HOW TO KNOW IF A STRATEGY IS WORKING — LUCK VS BROKEN EDGE.
EducationRisk~19 min readUpdated 30 September 2026
The short answer
A strategy is working if its expectancy is positive over a sample large enough for the result to be meaningful. That sample is usually 100+ trades, not 10 or 30. Below that threshold, the results are dominated by luck. Above it, the results reflect the edge. The question is not "am I winning?" It is "do I have enough data to tell?" Most traders abandon good strategies because they cannot distinguish bad luck from a broken edge.
Why this follows portfolio heat
Lesson 47 capped total risk. This lesson answers the next question: given that risk framework, is the strategy producing a real edge? A strategy with positive expectancy and 3% heat limit is a business. A strategy with negative expectancy and 3% heat limit is an expensive hobby. Both look identical over 10 trades.
The whole point of the risk framework in Block 7 is to keep you alive long enough to gather enough data to answer this question.
T
Written by the Trade To The Top team|Reviewed 30 September 2026
Strategy evaluation methodology cross-checked against Trade Your Way to Financial Freedom (Van Tharp), Evidence-Based Technical Analysis (Aronson), and the sample-size and confidence-interval frameworks used in statistical process control. Expectancy and standard deviation formulas verified against the standard performance metrics used in CTA reporting.
Most traders do not have a strategy problem. They have a sample-size problem. They take 20 trades, hit a drawdown, and conclude the strategy is broken. Or they take 20 trades, get lucky, and conclude they have found the edge. Neither conclusion is supported by the data. Twenty trades tells you almost nothing. One hundred trades tells you something. Three hundred trades tells you whether to keep going.
Key takeaways
Expectancy is the only metric that matters. Not win rate, not total P&L, not the last trade.
Expectancy = (Win rate × Average win) − (Loss rate × Average loss). Expressed in R.
Sample size determines whether expectancy is real or noise. 100+ trades is a minimum. 300+ is better.
Standard deviation matters. Two strategies with the same expectancy but different variance have different risk profiles.
Random chance produces streaks. Ten wins in a row is normal even for a 40% win rate system.
Bad luck and broken edge look identical over 20 trades. The difference only shows up over 100+.
Do not change the strategy mid-sample. Every change resets the count.
Grade plan adherence, not just P&L. A trade that followed the plan is valid data even if it lost.
The signal to stop is not drawdown. It is negative expectancy after 100+ trades.
The signal to continue is positive expectancy, even during drawdown.
A strategy is working if it has positive expectancy over a large enough sample to trust the result. That is the full definition. Everything else is a component of it.
Two common wrong definitions:
"I made money this month." Luck can produce a profitable month from a negative-expectancy strategy. It happens all the time.
"My win rate is 60%." Win rate alone tells you nothing. A 60% win rate with 0.5R winners and 2R losers is a losing strategy.
Expectancy is the only number that answers the question. Everything else — win rate, R:R, total profit — is a component of expectancy. None of them can answer the question alone.
WHAT "WORKING" LOOKS LIKE · EXPECTANCY IN FOUR SCENARIOS
Four strategies · different win rates, different R:R, four different expectancies
Two strategies with the same expectancy. Win rate alone cannot tell you which is working.
Common mistake
Celebrating a 65% win rate. Win rate without expectancy is a feel-good number. A system that wins 65% of the time but averages +0.4R on winners and −1R on losers has expectancy of (0.65 × 0.4) − (0.35 × 1) = 0.26 − 0.35 = −0.09R. It is a losing system dressed up as a winning one. Always compute expectancy, not win rate.
The expectancy formula
Expectancy is the average R you expect to gain (or lose) per trade, given your win rate and average win/loss. It is the closest thing to a single number that answers "is this working?"
Formula — expectancy in R
E = (W × Avg Win) − (L × Avg Loss)
Where W = win rate, L = loss rate (1 − W), Avg Win = average R on winners, Avg Loss = average R on losers (usually close to 1R for a disciplined trader).
Example — 45% win rate, 2:1 R:R
W = 0.45 · L = 0.55 · Avg Win = 2.0R · Avg Loss = 1.0R
E = (0.45 × 2.0) − (0.55 × 1.0)
E = 0.90 − 0.55 = +0.35R per trade
Over 200 trades: 200 × 0.35R = +70R expected
At 1% risk per trade, that is +70% expected over the year.
EXPECTANCY IS THE EDGE. EVERYTHING ELSE IS NOISE.
Expectancy also tells you how much variance to expect. A higher expectancy means more room for the drawdowns to be survivable. A lower expectancy means every drawdown eats further into the edge.
Expectancy
What it means
Verdict
Above +0.30R
Strong edge. Survivable drawdowns. Tradeable at 1% risk.
