Win/Loss Streak Analysis

Win/Loss Streak Analysis goes beyond counting consecutive losses — it applies statistical frameworks to determine whether a streak reflects normal variance or a genuine breakdown in trading edge. As a consistency metric, it answers one of the most practically important questions a trader faces: when should a losing streak trigger a process change, and when should it be ignored?

Formula & Calculation

Max Expected Losing Streak = log(N) / log(1 / (1 - p))

Where:

• N = total trades in the sample period
• p = win rate as a decimal (e.g., 0.52 for 52%)
• log = natural logarithm

This formula gives the longest losing streak you should expect before it becomes statistically anomalous. For the Wald-Wolfowitz runs test:

Z = (R - E[R]) / σ

Where:

• R = actual number of runs (alternating W/L sequences) observed
• E[R] = expected runs = 1 + (2 × n1 × n2) / (n1 + n2)
• σ = standard deviation of runs = sqrt((2n1n2(2n1n2 - n1 - n2)) / ((n1+n2)² × (n1+n2-1)))
• n1 = total wins, n2 = total losses

If |Z| exceeds 1.96, the sequence is non-random at 95% confidence. A negative Z means losses cluster together (streaks longer than expected); a positive Z means outcomes alternate more than expected.

Benchmarks

Level	Range	What It Means
Within Expectation	Streak at or below log(N)/log(1/(1-p))	Normal statistical variance — no adjustment warranted
Statistically Unusual	Streak exceeds threshold;	Z
Non-Random Clustering	Streak exceeds threshold and	Z
Severe Degradation		Z

Practical Example

A day trader closes out a quarter with 200 trades, a 52% win rate (104 wins, 96 losses). In week 6, they hit a 9-trade losing streak and immediately cut position size by 50% out of fear.

Step 1 — Calculate the threshold. Using the max-streak formula with N=200 and loss probability 0.48: log(200) / log(1/0.48) = 5.298 / 0.734 ≈ 7.2. A 9-trade losing streak exceeds the expected maximum of 7.2, so it is statistically unusual and warrants a formal test.

Step 2 — Run the Wald-Wolfowitz test. With n1=104 and n2=96: E[R] = 1 + 2(104)(96)/200 = 1 + 99.84 = 100.84. σ = sqrt((19968 × 19768) / (40000 × 199)) ≈ sqrt(49.6) ≈ 7.06. The trader counts 87 actual runs in their sequence. Z = (87 - 100.84) / 7.06 = -13.84 / 7.06 ≈ -1.97.

Result. Z = -1.97 just crosses the -1.96 threshold, confirming mild negative autocorrelation — losses are clustering slightly beyond chance. The size reduction was statistically justified, but for the right reason. After 20 additional trades, the trader retests; if Z normalizes above -1.96, full size resumes.

How to Track Win/Loss Streak Analysis

1. Record trade outcomes in sequence — Log each trade as W or L immediately after closing, preserving the exact chronological order. Out-of-order entry invalidates the runs test.
2. Calculate your win rate baseline — After each 50-trade block, update p so your threshold formula stays accurate as your edge evolves.
3. Apply the max-streak formula — Before each trading session, know your current threshold. If streak length approaches it, you’re entering review territory.
4. Run the Wald-Wolfowitz test at threshold — When a streak exceeds your formula threshold, count total runs R in your full sequence and compute Z. Do not act on the streak until you have a Z-score.
5. Retest every 20 trades during a flagged period — Non-random clustering can normalize as market conditions shift. Recompute Z regularly rather than holding a size reduction indefinitely.

How to Improve Win/Loss Streak Analysis

1. Audit streak composition by setup type — Export trades from the streak period and filter by setup. If one setup accounts for most losses, the streak reflects setup-specific degradation, not overall edge failure. Suspend that setup specifically rather than cutting all size.
2. Separate streaks by market regime — A trending strategy with a 55% overall win rate may have a 35% win rate in choppy markets. Tag trades by regime and recompute streaks within each regime to isolate regime-sensitive edge.
3. Confirm autocorrelation before changing size — Anti-martingale systems (increasing size after wins) only add value if positive autocorrelation is confirmed. Without |Z| above +1.96, you are sizing up on noise and increasing variance without improving expectancy.
4. Track emotional state during streaks — Log a 1-5 confidence score with each trade. Non-random loss clustering frequently coincides with declining pre-trade confidence scores, which reveals behavioral contribution to the streak versus pure market-driven variance.

Common Mistakes

1. Reacting to streak length without a threshold — A 6-loss streak sounds alarming, but at a 45% win rate with 200 trades, the expected maximum is about 7.7 trades. That streak is well within normal variance. Calculate the threshold first; then decide if a response is warranted.
2. Applying sizing changes without autocorrelation confirmation — Most retail traders’ trade outcomes show near-random sequencing. Reducing size after losses or increasing it after wins without statistical confirmation is trading on noise and adds volatility to your equity curve without improving expectancy.
3. Committing the gambler’s fallacy — After 6 consecutive losses, each new trade still carries exactly the same probability as before. A win is not overdue. This fallacy causes traders to oversize on “makeup” trades, which often become the largest individual losses.
4. Comparing streak lengths across different win rates — A 5-loss streak at 40% win rate occurs roughly every 13 trades (0.6^5 = 7.78%). The same streak at 60% win rate occurs approximately every 100 trades. Raw streak counts are meaningless without win rate context.
5. Running the analysis on too few trades — The Wald-Wolfowitz test requires at least 50 trades for reliable results. On 30 or fewer trades, Z-scores are unstable and will frequently generate false positives.

How JournalPlus Calculates Win/Loss Streak Analysis

JournalPlus tracks your win/loss sequence automatically from your trade log, displaying current streak length, maximum winning streak, and maximum losing streak on the analytics dashboard. The streak view updates in real time as trades are added. You can export your full trade sequence as a CSV with outcome columns preserved in chronological order, making it straightforward to paste into a spreadsheet and compute the Wald-Wolfowitz Z-score using your trade history. The equity curve analysis and win rate panels give you the N and p values needed to apply the max-streak formula directly. For traders who want to evaluate risk of ruin probability alongside streak data, both metrics pull from the same underlying trade log.