How to find your trading edge: a 100-trade review
The capstone. Take 100 trades of one written setup and test whether the result survives costs, sample-size uncertainty, regime splits, plan adherence and outliers. Then write a conclusion that states how confident you can be.
Start in plain English
This is the last lesson of the Advanced course, and it asks the question the whole course has been building toward: is there an edge in the way you trade one setup, and how confident can you be? For a small-cap scalper, an "edge" that is really a lucky month, a cost assumption that was too kind or a few outlier trades is expensive to discover with real money. The review here is designed to find those problems first.
The core idea, stated once: an edge is a positive average result after costs that survives the tests you wrote down before looking. Those tests are: an interval around the average, a harsher cost assumption, a split by market regime, a split by plan adherence, and a check for dependence on a few outliers. The output is not "yes" or "no". It is a written conclusion that says what the 100 trades support and how wide the uncertainty is.
You need 100 trades of one written setup and version, in R after costs, with the fields from earlier lessons. In the practice log these are setup_tag, regime (set before each session), plan_followed, fees_usd, slippage_penalty_usd and net_r. Simulated and replay trades come from your practice log. Real trades come from your broker-synced records. Keep the two datasets separate and review them separately.
Two paths from the same start
All learners, datasets and results are Illustrative.
The review protocol, written before any of the three learners looked at the totals.
- Average net R per trade, with a rough 95% interval.
- Cost stress: add one more cent of slippage per share to every trade.
- Regime split: hot vs. cold sessions, labelled before each session.
- Plan-adherence split: within plan vs. outside plan.
- Outlier check: remove the two largest winners.
- Conclusion: "edge supported", "no edge for the written plan", or "inconclusive", with the numbers.
All three learners risk $50 per trade; their average 1R is 20 cents per share, so one extra cent of slippage costs 0.05R per trade.
Path AAn apparent edge that survives the splits
Learner Q, 100 trades. Gross +0.21R, costs 0.08R: +0.13R net per trade, standard deviation 1.15R, rough 95% interval −0.10R to +0.36R.
Cost stress: +0.08R. Hot sessions (n = 55): +0.16R; cold (n = 45): +0.09R. Within plan (n = 91): +0.15R; outside plan (n = 9): −0.07R. Without the two largest winners: still positive.
Conclusion: "Consistent with a small positive edge for the written plan. The interval includes zero, so 100 trades do not establish it. No split turned it negative."
Path BAn apparent edge that disappears
Learner R, 100 trades. Gross +0.14R, costs 0.10R: +0.04R net, interval about −0.20R to +0.28R.
Cost stress: −0.01R. Hot sessions (n = 60): +0.18R; cold (n = 40): −0.17R. Within plan (n = 78): −0.03R; outside plan (n = 22): +0.29R.
The written plan, followed as written, averaged slightly below zero. The positive total came from hot sessions and from trades outside the plan, which cannot be repeated on purpose.
Conclusion: "No edge for the written plan in this sample. The positive total depends on regime and on unplanned trades."
Path CAn inconclusive result
Learner S, 100 trades: +0.10R net. The two largest winners were +6.0R and +5.0R. Without them, the other 98 trades average −0.01R.
Conclusion: "Inconclusive. The result rests on two trades. Another 100 trades are needed, with the outlier check repeated."
All three learners could have posted "positive after 100 trades". Only one result survived the tests written in advance, and even that one is not established. The rest of this lesson walks through each step of the protocol, then the decision that follows from it.
Step 1: average net R with an interval
Compute average net R per trade and its standard deviation. A rough 95% interval is average ± 1.96 × SD ÷ √n. For Learner Q: 0.13 ± 1.96 × 1.15 ÷ 10 = −0.10R to +0.36R (computed, Illustrative).
Read the interval, not only the average. If it includes zero, 100 trades have not separated your result from break-even, which is common for scalping edges of a tenth of an R. Trades from the same stock on the same morning are not independent, so the honest interval is wider. R distributions are skewed; a bootstrap (resampling your own trades many times, Efron 1979) gives an interval that does not assume a bell shape. How many trades before your numbers mean anything covers the method.
Step 2: cost stress
Your cost assumption is a guess, especially for simulated fills. Re-run the average with one more cent of slippage per share on every trade, converted to R with your average stop distance. If the result turns negative, your edge is smaller than your uncertainty about costs. Learner R's +0.04R became −0.01R. Paper trading that teaches covers why simulated fills are usually optimistic.
Step 3: regime split
Split by the regime label you recorded before each session, the regime column of the practice log (Hot and cold small-cap markets). A result that is positive in one regime and negative in the other is a conditional edge at best, and the subsamples are small: 40 trades give a rough interval of about ±0.37R at a per-trade standard deviation of 1.2R. Note how many weeks each regime covers, not only how many trades.
Step 4: plan-adherence split
Split by whether each trade followed the written plan (plan_followed in the practice log). Two patterns matter:
- Within plan positive, outside plan negative (Learner Q): the plan holds up and deviations cost money. The fix is behavioural (Overtrading; Revenge trading).
- Within plan negative, outside plan positive (Learner R): the written plan is not what produced the result. Either the plan needs rewriting as a new version, or the outside-plan trades were luck. Neither is an edge yet.
Step 5: outlier check
Remove the largest one or two winners and recompute. If the sign flips, the result depends on rare trades. That is not automatically bad (some setups earn their money in a few large moves), but it means 100 trades contain only a handful of the events that matter, and the sample is effectively much smaller (R multiple, R multiple, on reading R distributions).
