Two backtests, same asset, same year. Both returned about 40%. Both showed a worst drawdown near 12%. On paper they looked like the same strategy wearing different names.

One of them held a position 31% of the time. The other held one 94% of the time. They aren't the same strategy, and it isn't close.

Exposure is the percentage of the test period your system actually had money at risk. Most backtest reports print it somewhere near the bottom. Almost nobody reads it.

Here's why that's a mistake. Return is the number everyone stares at. Exposure is the denominator nobody fills in.

A system that makes 40% while sitting flat two-thirds of the year is producing that return out of a third of the risk budget. The other one produced the same number by being exposed almost continuously, which tells you it survived the year — not that it was safe.

Take the plainest example there is. A 200-day moving average filter on the S&P 500 has historically kept you invested roughly 65% of the time. Over the long stretch, that version's worst drawdown ran near 21% against about 55% for holding the index straight through. Same market, same asset, two-thirds the exposure, less than half the pain.

That gap isn't skill. It's arithmetic. You can't lose money in a crash you weren't in.

The mistake I see most often is comparing two systems on return and max drawdown alone, then picking the bigger return.

Max drawdown is one observation. It's the worst thing that happened on one path through history. If your system was in the market 94% of the time and happened to be flat during the two ugliest weeks of the sample, the drawdown number looks tame and tells you almost nothing about the next sample.

The low-exposure system didn't get lucky. It was structurally absent.

There's a second mistake, and this one was mine for years: assuming time out of the market is free.

It isn't. The standard argument against sitting flat is that the best days and the worst days live next door to each other. Since 1990, the median gap between one of the ten worst days and one of the ten best days has been about a week. Miss the ten best and $10,000 invested at the end of 1999 finishes around $33,000 instead of $75,000.

Low exposure cuts both tails. That's the actual trade. What you're buying with time on the sidelines is a narrower distribution — fewer disasters, fewer miracles — and you should want that on purpose, not by accident.

In crypto there's a cash cost stacked on top of the risk. If you're trading perpetuals, exposure is also a funding bill. A system holding a position 94% of the time pays funding 94% of the time, and at a mild 0.01% every eight hours that's roughly 11% a year bleeding out before the strategy does anything at all. The 31% system pays a third of that.

So here's the number I'd add to every strategy report you keep: annual return divided by exposure.

Call it whatever you want. A system returning 30% at 100% exposure scores 30. One returning 22% at 40% exposure scores 55. The second is doing more work per unit of time at risk, and it hands you back most of the year in dry powder — capital you can put behind a second, uncorrelated system, or simply not risk.

Run that on everything you're trading right now. My guess is at least one of them scores badly and you never noticed, because the headline return looked fine and you stopped reading there.

Our own BTC system runs near 100% exposure — it's always long or always short, never flat. I'm not going to dress that up as a low-exposure design. It means the risk comes from being wrong about direction rather than from sitting through a crash, and it gets sized with that in mind.

Whichever way your system leans, knowing the number is the part that matters. Full backtest data and methodology are at v33systematic.com.