You finished Essay 1. You're sitting back at the AAPL chart and it still looks like a winner. The thesis hasn't actually changed — small samples lie, the casino is right — but you have no replacement vocabulary. You know what you're not supposed to say (“it looks good”) and nothing concrete to say instead.
This essay is the replacement vocabulary. Specifically, it's four numbers. If you can't name all four for a trade, you don't have a trade — you have a tip.
The format is borrowed. Atul Gawande's The Checklist Manifesto opens with a story about central-line infections in ICUs.1. Gawande first laid this out in a 2007 New Yorkerpiece, “The Checklist.” The central-line checklist at Johns Hopkins reduced infection rates from 11% to 0% in a year. Five steps every doctor already knew. The intervention wasn't adding knowledge; it was forcing the steps to happen in order every time. Surgeons were getting infections in 11% of central-line insertions. Peter Pronovost wrote a five-step checklist of things the surgeons already knew. Infections went to zero in a year. Same knowledge. Same surgeons. The intervention was the checklist — forcing the four or five things to happen, in order, every time.
Trading is the same shape. The trader already knows that good trades have stops. The trader already knows that position size should be a function of risk. The intervention isn't adding knowledge — it's forcing the four numbers to be named, in order, before the click.
Number 1: Entry — what condition triggers the trade.
The first number isn't a number. It's a sentence. A sentence that describes a price-action condition specific enough that two different people looking at the same chart would agree on whether the condition is true.
“AAPL looks good” is not an entry rule. Two people can't agree on what “looks good” means. “AAPL made a higher low above the 50-day EMA and printed a bullish engulfing candle on the daily, with volume above its 20-day average” is an entry rule. You can either look at the chart and check the four conditions, or you can't.
The reason this matters is that checkable is the precondition for countable. If you can't check whether the rule was true, you can't count how many times it's been true historically, which means you can't know the next three numbers either. Every screen on ChartMath is one of these sentences, translated into code so it can be checked across thousands of historical bars without your involvement. That's the whole purpose of a screen.
Number 2: Stop — where you're wrong.
The stop is two numbers wearing a trench coat. It's a price (the level at which you'll exit) and a percentage (how much of your account you lose if you're wrong). The first number is technical; the second is account-level. You need both.
The first stop level is structural — it lives in the chart. For a bullish engulfing setup, the structural stop is usually below the engulfing candle's low. For a breakout, it's usually below the consolidation. The structural stop is the answer to the question: at what price does my thesis stop being true?
The percentage is what most retail traders skip. The percentage decides position size. If your stop is 2% below your entry, and you want to risk 1% of your account on this trade, your position size has to be 0.5 of your account. If your stop is 4% below, your position size has to be 0.25 of your account. The stop is what tells you how many shares to buy. Not your conviction. Not the chart pattern. The stop.
Risk is decided by your stop, not your conviction. The number of shares is what the stop tells you to buy.
This is the calculator that should run in the back of your head every time. The 1% number is conservative; many systematic traders use 0.25% or 0.5%. The point is that the number is fixed for the trade beforeyou click. It doesn't get bumped up because the setup feels especially clean. The system has already decided.
Number 3: R-expectation — long-run average R of trades like this.
R is the unit of trading. 1R is what you risked on the trade. If your stop is 2% below entry and you took the trade with 1% of your account at risk, a trade that hits the stop is -1R (you lost what you risked) and a trade that runs to 3% above entry is +1.5R. The R-multiple is the only honest way to compare a 1m scalp to a swing trade, because it normalizes for the size of the bet.
The R-expectation of a system is the average R per trade across a large sample.2. Expectancy = (win rate × avg win R) − (loss rate × avg loss R). A 60% win-rate system with 1.5R average win and 1R average loss has expectancy of (0.6 × 1.5) − (0.4 × 1) = +0.50R per trade. A 90% win-rate system with 0.2R average win and 1R average loss has expectancy of (0.9 × 0.2) − (0.1 × 1) = +0.08R per trade. The second one feels better; the first one compounds 6x faster.A system with +0.5R expectancy means that across many trades, you net half a unit of risk per trade. Across 200 trades at 0.5% risk each, that's 50% of account in expectation. Across 200 trades at 0.08R, it's 8%. Same number of trades, same risk per trade. The R-multiple decides the slope.
A high win rate is not a high expectancy. The system with the higher win rate isn't always the system that compounds.
That is the central confusion in retail. “80% win rate” is the most compelling lie on Twitter, because it's usually paired with tiny R wins and full-size R losses. The math underneath is brutal. The Expectancy Lab is what lets you see this directly.
The system with the higher win rate isn't always the one that compounds. Watch what changes when you trade win rate for R.
Run the three presets. The tip-botlooks elite — 80% wins, who wouldn't want that — and the expectancy is negative. The lottery ticket looks awful — losing 70% of the time — and compounds. The boring +0.5R system in the middle is what every real edge looks like. None of them feel good. All of them are math.
Number 4: Sample size — what backs the R-expectation.
The fourth number is the one Essay 1 was building toward. A system with +0.8R expectancy backed by 12 historical trades is a hypothesis. A system with +0.4R expectancy backed by 1,200 historical trades is a business. The first one has higher expected return per trade and you should still prefer the second one, because the second one is actually +0.4R and the first one might be anything.
The number you want to see on a screen is somewhere in the hundreds. A few dozen is a hint. Four-digit sample sizes are where you start treating the R-expectation as a forecast instead of a wish.
ChartMath backtests on the full historical window of every screen, not yet on a held-out walk-forward split. Rolling out-of-sample validation is on the roadmap; until then, the R-expectations you see on the screen pages are in-sample. We tell you this because the four-numbers framework should apply to us too. A system with in-sample +0.6R is a hypothesis worth running; the proof shows up in your own forward trades.
Putting the four together.
A trade with all four numbers reads like: “Bullish engulfing daily, stop below engulfing low at $182.40, R-expectation +0.7R per trade, 412 historical instances.”That's a system you can run. That's a trade you can take mechanically. That's a trade whose loss won't shake you, because the loss was priced into the expectancy you signed up for.
A trade without the four numbers reads like: “AAPL looks like a winner.” You already know what that is.
Go look at a screen's strategy analytics page. You should see all four numbers there: the entry conditions on the screen card, the stop logic and R-distribution in the strategy section, the sample size in the trade count. If a screen on the platform is missing any of the four, tell us — that's a bug.
Next: what a systematic week looks like when you've internalized this and you're running the program for real.