Independent strategy research
Do trend-following strategies actually work? We tested 280+.
Trend-following is the biggest category we test — more than 280 systems, over a quarter of the whole audit. About 79% were rejected, which is roughly the house average. What makes that number worth reading is the cause underneath it. The most common reason a trend system got cut has a name: trend-beta. It rode a bull market and billed the ride as skill.
What "trend-following" means in this test
Trend systems buy strength and sell weakness. Moving-average crosses, breakout continuation, Donchian channels, Supertrend flips, momentum filters — different plumbing, one bet: that a move already in motion keeps going long enough to pay for the ones that stall. It is the oldest idea in systematic trading and the most crowded corner of retail scripting. That popularity is exactly why it dominates our reject pile by raw count. We have tested more trend systems than any other type, and it is not close.
What happened when we tested 280+
About 79% failed once real spreads, commissions and slippage were in the model. That is the audit average, almost to the point. So trend-following is not special. It fails at the house rate, and because it is the biggest category, it fills more of the "no" pile than anything else.
Here is the part that separates it from the worst offenders. Grid and DCA bots die at a clean 100% — every configuration, no survivors. Trend systems do not. A real minority, call it one in five, cleared the bar with a conditional edge: an edge that shows up on specific instruments, specific regimes or specific timeframes, not everywhere its author claimed. Most fail. Some genuinely hold. The work is telling which is which.
Why most of them fail: trend-beta
Across the entire audit, trend-beta is the second most common reason we reject anything, behind only "no edge even before costs." And trend systems are where it concentrates. The mechanic is simple enough to be embarrassing.
Test a long-biased trend system over the last decade and it makes money. So does buying the index and going to sleep. The system did not forecast the trend. It inherited it. Strip out the market's own drift — the beta — and the "edge" leaves with it. Under walk-forward testing, the equity curve that looked like skill flattens into the benchmark with extra steps and extra cost. You paid commissions to underperform holding.
A large share of the rest showed no real edge at all, before a cent of cost went in. Those never forecast anything. They re-price the same trend everyone already owns and hand you its full drawdown with none of the diversification. Add costs and the thin ones tip over too.
The honest exception: the minority that held
We do not reject a category on principle, and trend-following earns its exception. Roughly one in five of the systems we tested kept a real edge after costs, after walk-forward, after the drift was stripped out. What survives tends to be narrower than the sales copy: a trend model that works on one liquid instrument at one timeframe, sized sensibly, not the "works on everything" claim printed on the script.
The survivors also earn their returns in drawdown you can actually sit through, which is the whole point. A curve that only looks smooth because the market went one direction for a decade is not a strategy you can trust into the next decade. The ones we kept still work when you take the tailwind away. That is a much shorter list than the category's reputation suggests — but it is not empty, and that honesty is the difference between us and a listicle ranking "the 7 best trend strategies."
How we test
Every strategy is ported to Python and run against real costs — spreads and commissions modeled from tick data, not a guessed flat number. 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 the strongest model tries to break every apparent winner, hunting look-ahead bias and impossible fills. We hash the code, so a trend system re-published under three names gets tested once, not three times. It is the same process that rejects roughly 78% of everything we test — and it is especially unkind to a curve that was only ever the market in disguise. This whole page is one slice of a running audit of 1,000+ strategies and indicators.
Research and education, not financial advice. No signals, no return promises. Independent, and not affiliated with TradingView.
Which trend systems survived, and who published them?
You now have the aggregate truth: about 79% rejected, trend-beta the top killer, a minority with a real edge. What this page does not give you is the names — which specific published trend strategies cleared our cost model, who wrote them, and the exact after-cost numbers behind every rejection. That is The No List: every strategy we audited, named, with its verdict and the reason it lived or died.
Get The No List →FAQ
Do trend-following strategies actually work?
Some do; most published ones don't. Of the 280-plus trend systems we tested, about 79% failed a real cost model — roughly the audit average. Unlike grid/DCA bots, they don't fail 100%: around one in five kept a conditional edge after costs and walk-forward testing.
What is trend-beta and why does it kill so many trend strategies?
Trend-beta is when a system makes money only because the market drifted one direction, not because it predicts anything. It's the second most common rejection reason in our whole audit. Strip out the market's own drift and the "edge" disappears with it.
If most fail, why isn't the failure rate 100% like grid bots?
Because a genuine trend edge is possible, just rare. Grid/DCA bets against trends and blows up in a fat tail. Trend-following bets with them, so a minority that adds real timing or filtering on top of the drift can survive costs. About one in five did.
How can I tell a real trend edge from a bull-market ride?
Run it walk-forward, and compare it to simply holding the instrument over the same window. If the curve flattens toward buy-and-hold once you subtract the market's drift, it's trend-beta. If it still adds return after costs and after the tailwind is removed, it may be real.