I backtested every strategy sold in trading courses. 13 of 16 lose to a coin flip.
There is a large industry built on selling trading strategies — RSI bounces, MACD crosses, order blocks, liquidity sweeps, the whole vocabulary of the $500-to-$2,000 course. Almost none of these products show you the one number that matters: how the strategy performs against firing the same trade at random moments. So we built the test they skip. We ran sixteen of the most-marketed strategies on 90 days of real 5-minute data across the twenty most-liquid crypto perpetuals, applied real trading fees, and benchmarked every one against a random-entry control with identical risk. Thirteen of the sixteen performed no better than coin-flipping. Here is the full data — and the free tool that lets you run it on your own strategy in ten seconds.
The one benchmark every course skips
Here is the trick that makes almost any long strategy look profitable in a screenshot: the market drifts upward over the sample. If crypto rallied during your test window, then buying at random times and holding also made money. So a strategy showing a positive return has proven nothing until you know how random entries did with the same stop-loss and take-profit. That comparison — the random-entry control — is the difference between "this signal works" and "the market went up while I happened to be long."
No retail backtester we know of shows it. TradingView's strategy tester lets you optimise parameters until the curve looks great, then hides the fact that you just fit to noise. Course sellers show you the one cherry-picked configuration that worked and never the hundred that didn't. The random control is the cheapest, most brutal honesty check in trading, and it is conspicuously absent everywhere money changes hands.
The setup
Every result below uses the same harness. Ninety days of 5-minute candles across the twenty highest-volume USDT-margined perpetuals. A flat 0.20% round-trip cost (fees plus slippage) applied to every trade — generous, if anything, for a taker. Each strategy formalised mechanically so there is no discretion to hide behind. And a time-split so we can see whether an effect that appeared in the first two-thirds of the window survived into the last third. The random control: the identical stop and target fired at random moments, which in our window netted −0.17% per trade. That is the line every strategy has to beat.
The results
Net return is per-trade, after fees. The t-statistic tells you whether the result is signal or noise — below an absolute value of 2, it is indistinguishable from luck.
| Strategy (as taught) | Trades | Net / trade | t-stat | Verdict |
|---|---|---|---|---|
| Bollinger lower-band bounce | 11,938 | −0.20% | −19.9 | Dead |
| Bullish engulfing | 11,303 | −0.19% | −16.6 | Dead |
| Pin bar / hammer | 5,866 | −0.20% | −14.7 | Dead |
| MACD signal cross | 13,430 | −0.17% | −10.9 | Dead |
| Wyckoff spring | 2,058 | −0.21% | −10.5 | Dead |
| VWAP bounce | 2,592 | −0.18% | −8.0 | Dead |
| EMA 9/21 golden cross | 9,373 | −0.18% | −7.7 | Dead |
| RSI oversold bounce | 4,012 | −0.19% | −6.9 | Dead |
| Liquidity sweep (SMC flagship) | 1,496 | −0.26% | −4.4 | Worse than random |
| Break of structure (BOS) | 2,284 | −0.20% | −4.1 | Dead |
| Volume-spike breakout | 1,988 | −0.17% | −4.0 | = random |
| Fair value gap (imbalance) | 1,151 | +0.05% | 1.1 | Noise |
| Ichimoku Tenkan/Kijun | 6,785 | −0.16% | −6.7 | Dead |
| Golden pocket (0.618 fib) | 462 | +0.80% | 2.6 | Partial |
| Demand zone (rally-base-rally) | 59 | +0.57% | 2.6 | Partial |
| Order block retest (SMC) | 604 | +0.83% | 5.0 | Survives |
What the survivors have in common
Three strategies came out positive: order block retest, golden pocket, and demand zone. Here is the punchline that the marketing never mentions — they are the same effect wearing three costumes. Strip away the vocabulary ("institutional order block", "the golden 0.618", "supply and demand") and all three are the same mechanical setup: buy a pullback to the origin of a fresh upward impulse. That is momentum-pullback, one of the oldest documented effects in the academic literature. It works not because of anything mystical about a Fibonacci level or "smart money footprints", but because a market that just moved impulsively tends to continue after a shallow retrace.
And note the irony sitting in the table: the single most-marketed setup in the entire Smart Money Concepts canon — the liquidity sweep — tested worse than random entries. The story ("smart money grabbed the stops below the low, now price reverses") is compelling and unfalsifiable. The cash flow is −0.26% per trade across fifteen hundred trades. If a mechanism story cannot beat a coin flip with the same risk, the story is decoration on a losing pattern.
Test your own strategy — free, in ten seconds
We turned this exact engine into a free public tool. Pick a strategy, set your stop and target, and see the honest result against the random-entry control and the full parameter landscape. No signup, no paywall. Nobody else gives you the random control — that's the point.
Open Reality Check →Why "high win rate" is the most expensive lie in trading
Look again at RSI oversold: 48% win rate, and it still loses money. This is the trap that sells more courses than any other. A strategy can win most of its trades and still bleed, because the winners are small (a bounce to the mid-band) and the losers are the trades where "oversold" kept being right all the way down. Crypto downtrends routinely pin RSI below 30 for hours; the strategy systematically buys the middle of liquidation cascades. Win rate tells you nothing on its own. Only the full distribution — net expectancy after fees — tells you whether you have an edge.
How to not get fooled
Before you pay for any strategy, demand three numbers from whoever is selling it: the sample size, the net expectancy after realistic fees, and the performance versus a random-entry control with the same stop and target. If the vendor cannot produce all three, you are the exit liquidity. Most cannot, because most of what is sold is a −0.2%-per-trade pattern packaged with a confident narrative and a screenshot of the one week it worked.
We publish the opposite. Every hypothesis we have tested — the dead ones and the handful that survived — is documented with the math and the verdict in our research log. We do this because intellectual honesty is the only durable moat in a field this crowded with mirages, and because a firm that shows you its failures is telling you something a firm that only shows winners never can.
Trade with a desk that kills its own mirages
We run our strategies on our own capital and report them honestly — including the ones that failed. No fabricated win rates, no cherry-picked screenshots.
Try the free backtester →