Working.
+0.15R to +0.30R
Modest edge. Drawdowns need discipline but are recoverable.
Working.
+0.05R to +0.15R
Thin edge. High sensitivity to spread, slippage, and rule violations.
Marginal.
−0.05R to +0.05R
Indistinguishable from zero. Requires far more data to conclude.
Unproven.
Below −0.05R
Losing system. No amount of position sizing saves it.
Not working.
Sample size
The expectancy formula gives you a number. But the number is only meaningful if the sample is large enough. Over 10 trades, expectancy can be +2R or −2R by pure chance, regardless of the true edge.
The rule of thumb: you need at least 100 trades before you can trust the expectancy estimate. 300 trades is better. 500 gives you statistical confidence.
SAMPLE SIZE · HOW THE ESTIMATE SETTLES
Rolling expectancy measured over a growing sample of trades
The expectancy estimate swings wildly for the first 30–50 trades. It stabilises around 100–150.
At 20 trades, you are reading noise. At 200 trades, you are reading the edge.
Luck vs broken edge
Here is the trap. Bad luck and a broken edge look identical over a small sample. Both produce losing streaks. Both produce negative equity curves. Both feel like "the strategy stopped working."
The only way to tell them apart is to wait for the sample to be large enough. If you have 100 trades and expectancy is positive, the drawdown is bad luck. If you have 100 trades and expectancy is negative, the edge is broken.
BAD LUCK VS BROKEN EDGE · SAME 10-TRADE RESULT, DIFFERENT CAUSES
Two equity curves over 10 trades · both down −4R · the difference shows up at trade 100
Same 10-trade drawdown. One strategy recovers. One keeps declining. You cannot tell which is which at trade 10.
Signal
Broken edge
Bad luck
Sample size
100+ trades, expectancy negative
Under 100 trades, positive expectancy on the sample so far
Rule adherence
Multiple rule violations during the losing stretch
Rules followed, trades match backtested patterns
Setup quality
Setups look structurally different from backtested ones
Setups look identical to backtested ones
Market regime
Strategy no longer matches current market state
Same regime the strategy was built for
Drawdown depth
Exceeds historical max by 50%+
Within historical range
Recovery signal
No recovery after 20 additional trades at reduced size
Recovery begins once the streak ends
Variance and standard deviation
Two strategies can have the same expectancy but completely different risk profiles. This is because variance determines how wild the equity curve will be around the expected return.
Standard deviation is the measure of variance. A strategy with the same expectancy but higher standard deviation has bigger swings — bigger drawdowns and bigger winning streaks.
Strategy
High variance
Low variance
Win rate
30% — more losses, bigger wins
65% — more wins, smaller wins
R:R per trade
3:1 — fewer but larger winners
0.7:1 — many small winners
Expectancy
Same
Same
Typical losing streak
8–12 losses
3–5 losses
Drawdown depth
Deeper, longer
Shallower, shorter
Psychological difficulty
High — many losses before the win
Lower — frequent wins
Both strategies are equally profitable in theory. But most traders cannot execute the high-variance strategy, because the losing streaks are longer than their patience allows. They abandon it during the drawdown, thinking the edge is broken — when in reality, the edge is intact and they simply could not sit through the variance.
Keep, pause, or abandon
Based on everything above, here is the decision framework.
DECISION FRAMEWORK · KEEP, PAUSE, OR ABANDON
Three possible actions based on sample size and expectancy
Below 50 trades, the decision is almost entirely noise. Above 150 trades, act on the expectancy.
The three actions
01
Keep trading. Under 50 trades or positive expectancy in the sample. The rules stay the same. Change nothing. Continue gathering data.
02
Pause and diagnose. 50–150 trades with a concern. Check rule adherence. Check setup quality. Check market regime. Do not change the strategy yet — the sample is not large enough.
03
Act on the data. 150+ trades. Positive expectancy → continue at full size. Negative expectancy → stop, diagnose, rebuild. Near zero → test specific changes on paper before live.
Worked example — 50 trades, two conclusions
Strategy
Range breakout with retest
Historical expectancy
+0.35R per trade
Trades logged
50
Current drawdown
18%
Historical max DD
16%
Trader A — Abandons at 50 trades.
Sees 18% drawdown > 16% max. Concludes the strategy is broken.
Abandons, starts new strategy, loses again after another 50 trades.
Cycles through three strategies over six months.
Total result: −14R and no data on any strategy.
Trader B — Diagnoses at 50 trades.
Sees 18% DD. Reviews journal:
• 50 trades logged, 42 followed the plan, 8 did not.
• Among the 42 rule-followed trades: expectancy +0.28R.
• The 8 violations averaged −1.4R each.
