The charge. Every crypto newsletter, signal group, and momentum bot rests on one
assumption: that a coin which has been rising will keep rising — that recent price trend
predicts the next move. Our own model, mom-1.0, is built on exactly that belief. So we
did the thing almost nobody does with their own product: we put it on the stand.
The evidence
We ran the exact model that posts on this site — no changes, no tuning — across two years of daily price history and eleven coins: 7,023 graded calls from October 2024 through August 2026. The model has no fitted parameters; its weights are fixed constants. That means history it never saw is a fair trial, not a rigged one.
Here is the record:
| Strategy | Hit rate | Margin |
|---|---|---|
| Momentum (what we publish) | 49.2% | ± 1.2 points |
| Mean-reversion (bet the opposite) | 50.8% | ± 1.2 points |
| Only the high-confidence calls (≥58%) | 48.9% | ± 1.9 points |
A coin flip is 50%. Every number above is a coin flip. The model's most confident calls — the ones it shouts loudest — do no better than its quietest. Conviction buys you nothing.
Cross-examination
Three objections, each one tested.
"Isn't the model just broken?" No. We re-derived every recent call's direction from raw exchange candles: 92 of 92 matched what the model stored. We re-checked every grade against the real closing price: 92 of 92, within a rounding error. The plumbing is sound — the losses are real.
"Then just flip it — bet against the model." Tempting, because over the last month, fading it would have won about 78% of the time. But over two years the fade is 50.8% — another coin flip. That hot streak was four weeks of noise, not a strategy. We even tested a version that fades only in volatile conditions, and one that switches to whatever worked recently: 49.7% and 49.1%. Nothing rescued it.
"Maybe this month is the anomaly." It's the opposite. Momentum's hit rate swings from 38% in a bad month to 63% in a good one, with a median of 48%. It reverts to its mean — and its mean is a coin flip.
The verdict
On the daily timeframe, momentum does not predict crypto. Neither does its reverse. The market is efficient enough at this horizon that yesterday's trend tells you almost nothing about tomorrow's close.
Why we're telling you this. Because it's true, and the alternative is lying. Liquid markets are hard to beat — that isn't a scandal, it's finance. What separates an honest signal site from a sales funnel isn't a magic edge; it's whether they'll publish the guilty verdict on their own model. We just did. Every call we post is still graded here in public, coin flip and all — see the record. If someone selling you daily crypto calls won't show you a graded, calibrated track record, you already have your verdict.