I Found A Strategy That Works. Buy-And-Hold Still Beat It By 5x.

This is the story of the second of two strategies that survived out of 83 I tested — and why finding a winner turned out to be a more complicated result than finding another loser.

The setup

10.6 years of QQQ daily bars: 2016-01-04 through 2026-08-21, 2,674 sessions. I used the long history deliberately so the sample would span both the 2020 crash and the 2022 bear market, rather than testing a dip-buying idea exclusively on a decade that went up.

Eight strategies — three trend-following, five mean-reversion. Parameters fit on the first 60%, scored on the last 40%. Costs of 0.01% round-trip.

The scoring number is a skill score: how the days the strategy held compare against the days it skipped. It's a deliberately awkward measure, because it makes it hard to look good simply by being in the market during a bull run.

The winner

Buy after three consecutive down daily closes. Sell on the first up close.

That's the whole thing. No indicators, no parameters to tune beyond the number three.

skill score	+2.02 train / +2.05 test
trades	126 over 10.6 years
win rate	75%
average trade	+0.60%
median hold	1 day
time in market	8%

It was the only one of the eight to clear +2 on both halves. It survives deleting its ten best trades — t still 2.51 without them, so it isn't one lucky week wearing a trenchcoat. Ten of eleven calendar years were positive.

Its best year was 2022. The bear market. A dip-buying strategy having its best year in the worst year is the kind of thing that makes you trust a result.

And now the part nobody puts in the thread

Over the same 10.6 years, buying QQQ and doing nothing returned +552%. The strategy returned +109%.

Five times worse. From doing considerably more work.

That's not a rounding error or a cost assumption. It's structural, and it follows directly from the stat I highlighted above: the strategy is in the market 8% of the time. You cannot capture a decade of upward drift while sitting in cash 92% of it. Every day you're flat is a day the index is compounding without you.

So is it good or not?

It depends entirely on which question you were asking, and I think most people never actually decide which one they're asking.

	strategy	buy & hold
total return	+109%	+552%
worst drawdown	−6.1%	−35.6%
Sharpe	1.04	0.90

It wins on risk and loses on dollars.

If your goal is maximum money over ten years and you can genuinely stomach watching a third of your account evaporate in 2022 without touching it — buy and hold, and stop reading backtests. Nothing I tested beat it.

If your goal is a smoother ride, or you have a mandate that a −35% drawdown would end, then a 1.04 Sharpe with a −6.1% worst case is a real, defensible product. It's just a different product than "beat the market."

What it is not is a strategy that makes you more money than the boring thing. I had to sit with that for a while.

Four catches I'd need to solve before trading it

It holds overnight. My own evaluation rules require everything flat by 4:45 PM ET, no exceptions. This strategy is structurally incompatible with that — it's a genuine fork in the road, not a detail.

It has no stop loss. The exit is "the first up close," whenever that arrives. The worst single trade was −4.9%. Unleveraged that's survivable. On one MNQ contract it's roughly $2,940, which instantly busts a $2,000 maximum drawdown limit. Safe in cash, dangerous with leverage — same rules, completely different risk.

It's QQQ, not futures. Percentage moves transfer between them reasonably well. Overnight gaps and margin treatment do not.

126 trades in a decade. That's about twelve a year. Even a real edge at that frequency takes years to distinguish from luck in live trading.

The useful contrast

One of the other seven is worth showing, because it's a trap I nearly fell into.

A simple trend filter — hold only above the 200-day SMA — produced a higher Sharpe ratio (1.25) than the winner. On the surface it looks like the better strategy.

Its skill score was +0.30, off five trades.

It scores well by being absent, not by picking anything. Five events cannot tell you about skill no matter how attractive the ratio attached to them looks, and Sharpe doesn't care whether the number came from 5 decisions or 500. A great-looking metric on a handful of events is not a result. It's a coin landing heads twice.