The record
Every signal each model has published, and what the price did afterwards — hit, missed, or still inside its window. Every number opens the trades behind it.
Hourly — honest record on the EXTENDED-hours basis.
Hourly is shown on its HONEST basis — the EXTENDED-hours session it actually trades (04:00–20:00 ET, touch on ext-session Highs). The 85% claim lives at the RULED 10-day swing window, where ELITE reaches 86.9% (rate is the bar per the 8/10 ruling — Wilson LB is dev-side only, not the display gate). The earlier bands are the speed profile. The conservative-daily master figures were known-LOW; these supersede them for the record.
BASIS DISCONTINUITY (disclosed, not hidden): the Hourly cohort is the pyrois board-tier fold graded on the extended basis through 2026-07-30; where that history meets the forward daily book the series is NOT continuous — figures on either side of that seam are not one series.
Return outcomes
how far Hourly signals ran, measured from the signal to the end of tracking (5 trading days)| Reached | Share of signals | Count | Median trading days to peak | 25th–75th |
|---|---|---|---|---|
| +10% | 64.6% | 217 | 3 | 1–4 |
| +20% | 47.9% | 161 | 3 | 2–5 |
| +30% | 35.7% | 120 | 3 | 2–5 |
| +50% | 20.5% | 69 | 4 | 2–5 |
| +100% | 9.5% | 32 | 4 | 2–5 |
| +150% | 7.7% | 26 | 4 | 3–5 |
| +200% | 5.7% | 19 | 4 | 3–5 |
| Reached | Share of signals | Count | Median trading days to peak | 25th–75th |
|---|---|---|---|---|
| +10% | 53.8% | 1381 | 3 | 2–5 |
| +20% | 30.5% | 783 | 4 | 2–5 |
| +30% | 18.8% | 482 | 4 | 2–5 |
| +50% | 7.3% | 188 | 4 | 2–5 |
| +100% | 2.8% | 71 | 3 | 2–5 |
| +150% | 1.6% | 42 | 2 | 1–4 |
| +200% | 1.2% | 30 | 2 | 1–4 |
| Reached | Share of signals | Count | Median trading days to peak | 25th–75th |
|---|---|---|---|---|
| +10% | 44.0% | 5206 | 4 | 3–5 |
| +20% | 18.5% | 2196 | 5 | 3–5 |
| +30% | 8.2% | 969 | 5 | 4–5 |
| +50% | 2.2% | 260 | 5 | 4–5 |
| +100% | 0.4% | 44 | 5 | 3–5 |
| +150% | 0.1% | 11 | 4 | 2–5 |
| +200% | 0.0% | 4 | 3 | 2–4 |
The top row is capped at +200%. Peak is measured on the adjusted daily high, so a stock split never reads as a move. Past performance does not predict future results. This is historical data, not investment advice. You are responsible for your own trading decisions.
Past performance does not predict future results. Historical blind walk-forward evidence, not investment advice.
The method behind these signals is published: OLYMPIA: Market-Regime-Conditioned Machine Learning for Equity Signal Generation — Rupert, Michael (Zenodo, 2026), DOI 10.5281/zenodo.18987223. The figures on this page are measured from the blind walk-forward record shown above, not taken from the paper.