The strategy is not broken. The trader broke the rules.
Action: keep the strategy, tighten discipline on entry criteria.
Result over the next 100 trades: +32R, account recovers.SAME 50-TRADE DRAWDOWN. OPPOSITE DECISIONS. OPPOSITE OUTCOMES.
Trader B did not change the strategy. Trader B checked whether the strategy was being executed correctly. The 50-trade drawdown was not evidence of a broken edge. It was evidence that the trader had broken the rules.
The seven evaluation rules
The rules — print these
01
Compute expectancy, not win rate. Every 25 trades, recompute expectancy in R. That is the only metric that matters.
02
Do not judge the strategy before 50 trades. Below that threshold, results are dominated by luck. Keep trading.
03
Do not change the strategy before 150 trades. Any change resets the sample. Frequent changes produce no data.
04
Grade plan adherence, not just P&L. Count plan-followed trades separately. Expectancy on plan-followed trades is the real number.
05
Expect losing streaks. A 45% win rate system will produce 8-loss streaks regularly. That is not a broken edge. It is the math.
06
Compare variance, not just expectancy. A high-expectancy, high-variance strategy is only tradeable if you can sit through the streaks.
07
Stop only on 150+ trades with negative expectancy. Not on drawdown. Not on a bad week. On the data.
When this fails
When this fails
The strategy has a very low trade frequency. If the strategy only produces 3 trades a month, gathering 150 trades takes 4 years. In that case, judge on process quality and rule adherence rather than expectancy. The sample will never be large enough on a reasonable timeline.
The market regime has clearly shifted. A strategy designed for ranging markets will not produce data in a trending market. Sample is invalid when the market regime the strategy was built for no longer exists.
The trader changed rules mid-sample. Every time the rules change, the sample resets. The 150-trade count only applies to trades taken with the same rules. If you changed the plan twice in 100 trades, you have three samples of ~30 trades each. None are reliable.
Expectancy is very near zero. A strategy with expectancy of +0.02R is not tradeable. Spread and slippage will eat it. The threshold for "working" is +0.10R, not +0.01R.
The sample includes rule violations. If 30% of trades broke the plan, the expectancy calculation is contaminated. Compute expectancy separately for plan-followed and non-plan-followed trades. The plan-followed number is the strategy's expectancy.
The trader is measuring the wrong thing. Total P&L is a function of expectancy × number of trades × risk per trade. It is not a direct measure of edge. Always measure expectancy in R.
If you remember nothing else: expectancy is the edge. Sample size determines whether the expectancy number is real. Do not judge a strategy on fewer than 100 trades.
In one box
Expectancy = (W × Avg Win) − (L × Avg Loss). In R.
Win rate alone is meaningless. Compute expectancy.
Minimum 100 trades before you trust the number. 150 is better.
Do not change the plan before 150 trades. Every change resets the count.
Grade plan adherence separately. Compute expectancy on plan-followed trades only.
Expect losing streaks. 8 losses in a row is normal at 40% win rate.
Variance matters. Same expectancy, different variance = different tradeability.
Stop only on 150+ trades with negative expectancy. Not on drawdown.
See it in practice. Our free trading journal computes expectancy, win rate, and average R automatically. It separates plan-followed trades from violations so you can see the real edge. No more guessing whether a strategy is working.
5 questions · immediate feedback · retake any time
Question 01 of 05
What is the formula for expectancy?
Correct: C. Expectancy in R = (Win rate × Average win in R) − (Loss rate × Average loss in R). This is the single number that answers "does this strategy have an edge?"
Question 02 of 05
How many trades do you need before you can trust the expectancy estimate?
Correct: B. 100–150 trades is the minimum. Below 50, the sample is dominated by luck. Above 100, the estimate stabilises. Above 150, act on the data.
Question 03 of 05
You have 30 trades and a 15% drawdown. What should you do?
Correct: D. 30 trades is not enough data to conclude anything. Continue, reduce size per the drawdown protocol, and keep logging trades. The sample needs to reach 100+ before making a decision.
Question 04 of 05
What is the single signal that a strategy should be stopped?
Correct: A. The only evidence-based signal to stop is negative expectancy over a large sample. Losing streaks, drawdowns, and losing weeks are all expected variance, not evidence of a broken edge.
Question 05 of 05
Two strategies both have +0.35R expectancy. Strategy A has a 45% win rate. Strategy B has a 30% win rate. What is the main practical difference?
Correct: C. Same expectancy, different variance. The 30% win rate strategy will produce longer losing streaks and deeper drawdowns. Many traders cannot sit through that variance, even when the edge is real.
Why the strategy that worked in the backtest stops working in live trading. Slippage, spread, look-ahead bias, and the reality gap. Final lesson of Block 7.