I tested my own strategy and it didn't work

StrategyLens started as a way to test one idea properly. I'd picked up a handful of indicators over the years, and each one seemed useful on its own. I wanted to combine them into one set of rules and see whether they did better together.

That was harder to try than it sounds. On TradingView, running several indicators at once means a pricier plan, and testing a combination across many stocks by hand is slow and awkward. So I built something that could do it at scale.

What I expected

Nothing spectacular. Just that a disciplined, rules-based version would beat simply buying and holding the same stock, often enough and by enough to justify the extra effort and risk.

What the evidence said

It didn't. Across the board, the strategy didn't show it could reliably beat buy-and-hold.

That's a narrower claim than "this idea is rubbish", and I want to keep it narrow. Some version of it might work somewhere. What I set out to show, I couldn't show.

The part that actually mattered

Earlier results looked better. Working out why is the most useful thing to come out of this project.

Some of the checks I'd built to stop me fooling myself were too generous, and they all leaned the same way: towards making a result look good. None of it was deliberate, and none of it was obvious. I only found them by feeding each check examples where I already knew the right answer.

Once they were fixed and everything re-ran, the good numbers didn't survive.

That's how backtests usually go wrong. It isn't fraud or one big mistake. It's a few small, generous choices that all point the same way, because the choices that flatter you are the ones you're least likely to question.

Why publish this

A site that has never shown a failure hasn't found that markets are easy. It's showing you a filter. If you only see the wins, you can't know how much was thrown away to produce them.

Every ticker page on StrategyLens says how much testing sits behind its verdict, and flags when there's too little to trust. It's the same principle, one stock at a time.

Pick a stock you know and see what the evidence says. For example, SPY.

Not financial advice, and not a prediction. It's a record of what I measured and what it said.

Did the strategy beat buying and holding?

Not reliably. On SPY, 3 of 553 backtests beat holding (0.5%). On QQQ, 97 of 553 did (17.5%). The median strategy trailed holding on both: 1.3% a year against 10.8% on SPY, and 3.8% against 10.8% on QQQ.

Why did the early results look better?

Some checks were too generous, and they all leaned the same way. Once they were fixed, the good numbers did not survive. Every ticker page now shows how many backtests sit behind its verdict.

How many tests sit behind a verdict?

On the current cards, each ticker has 553 backtests, run with default settings. The cards are dated 29 October 2026.

Is this financial advice?

No. This is research, not financial advice. Past results, real or simulated, do not tell you what will happen next. Capital at risk.