The Only Number I Track That Isn't P&L

Last Friday I finished the week down 2.1%. I wrote "96" next to it in my log, closed the laptop, and went to dinner in a good mood.

96 isn't a return. It's the percentage of that week's trades that followed the plan exactly — right trigger, right size, right exit, no improvising. Twenty-four trades, one of which I sized up because the setup looked cleaner than the rest. One break out of twenty-four.

That's the only number I grade myself on now, and it took me about three years of staring at P&L to work out why.

Here's the problem with profit as a score. Over a week, or even a quarter, your P&L is mostly a report on what the market did, not on what you did. A trend system in a chop market loses money while executing perfectly. That same system in a clean expansion makes money while you break half your own rules.

Profit and skill do converge eventually. But the window where "eventually" arrives is far longer than anyone's attention span, and you have to make decisions in the meantime.

Adherence doesn't have that problem. Nothing the market does can move it. It's the one number on your screen that's entirely a report on you.

Richard Dennis and William Eckhardt ran the cleanest test anyone's run on this. In 1983 they handed a group of beginners an identical rule set — same signals, same markets, same position-sizing formula — and funded them with real money. Results still spread out. Some of the Turtles got cut from the program for not taking trades the system told them to take.

Eckhardt's standard was that two traders with the same inputs should produce the same outputs, and if they don't, the system isn't the system anymore. What varied in that room wasn't the edge. It was compliance with the edge.

There's a version of this outside markets that I think about a lot. When the WHO surgical checklist was rolled out across eight hospitals, inpatient deaths after major surgery fell from 1.5% to 0.8%. Nobody got better at surgery that year. What changed was whether a short list of steps happened every single time — and the thing being measured was the steps, not the outcome of any one operation.

Most traders have exactly one number, and it refreshes every second. That's the whole mistake. As a feedback mechanism, P&L is a slot machine: it pays out for sloppy process on lucky days and punishes clean process on unlucky ones. Learn from it trade by trade and you're training yourself on noise.

Which is why the dangerous week isn't the losing one. It's the week you made 6% with four rule breaks in it.

You'll remember that week. Some part of you will file it as evidence that your judgment beats your plan, and you'll reach for that override again later, at a worse moment, with more size on. A clean losing week teaches you nothing harmful. A dirty winning week teaches you something that costs money for years.

So score it. One question per trade: did every step happen as written? Yes or no, no partial credit — partial credit is how you talk yourself into a better grade.

Count the yeses, divide by the number of trades, and write the result down before you look at your P&L. Check the money first and the money grades the trades for you.

Then watch the trend, not the level. A single 88 means nothing on its own. But 95, 91, 84, 78 across four weeks means something is happening to you — fatigue, a drawdown you're quietly fighting, boredom, whatever it is — and it shows up in that number weeks before it shows up in your equity curve.

That's the real value of tracking it. It's the only leading indicator you'll ever have on yourself.

Most traders I know who use this target 90% and treat anything under 80 as a signal to stop for the day rather than a signal to try harder. In my experience that's about right. Below 80 you're not running a strategy anymore, you're running a strategy plus a second, undocumented strategy made of moods, and there's no way to backtest that one.

If your score sits in the 70s month after month, the honest read might be that you shouldn't be the one pressing the buttons. That's most of why I ended up running systems instead of trades — a bot scores 100 by construction, and my only remaining job is to leave it alone. Full backtest data and live results are at v33systematic.com if you want to see what the other side of that looks like.