy is better?
No. A high absolute Z-score only means the sequence of wins and losses is non-random (streaky or alternating). Profitability is measured by other metrics such as R-expectancy and profit factor.
How many trades do I need for a reliable Z-score?
Around 30–50 closed trades at minimum. Below that, the normal approximation behind the ±1.96 threshold is unreliable and a Z-score reading is mostly noise.
What is the difference between Z-score and serial correlation?
The trading Z-score looks only at the sequence of win/loss outcomes (a runs test), while serial correlation looks at the sizes of consecutive trade results. The Z-score answers “are my streaks random?”; serial correlation answers “does a big result tend to follow another big result?”
How do I use the Z-score for position sizing?
If losses cluster (significantly negative Z), reduce risk during a losing run and restore it afterwards. If results alternate (significantly positive Z), you can lean the other way. With a non-significant Z-score, keep sizing constant — the streaks are random.
The Z-score is one of several strategy-robustness metrics. Read it alongside the Sharpe ratio, R-expectancy and profit factor and the System Quality Number (SQN), and see how we apply them in our live trading portfolio.
