THE FULL REGISTER / EVERY TEST, GRADEDCRYPTO · OUT-OF-SAMPLE

We backtested 48 trading strategies. 8 survived.

Every famous signal, plus a pile of our own research, run through the identical honest test: multi-year, fees included, judged on data it never saw, measured against simply holding. Here is the whole scoreboard, nothing hidden.

Failed outright

No edge · coin flip

Survived the test

01 What 48 honest tests actually showed

The headline is not that trading is hopeless. It's that almost everything sold as an edge isn't one. Of the famous signals everyone quotes, most beat simply holding on roughly half the coins, which is a coin flip you pay fees to make. A few lost money outright. And a small handful genuinely held up, on data they were never fitted to.

The failures died in a few repeating ways. The mean-reversion crowd (RSI, Bollinger, Stochastic) bought the knife: on a trending asset, "oversold" just means cheaper tomorrow. The busy trend-followers (MACD, Donchian) paid the whipsaw tax, catching the big moves and handing the gains back one crossover at a time. And the famous ones people trust most (the golden cross, the 200-day) turned out to be coin flips in a respectable-looking wrapper. Our own research died more quietly, mostly overfit to one regime or killed by costs once tested honestly.

02 The whole scoreboard

Survived8 of 48

Beat holding across most coins and stayed ahead out-of-sample. Not money printers, and past results still aren't the future, but they didn't fall apart the way the famous ones did.

9/21 EMA CrossoverMomentumsurvived — 7/10, held out-of-sampleRead autopsy → Ichimoku CloudMomentumsurvived — 7/10, held out-of-sampleRead autopsy →

EMA-50 daily flipMomentumout-of-sample validatedrecord only

Retail L/S contrarian gateAlt-dataOOS ~2x, drawdown halvedrecord only

FVG displacementMomentumpositive every year 2020–26record only

FVG hybrid (bull-gated)Momentumdrawdown 65%→42%record only

Directional J-line (BTC-gated)Momentumthe resurrectionrecord only

Gold J-line (directional)Momentum+162% / 2.9x, uncorrelatedrecord only

No edge — a coin flip10 of 48

Beat holding on four to six coins out of ten. You can't tell the result apart from luck, and you pay fees for the privilege.

The Golden CrossMomentumbeat holding on 5/10 — a coin flipRead autopsy → MACD CrossoverMomentum80 trades, no edgeRead autopsy → SupertrendMomentumcoin flip in a green wrapperRead autopsy → Donchian / Turtle BreakoutMomentumbig returns, still 6/10Read autopsy → 200-Day Moving AverageMomentuma seatbelt, not an edgeRead autopsy →

Ranging scalper (retuned)Mean reversionneeds >70.7% winrecord only

Funding-neutral carryArbitrage~15% CAGR, cost-sensitiverecord only

Catapult breakoutMomentumregime-gated pyramidrecord only

Catapult dualMomentumlong-only vs dual A/Brecord only

Listing sniperAlt-datafade-then-flip, on trialrecord only

Failed30 of 48

Lost money outright, or beat holding on three or fewer coins. Including the extra tweaks and filters that were supposed to help and didn't.

RSI Oversold BounceMean reversion67% win rate, still lostRead autopsy → Bollinger Band BounceMean reversion1/10 — bought the knifeRead autopsy → Stochastic OscillatorMean reversionlost money outrightRead autopsy →

KDJ J-line + ADX (plain)Momentum+661% in-sample → −97% full cyclerecord only

Signal copierAlt-datareal −$15 over 10 daysrecord only

Meme launch sniperAlt-data57,408 seen, 0 worth enteringrecord only

Smart-money copy-tradeAlt-data−70% in 4 days, slippage ate itrecord only

OI + funding squeezeArbitrage36 variants, no edgerecord only

Cointegration pairsArbitrageloses out-of-samplerecord only

Retail vs institutional divergenceAlt-datatop-trader is noiserecord only

Instant intra-candle entryMomentumworse than closed-candlerecord only

Fast 1h exitMomentumexits winners too earlyrecord only

Slow 6h/8h exitMomentumgives the gains backrecord only

1h timeframeMomentum4h is the peakrecord only

2h timeframeMomentum4h is the peakrecord only

6h timeframeMomentum4h is the peakrecord only

8h timeframeMomentum4h is the peakrecord only

CHOP filterMomentumhurt the lean strategyrecord only

BBW-squeeze filterMomentumhurt the lean strategyrecord only

Kaufman efficiency filterMomentumhurt the lean strategyrecord only

Parabolic SAR filterMomentumhurt the lean strategyrecord only

DMA filterMomentumhurt the lean strategyrecord only

MACD + CHOP + SAR stackMomentumevery filter hurtrecord only

More pairs (0.77 corr)Momentumcorrelation kills diversificationrecord only

Single-pair concentrationMomentumruin at 10xrecord only

10x leverageMomentumcompounded blow-uprecord only

Tight hard stopsMomentumwhipsawed outrecord only

Silver J-lineMomentumrejectedrecord only

TradFi stocks / indicesMomentummarket-hours gapsrecord only

Gold-native confluenceMomentumlean beats the stackrecord only

03 The verdict on the whole lot

Out of 48 honest tests, 8 survived. That's the real number, and it's why "I backtested it and it works" means almost nothing on its own. The test is easy to pass by accident and hard to pass for real. Everything here is graded the same way, and you can read the full autopsy on the famous ones.

How each was tested — spot, long/flat, no leverage · daily bars over each coin's full history · a signal read at the close is filled at the NEXT close (no look-ahead) · 0.06% fee each side, buy-and-hold charged one entry fee too · out-of-sample = the most recent ~45% of each coin's history held back. The full rules, and exactly what earns each verdict, are on the how we grade page. Every figure is a simulation on historical data; past performance is not the future.

Think yours is one of the survivors?

Soon you'll be able to run your own strategy through this exact test and see which side of the scoreboard it lands on. Get notified when it opens →

Or see the trick behind every "amazing" backtest first: how a backtest lies →