Ignition analytics
What each signal type actually did next — auto-recorded at +1h / +6h / +24h. This table decides what becomes tradeable.
Volume ignitions — 32 events in 90 days
Fire-alarm events: 50x+ volume with price moving up
| Horizon | Measured | Median | Win rate (≥1x) | Best | Worst |
|---|---|---|---|---|---|
| +1 hour | 32/32 | 1.00x | 50% | 1.27x | 0.46x |
| +6 hours | 32/32 | 0.98x | 31% | 1.45x | 0.38x |
| +24 hours | 31/32 | 0.88x | 29% | 1.84x | 0.39x |
AI 10/25, 11:37 AMORBIO 10/20, 5:55 PM → 0.986x/24hSPCX 10/20, 2:51 PM → 1.033x/24hSPCX 10/20, 2:48 PM → 1.05x/24hTORCH 10/18, 8:25 AM → 1.373x/24hVRAX 10/14, 4:26 AM → 1.536x/24hAI 10/13, 10:20 PM → 0.704x/24hVRAX 10/12, 12:04 PM → 1.393x/24hAI 10/11, 11:25 AM → 0.812x/24hHOODIE 10/8, 3:59 PM → 0.947x/24hAI 10/8, 5:45 AM → 0.879x/24hPONS 10/7, 8:40 PM → 0.842x/24h
RH page adds — 2 events in 90 days
Robinhood published a new asset page — the catalyst itself
| Horizon | Measured | Median | Win rate (≥1x) | Best | Worst |
|---|---|---|---|---|---|
| +1 hour | 2/2 | 0.96x | 50% | 1.00x | 0.92x |
| +6 hours | 2/2 | 0.85x | 50% | 1.00x | 0.70x |
| +24 hours | 2/2 | 0.65x | 50% | 1.00x | 0.30x |
XDC 10/8, 7:50 AM → 1x/24hMON 10/5, 3:43 PM → 0.3x/24h
Ramps (data-only) — 57 events in 90 days
Slow grinds +40%+/3h — NOT a trade signal until this table earns it
| Horizon | Measured | Median | Win rate (≥1x) | Best | Worst |
|---|---|---|---|---|---|
| +1 hour | 57/57 | 0.93x | 25% | 1.74x | 0.44x |
| +6 hours | 57/57 | 0.91x | 26% | 3.09x | 0.00x |
| +24 hours | 56/57 | 0.80x | 30% | 2.45x | 0.00x |
SEASON 10/24, 9:53 PMSEASON 10/24, 6:29 AM → 1.425x/24hRAETO 10/22, 2:20 PM → 0.771x/24hANTI 10/20, 2:53 AM → 1.032x/24hHARMONIC 10/20, 12:16 AM → 1.004x/24hTORCH 10/19, 8:53 PM → 1.192x/24hTORCH 10/18, 10:16 AM → 1.103x/24hPORKU 10/16, 10:15 AM → 0.991x/24hORBIO 10/16, 6:46 AM → 0.773x/24hVRAX 10/16, 4:40 AM → 0x/24hASTRO 10/15, 8:48 AM → 1.081x/24hIGNITION 10/14, 3:32 PM → 1.393x/24h
How to read this: outcomes are price-vs-alert multiples, snapshotted automatically — no cherry-picking, skipped alerts included. A signal type earns real money only when its win rate and median hold up over weeks of data. Until then it's a hypothesis. Sample sizes under ~10 are noise; don't conclude anything from them.