Reading Backtest Results: Expectancy, Drawdown and Streaks

The test finishes. The summary line reads: win rate 40%, average win 2.5R, average loss 1R. The rules and the figures are hypothetical. They still raise the questions any real summary raises, and each of them has an answer you can work out with a calculator from the trade list before you decide whether the rules deserve paper, a small account or the bin. Is 40% bad? What does a normal bad month look like? How much of the result could be luck?

R, the unit that makes trades comparable

R is the amount risked on a trade: the distance from entry to stop times the share count. With the fixed $500 risk from the first lesson, 1R is $500. A $1,250 gain is a 2.5R win. A stop-out is a 1R loss. Measuring in R lets you compare a trade in a $20 stock with one in a $200 stock without the share prices getting in the way.

Expectancy

Expectancy is the average trade’s result in R.

So a 40% win rate is fine here. The big wins cover the frequent losses. A 60% win rate with average wins of 0.5R would be worse: 0.6 × 0.5 − 0.4 × 1 is −0.1R, a strategy that wins most of the time and loses money. Win rate on its own tells you almost nothing.

Expectancy also scales into a rough yearly figure. At 0.4R a trade and 200 trades a year, the hypothetical rules would average 80R, or $40,000 at $500 per R, and on a $50,000 account a number that large should make you suspicious of the test before it makes you excited about the strategy.

Expectancy must be after costs. Charge the commission and slippage from the previous lesson first, then calculate.

Maximum drawdown

Maximum drawdown is the largest peak-to-low fall in the account. It tells you how bad the worst stretch was in money terms.

Say the hypothetical curve peaks at $62,000. It falls to $53,000, then recovers.

Divide by the peak, never the low. Would you have kept trading through a $9,000 loss? If the honest answer is no, the position size is too large for you, whatever the expectancy says, and the fix is a smaller risk per trade before any live money goes in.

The backtest’s worst drawdown is only the worst so far. It is the worst in one particular sequence of history. A live run can do worse. Future prices will arrive in a different order, and the order is what decides how deep the hole gets, so treat the tested figure as a floor for planning and ask whether you could stand a drawdown noticeably larger than it.

Losing streaks

At a 40% win rate, losses come in bunches. Take any given eight trades. The chance all eight lose is 0.6 to the eighth power.

About 1.7% sounds rare. It is the chance for one particular block of eight. Over a few hundred trades there are hundreds of overlapping blocks, so the chance that an eight-trade losing run appears somewhere in the list is far higher than 1.7%, and a trader who sizes from the single-block figure is planning for a streak that the arithmetic says to expect sooner or later. The losing streak calculator runs that longer sum for your win rate.

Plan for the streak. Eight losses at $500 is $4,000. On $50,000 that is 8%.

How many trades stand behind it

Last, count the trades. A summary built on 20 trades is fragile.

Take 20 hypothetical trades with 8 wins at 2.5R and 12 losses at 1R. That is 20R − 12R = 8R, or 0.4R a trade. Change one win into a loss. Now 7 wins give 17.5R and 13 losses cost 13R. The total is 4.5R, or 0.225R a trade. One trade nearly halved the expectancy. With hundreds of trades, a single result barely moves it.

The strongest check is data the rules never saw. See out-of-sample testing is the only part that counts. The stock trading hub has more on sizing. Numbers that hold up move on to paper trading and small size, where the rules meet a real market.