What a Year of Journaling in R Taught Me

I had kept a trading journal before, in the way most people do. Ticker, date, dollars in, dollars out, a sentence about why. I filled it in for a few months and then stopped, because reading it back told me nothing. Some trades made money, some lost, and the sentences all sounded reasonable.

The second attempt was different in one way: every result was recorded as a multiple of the amount I had risked. A full stop-out was −1R. A trade that made twice the risk was +2R. I kept it for a year without missing a trade. What follows is what it showed me, in roughly the order I noticed it.

The first month: the losses were not −1R

The plan was that every loss would be a planned loss, close to −1R. The first thirty rows said otherwise. Losses of −1.4R, −1.8R, one of −2.6R.

In dollars, these had looked like ordinary bad trades. In R, they were something else: trades where the outcome was worse than the plan allowed. The −2.6R was a stop I had moved. The −1.8R was a stop I had placed as a mental level and then argued with. The −1.4R was a gap, which is bad luck, but the other two were me.

That was the first lesson, and it arrived within a month. My problem was not that I lost. My problem was that I lost more than I had agreed to, and the journal in dollars had hidden it because a $340 loss and a $210 loss look like the same kind of thing. In R they are not.

The third month: I was cutting winners

Once the losses were under control, the next pattern was in the wins. My average winner was about +1.3R. My targets, on the same trades, had averaged about +2.5R.

I was exiting early. Not always, but often enough that it was the dominant feature of the results. The journal notes told me why, once I read them together: “took profit before earnings”, “looked heavy”, “did not want to give it back”. Each one sounded sensible in isolation. In aggregate, they described a trader who set a target from the chart and then overrode it from his stomach.

The fix I settled on was partial exits. Half at a fixed 1.5R, the rest at the chart target or a trailing stop. It did not maximise anything. What it did was give the nervous part of me something to do that was not closing the whole position.

The sixth month: the setup breakdown

Six months in, with about a hundred and twenty rows, I could split the results by setup. I had been trading three: pullbacks in trends, breakouts from consolidation, and reversals at levels.

The pullbacks had a clear positive expectancy. The breakouts were roughly break-even after costs. The reversals were negative, and not slightly.

I had believed, before looking, that the reversals were my best setup. They produced the trades I remembered, because a reversal that works is dramatic. The journal was not interested in what I remembered. I stopped trading reversals for the rest of the year and the overall expectancy went up by more than I would have guessed.

The end of the year: the number and its error bars

By the end I had a little over two hundred trades and an expectancy of roughly +0.3R per trade. That number is the one everybody wants. The more useful thing I learned was how uncertain it was.

Graham, who writes the statistics guides on this site, showed me how to put a confidence interval around it. With the spread of R values in my log, the honest range was wide, roughly zero to a bit over half an R. Two hundred trades, a year of work, and the data could not rule out that the edge was very small. It could rule out that it was negative, which is something. But the precision I had been assuming did not exist.

That changed how I think about size. If I do not know whether my edge is 0.1R or 0.5R, I should not be sizing as though it is 0.5R. The expectancy and position sizing guide covers the arithmetic; the journal is what made me care about it.

The three columns that mattered

If you keep only three things, keep these:

1. The result in R, so that every trade is comparable and losses larger than −1R stand out.
2. The setup name, one word, so that after a hundred trades you can see which ones are paying for the others.
3. The reason for the exit, in a few words, so that when the average winner is smaller than the average target you can find out why.

Everything else I recorded was interesting and changed nothing. Those three changed how I trade. The stock profit calculator will convert a closed trade into R if you give it the original stop, which removes the last excuse for not doing it.