Do trading indicators actually work?
In our tests, no. We backtested 1,342 trading strategies built on common indicators and chart rules, and after correcting for luck across 1,774 tests, 0 passed. This is a backtest, in-sample, not a promise.
Last updated
What we tested
We wrote each popular rule down as exact code, ran it on past prices of stocks, ETFs and crypto, and charged it trading costs. That gave 1,342 strategies. Then we checked them all together, counting every try. This is a backtest, in-sample, not a promise.
Why a high win rate misleads
Many indicator rules look great because they win most trades. But a win rate says nothing about size. A rule can win nine trades in ten by taking tiny profits, and lose it all back on the tenth. In our quick tests, 1,295 rule and target pairs won at least 60% of their trades. After costs, none of them made money. A few examples:
- RSI(2) extreme with trend filter: won 86.6% of 71,970 trades, yet lost 0.59R a trade on average after costs.
- Exhaustion fade (high hit-rate scalp): won 86.6% of 95,423 trades, yet lost 0.58R a trade on average after costs.
- Bollinger %B reversion on large-cap crypto: won 87.0% of 110,512 trades, yet lost 0.59R a trade on average after costs.
R is the size of the loss if the stop is hit. So -0.50R means an average trade lost half of that. This is a backtest, in-sample, not a promise.
It happens in the claims people share, too. A viral clip says 81% wins. On NQ it won 90.4% of 1,641 trades and still lost 0.84 points a trade over all of them, 1.15 on the later test split, net of costs. And: Claimed 73 to 90% wins. We measured 30.6% over 252 trades, net of costs. This is a backtest, in-sample, not a promise.
What multiple testing means
Flip a thousand coins ten times each. A few will land heads ten times in a row. Those coins are not special. You just flipped a lot of them.
Testing indicator rules works the same way. Try enough settings and timeframes and one will look brilliant on past prices. The fix is to count every rule you tried and ask how good the best one would look if none had any edge. That is the luck check we ran on all 1,774 tests. 0 passed. This is a backtest, in-sample, not a promise.
What would count as evidence
- The rule is written down before you test it, with no tweaking after you see results.
- It makes money after realistic costs: fees, spread and slippage.
- It still works on data it never saw while you built it.
- It beats random entries that use the same exits.
- It survives the correction for how many rules you tried.
- It works on more than one market, and other people can repeat it.
Our own indicator score is not exempt. It describes what ten indicators say about a chart. It is not a tested trade signal.
What this does not mean
It does not prove that no indicator ever works. We tested popular and published rules, on daily prices, over one stretch of history. A different rule or market could behave differently. It does mean that an indicator looking good on a chart is not evidence.
Questions
Do trading indicators actually work?
In our tests, no. We backtested 1,342 trading strategies built on common indicators and chart rules, and after correcting for luck across 1,774 tests, 0 passed. This is a backtest, in-sample, not a promise. It does not prove that no rule ever works.
Can an indicator with a high win rate still lose money?
Yes. In our quick tests, 1,295 rule and target pairs won at least 60% of their trades, and none made money after costs. Win rate ignores how big the wins and losses are. This is a backtest, in-sample, not a promise.
What is multiple testing?
If you test many rules on the same past prices, some will look good by luck alone. Multiple testing is the fix: you count how many rules you tried, then ask how good the best one would look if none of them had any edge.
Keep going
General information only. It doesn't consider your objectives, finances or needs. Tickfloor holds no financial services licence and never places trades.