Step 6: the written conclusion
Write three to five sentences, in this order:
- What was tested: setup, version, period, n, data source (practice log or broker-synced records).
- The average net R and its interval.
- What each split showed.
- The conclusion in one of three forms: supported, not supported for the written plan, or inconclusive.
- The next step, stated as a test rather than a promise: "another 100 trades at the same size", "a new version with a changed stop", or "retire the entry".
Avoid words that claim more than the data: "proven", "works", "consistent profits". Write "consistent with", "not established", "depends on".
Observe vs. infer
Learner R's first draft reads: "100 trades, positive, profit factor 1.07. Edge confirmed, moving to live with double size."
| What the review shows (observed) | What is inferred | Why it is not established |
|---|---|---|
| +0.04R net over 100 trades | "Edge confirmed" | The interval runs from about −0.20R to +0.28R |
| Profit factor 1.07 | A margin of safety | One more cent of slippage turns the average negative |
| Within-plan trades at −0.03R | — | Left out of the draft; it is the most important line |
| Simulated trades | "Ready for live at double size" | Live fills, fees and emotions differ; doubling size doubles a result that is not established |
What shifts the odds
These change how far a 100-trade review can be trusted. They are checks, not findings:
- Independence. How many sessions and different stocks the 100 trades came from.
- Definition stability. One version throughout, with no mid-block edits.
- Skipped setups logged. Without them, the 100 trades may be a selection of the ones you liked.
- Cost realism. Simulated fills with a slippage penalty, real fills from broker records.
- Regime coverage. At least a few weeks in each regime label.
Cost of being wrong
At $50 risk per trade, 200 trades a quarter (Illustrative):
| Conclusion acted on | If the true average is +0.13R | If the true average is −0.05R |
|---|---|---|
| Trade it at the same size | +$1,300 | −$500 |
| Trade it at double size | +$2,600 | −$1,000 |
| Stay in sim for another 100 trades | $0 (and more evidence) | $0 (and more evidence) |
Learner Q's interval covers both columns. Doubling size widens the range of outcomes in both directions; Risk of ruin shows how risk per trade shapes the chance of a deep drawdown even when the edge is real.
How many trades before this means anything
One hundred trades is a checkpoint, not a finish line. At a per-trade standard deviation of about 1.2R, narrowing the interval to ±0.10R takes about 550 independent trades (Illustrative, normal approximation; see Backtesting intraday ideas). Run this review every 100 trades per playbook entry, keep the earlier conclusions, and watch how the interval narrows (Building a playbook).
What this lesson does not claim
- It does not claim any learner has an edge. All datasets are Illustrative.
- Passing every split does not establish an edge; it removes several common ways of being fooled.
- A past edge does not indicate future results.
- Deciding whether to trade live, stay in a simulator or change style is a personal risk decision. This lesson describes evidence, not what you should do.
Go deeper
- FennEdge first.
- For real trades, the Trading Journal syncs fills from supported brokers; on Plus and above it shows net P&L, profit factor, maximum drawdown and maximum consecutive losses computed from dollar P&L, with fee detail on Pro and Max. In its Journal Review view (Pro or Max, as of 2026-10-05) you can tag each trade by playbook entry and mark it "Within plan" or "Outside plan", which supplies Step 4 for real trades. R, regime labels and intervals stay in your own sheet; as of 2026-10-05 the Journal has no planned-stop field. The Journal holds broker-synced trades only; simulated and replay trades stay in the practice log.
- Desk can hold your written conclusion as an ongoing note (Plus and above; saved when signed in), next to the playbook entry it reviews.
- Level Up (Plus and above) is a private challenge scored on your Journal net P&L, with an optional daily max loss per level. It does not track tags or entry timing, so keep process goals such as "100 tagged trades of one setup" in your own log.
- Where this started. Capstone: a measured 50-trade sim block is the Foundations version of this review.
- Excursions. Add MAE and MFE to the review if you are deciding about stops or exits (MAE and MFE).
- Trading live vs. staying in sim. Moving from simulated to real fills changes costs, queue position and pressure. Many traders repeat this review on a small real-money sample before changing size; whether to do that is your decision.
Exercises
Nine exercises follow this lesson: three concept checks (reading an interval that includes zero; why the adherence split matters; wording a conclusion), five calculations on a fictional learner's 100-trade summary (net expectancy, the interval's lower bound, cost stress, the cold-regime average and the within-plan average), and one structured practice task. After each calculation, repeat it with your own numbers. Exercises, answers and progress are available after signing in.
Practice task: your 100-trade edge review. This is the ninth exercise (ungraded). Choose one playbook entry with at least 100 trades under one version. Use the FennEdge practice log for simulated and replay trades (download the practice log (CSV), with a column guide); review real broker-synced trades separately, using your Journal data plus your own R sheet. Write the six-step protocol before computing any totals, then complete each step and the written conclusion. Keep practice records out of the Trading Journal, which holds only real broker-synced trades.
Primary sources
- Efron, B. (1979). Bootstrap Methods: Another Look at the Jackknife. The Annals of Statistics 7(1).
- FINRA. Frequent Intraday Trading: Understanding the Basics.
Sources checked 2026-10-05. All learners, datasets, splits and costs in this lesson are fictional and Illustrative. This lesson is education, not a recommendation to buy or sell any security. Past patterns do not indicate future results.