The test that decides whether an edge is real
Every signal seller frames the same fantasy: their system “beats the market.” So we ran the only fair version of that test — every strategy in our library, head-to-head against the dumbest possible benchmark: buy it and hold it. No timing, no genius, just sit there.
We did it across BTC, ETH, SOL and BNB over the 2017–2026 window, two full bear markets (2018 and 2022) included — because a strategy’s real test is whether it survives a crash, and a bull-only sample flatters everything.
What this table is, and is not — first. The numbers below are a nine-year Binance daily-bar study (harness:
research/grid/fetch_binance_daily.py, collection starts 2017-07-14; each asset contributes from its own listing date). It is not the deployed strategy. The live trend ensemble runs on three OKX assets (BTC, ETH, SOL) and BNB is not in it — because the study did not pass BNB (per-asset verdicts 3/4: BTC ✅ ETH ✅ SOL ✅ BNB ❌ — holding BNB was unusually strong, so de-risking cost more than it saved). Why study it on Binance: OKX’s own daily bars only start in 2023-12, giving 2.4 bull-only years, and in that window this edge (crash avoidance) cannot be validated at all. The research write-up says so itself.
The result is blunt:
No directional strategy beat holding. After 17 strategy families across 15 research campaigns, the count of deployable directional strategies that beat holding on return is still 0. (One cell below looks like an exception: the floor-0.5 CAGR of 69% edges the 68% of holding. At a 1%p gap that reads as matching, not beating, and what that row actually sells is drawdown. So we do not flatly say “zero.”)
The head-to-head scoreboard
Here is the equal-weight BTC/ETH/SOL/BNB portfolio — hold versus our trend-following ensemble — over the full nine-year, bear-inclusive window:
| Metric | Buy-and-hold | Trend ensemble | Trend ensemble (floor 0.5) |
|---|---|---|---|
| CAGR (annual return) | 68% | 63% | 69% |
| Max drawdown | −81% | −46% | −65% |
| Sharpe (risk-adjusted) | 1.09 | 1.29 | 1.21 |
| Calmar (return ÷ drawdown) | 0.84 | +63% vs hold | +26% vs hold |
Read the top row carefully. On raw return, buy-and-hold won or tied every time. The ensemble’s CAGR was lower (63% vs 68%) until we dialed in an exposure floor — and even at its best (floor 0.5), it only matched holding’s return (69% vs 68%). It never beat it.
So if the only question is “did anything make more money than holding?” — the honest answer is no.
What actually changed — and why it still matters
The return column is not the whole story. Look at the drawdown row.
Holding BTC/ETH/SOL/BNB meant living through an −81% drawdown. The ensemble cut that almost in half, to −46% — same neighborhood of return, roughly half the pain. That is the entire edge: it does not predict price; it sidesteps the worst of the crash. We call it crisis-alpha, not a crystal ball.
The year-by-year breakdown shows exactly what kind of trade-off that is:
- 2021 bull: +652% vs +1311% — the ensemble gave up half the upside.
- 2022 bear: −25% vs −74% — but it defended ~50 percentage points of capital.
- 2026 drawdown: −16% vs −37% — another 21-point defense.
It is not magic. It is a deliberate exchange: surrender some of the bull to survive the bear. Whether that is a good trade depends entirely on whether you can stomach an 81% drawdown — and on a leveraged product, you literally cannot (−81% is a liquidation; −46% is survival).
”Verified” does not mean “deployable”
We are precise about status. Verified means reproducible — the numbers above come from a fixed spec on nine years of multi-asset data, and you can re-run the logic yourself. It does not mean “live and printing money.” The size of the risk-adjusted edge depends on the window: it is largest in bear-inclusive periods and shrinks in a strong bull run. The robust, honest claim is survival, not a return multiple.
That is why we publish the failures instead of burying them. A platform that only shows winners is indistinguishable from the signal sellers it competes with. The full scoreboard — every loss, and the one thing that survived — is the product.
Try to beat hold yourself
Don’t take our word for it — but the table above does not reproduce in the simulator. Three things block it: (1) this ensemble is not one of the simulator’s presets; (2) the simulator’s benchmark is fixed to BTC, so you cannot build an equal-weight four-asset hold; (3) the simulator’s OKX data starts in 2023-12, so the two bear markets in that table are simply absent. Reproducing it means running the research harness (research/grid/multiasset_ensemble.py).
What the simulator can show you is what an individual strategy does in its own window — go run that. See the full method and every result in the open on our trust page, read the deeper writeup of why directional prediction fails, and see the one strategy that controlled drawdown without sacrificing return: the trend ensemble. If you decide holding is the right call, do it intelligently with a portfolio plan.
Don’t believe us. Verify.