Independent strategy research
Most trading strategies fail before costs. The #1 killer is no edge at all.
Across the 1,000+ strategies and 1,700+ indicators we've run through a real cost model, about 78% of the strategies get rejected — and most were dead before a cent of cost came off. The biggest single reason we throw a strategy out is no real edge at all: the rules never beat a coin-flip entry on the same bars, gross, before spread or commission. Cost, the thing every trader blames, is only our fourth most common reason. Most rejected systems were never alive.
Here's the part the listicles skip. We didn't just count how many strategies failed. We logged why each one died, ranked the reasons, and the order is not the one you'd guess.
What "fails before costs" actually means
Start with the phrase, because it does the work. A strategy has a gross edge when its entries beat a random coin-flip on the same bars, with no spread, no commission and no slippage in the math yet. No real edge means the signal never forecast the next move. The entries were noise wearing the costume of logic.
Add costs to a strategy like that and nothing changes. There was nothing there to erode. It didn't lose its edge at the broker. It never had one.
The order strategies die in
We tag every rejection with a single primary reason, then count. The ranking above is the result, and four things stand out.
No edge at all is number one by a wide margin. Roughly four in ten rejected strategies land here. Trend-beta is second: the system made money because the market drifted up while it happened to be long, and buy-and-hold would have matched it with fewer rules to break. Tail-concentrated profit is third. Strip out a handful of jackpot days and the curve goes flat or negative. Then, and only then, cost-fatal: a genuine edge on paper that real spread and commission ate whole.
Group the reasons that all mean no edge existed: no gross edge, systems that fail a placebo test, and ones firing on too few trades to trust. Together they're nearly half of everything we reject. That's the honest headline. Not costs killed them. Most were never alive.
Why so many never had an edge
Published strategies are selected for one thing: a good-looking equity curve on the chart the author posted. Nobody screenshots the version that lost. So the pool you're browsing is survivorship all the way down, the curves that happened to fit a past that's already gone.
A rising decade flatters almost any long-only rule, which is why trend-beta shows up so often. A lucky cluster of outlier days makes a mediocre system look spectacular, and that's tail-concentrated profit. A rule tuned on noise looks immaculate until you shuffle the data underneath it. Our placebo test randomizes the trade timing and re-runs, and a meaningful share of profitable-looking systems collapse the moment their timing is scrambled — because the pattern lived in the fit, not the market.
Most fail. But not all.
Here's where this differs from the bleakest corner of our audit. Grid, DCA and martingale bots fail at a clean 100%, the only whole category that wipes out with zero survivors (we wrote up how that one dies). The broad population isn't that. About one in five strategies clear the no-edge bar and show something real before costs. A smaller slice keep a conditional edge once real costs are in. Fewer than one in a hundred come out cleanly deployable across our full 13-year confirmation.
So most fail, and a thin minority don't. The job was never to sneer at the whole field. It's to find the few systems that survive every check, which is what the rest of our research is about.
How we test
Every strategy is ported to Python and run against real costs — spread and commission modeled from tick data, not a flat guess. Futures come from Databento, 13 years of CME. FX comes from Dukascopy with real bid/ask. Stocks get liquidity-aware fills; crypto runs as spot and perps. A fast model does the bulk porting, then our strongest model tries to break every apparent survivor, hunting look-ahead bias and fills that peeked at the future. We hash the code, so a strategy re-published under three names gets tested once. And we run a placebo control on everything: if a system can't beat a shuffled version of itself, it found a coincidence and named it a strategy. It's the same process that rejects roughly 78% of what we test.
Research and education, not financial advice. No signals, no return promises. Independent, and not affiliated with TradingView.
Which ones survived? Named, with verdicts.
You now have the aggregate truth for free: most strategies fail, and usually because they never had an edge — not because costs killed them. What this page can't give you is the names. Which specific published strategies cleared no-edge, trend-beta, the tail test, the placebo control and real costs, and which of the "profitable" ones we rejected, with the exact after-cost reason for each. That is The No List: every strategy we audited, named, with its verdict.
Get The No List →FAQ
Do most trading strategies actually work?
Most don't, in our testing. About 78% of the strategies we've put through a real cost model get rejected. Only a thin minority keep an edge after costs, and fewer than one in a hundred come out cleanly deployable.
What is the number one reason trading strategies fail?
No real edge at all — the rules never beat a coin-flip entry before a cent of cost. It's our most common reject reason by a wide margin, and combined with placebo and sparse-signal failures it accounts for nearly half of everything we throw out.
Don't trading costs kill most strategies?
Less often than people think. Cost-fatal is only our fourth most common reject reason, behind no edge, trend-beta and tail-concentrated profit. Most rejected systems had no edge for costs to erode in the first place.
Do all strategies fail, like grid and DCA bots?
No. Grid, DCA and martingale bots are the only category we've seen fail at 100%. Across the broader field roughly one in five clear the no-edge bar and a smaller slice survive costs, so most fail but not all.