The Hidden Cost of Switching Strategies Every Few Months

I once counted the strategies I'd traded in a single year. Eleven. Eleven different systems, none of them given more than six weeks before I decided it was broken and moved on.

Here's what I didn't understand at the time: I never actually tested any of them. I tested my patience, and my patience kept losing.

A strategy isn't a switch you flip on Monday and judge by Friday. It's a distribution. Any approach with a real edge still loses plenty of individual trades, and those losses don't politely space themselves out. They cluster. You can have a system that wins over 200 trades and still hit a stretch of eight losers in a row inside that run. Most traders meet that stretch in week three and conclude the strategy is dead.

It isn't dead. It's just early.

Most strategies need somewhere between 50 and 100 trades before the numbers mean anything. Less than that and you're reading noise as if it were signal. If you take three trades a week, 50 trades is four months of consistent execution before you even have the right to an opinion about whether the thing works. Switch at week six and you've thrown out the experiment before collecting the data.

Now here's the part that actually costs you. Every time you start over, the clock resets to zero.

You don't just lose the new strategy's runway. You lose the skill you were building with the old one. Knowing where a system tends to give false signals, how it behaves in chop, when to size up and when to sit on your hands — that knowledge only comes from reps. Strategy number two doesn't inherit anything from strategy number one. You're a beginner again, every single time, paying the beginner's tuition again, every single time.

I think this is why so many genuinely smart, hard-working people quit trading for good. It's not that they couldn't find an edge. It's that they kept abandoning edges right before those edges had a chance to show up in the equity curve. They mistook the natural shape of variance for personal failure, and after enough of those false verdicts, they decided they just weren't cut out for it.

The mistake underneath all of this is a simple confusion: treating a drawdown like a verdict.

A drawdown is what a working strategy looks like from the inside during a bad stretch. A broken strategy is something else entirely — a system whose underlying assumption stopped being true, like a mean-reversion bot running into a market that suddenly trends for months. Those two things feel identical when you're in them. Your account is bleeding either way. The only way to tell them apart is to have decided, in advance and in writing, what failure actually looks like. How many trades. How deep a drawdown. Outside what range of conditions.

If you don't define that ahead of time, your gut defines it for you in the moment. And your gut's threshold for "this is broken" is almost always "I'm down and I'm uncomfortable." That's not a system evaluation. That's an emotional one.

So here's the one thing I'd change if I could go back and talk to the version of me running eleven strategies a year. Before risking a dollar on any approach, write down the exit criteria for the strategy itself, not the trade. The sample size you'll judge it on. The maximum drawdown you'll tolerate before walking away. The market regime it's actually built for. Then hold yourself to it the way you'd hold yourself to a stop loss.

When you've defined failure in advance, a losing streak stops being a crisis and becomes data. You can sit through week three because you already know week three is inside the range you agreed to. You let the sample fill up. You give the edge enough trades to either prove itself or fail honestly, on the terms you set, instead of the terms your anxiety set.

This is one of the quieter reasons I moved to a fully systematic approach. A bot doesn't get bored at week six. It doesn't see eight red trades and panic-switch to whatever looked good on Twitter that morning. It runs the same logic across the full sample, which is the only way you ever find out if the logic was any good.

Pick one thing. Define how it's allowed to fail. Then give it long enough to answer the question you're actually asking.