Aleph Edge · evidence
What we tested, and what failed
Updated 17 Oct 2026 · 9 ideas
Every idea we consider for Aleph Edge is written down before we look at the result: the data, the split into train, validation and test periods, the costs, and the bar it must clear. Most ideas fail. We publish those too, because a tool that only shows its winners is hiding its losers.
| Idea | What we measured | Result | Verdict |
|---|---|---|---|
| DCA Long Classic on major coins | Bot-months, test period | +1.90% per bot-month, 8 of 8 months positive | Passed |
| DCA Long Safe on major coins | Bot-months, test period | +1.45% per bot-month, 8 of 8 months positive (train +2.19%, validation +1.69%) | Passed (lower return than Classic) |
| The same template on the top 30 coins | Worst bot | −97.7% | Majors only |
| Predict the next 15m / 30m / 1h / 4h candle | 2,022 setups, after 0.12% cost | 0 passed. Break-even needs 68.6% / 63.3% / 59.4% / 54.6% hit rate | Failed |
| Short yesterday's biggest gainers | Perp, after fees and funding | Validation −0.78%, test +0.97% per trade, CI includes 0 | Failed |
| Start DCA after a −20% day | Bot-months | Validation −0.15%, below the control | Failed |
| Limit (maker) entry for signal bots | Net per trade | Fee saving lost to missed best signals | Failed |
| Market news moves BTC | 889 stories, 1h/4h/24h vs random hours | No type differs after correction | No effect found |
| Paper signal-bot variants (fresh only, AI risk gate, ranking, split risk) | Forward paper, 3 weeks | None beat its control | Failed |
What runs now: a paper farm of every template on live data since 17 October 2026. Its ranking appears only after 30 days and a minimum number of closed cycles per bot. Until then it shows "measuring".
Aleph Edge · Risk Lab · Today