Can we tell at 09:30 whether the RTH range will come in above the median of the last 20 days (relr ≥ 1)?
Criterion met — BSS 13.8 % [9.9 ; 17.7], AUC 71.1 %, accuracy 65.0 % against a base rate of 51.8 %, n = 905 days pooled forward.
Criterion: BSS > 3 and AUC ≥ 60, pooled forward across 2023–2026
| Decision time | 09:30 New York — bars up to t−1 complete, the opening print and every tick of session t before 09:30 |
| Target | relr ≥ 1 — RTH range divided by the median of the 20 previous RTH ranges |
| Data window | 2020-05-29 to 2026-08-14 · 1607 daily bars · last day 2026-08-14 |
| Walk-forward | Dev ≤ 2022 (definitions only), test 2023 / 2024 / 2025 / 2026, expanding with a yearly refit |
| Model | L2 logistic regression on 15 preregistered set C features, standardised on the training window; C from inner validation (the last training year). Features: gap_atr, abs_gap_atr, on_range_atr, ln_on_relr, on_vol_rel, on_delta_norm, open_pos_on, on_ext_recency, pre_range_share, on_high_vs_rh1_atr, on_low_vs_rl1_atr, open_pos_va, prev_ln_relr_rth, prev_ln_relr_mean5, prev_atr5_atr20 |
| Holdout | The PX holdout from 2026-05-01 concerns the PX theses on the tick store. This study is a walk-forward over 2023 to 2026 on daily data — explicitly approved by the user for regime rounds 1 and 2. Every test year is computed from data before it only; no fit ever sees its own test slice. |
Every row is a forecast over the same days. Base rate, majority and persistence are the baselines; the model has to beat them, otherwise it does not count.
| Row | Year | n | Base Rate % | Acc % | AUC % | Brier | BSS % |
|---|---|---|---|---|---|---|---|
| Base Rate | 2023 | 248 | 49.6 | 49.6 | 50.0 | 0.2500 | 0.0 |
| Majority | 2023 | 248 | 49.6 | 49.6 | — | — | — |
| Persistence | 2023 | 248 | 49.6 | 55.2 | 55.1 | 0.2496 | 0.2 |
| Model (LogReg) | 2023 | 248 | 49.6 | 60.5 | 64.6 | 0.2324 | 7.0 |
| Base Rate | 2024 | 251 | 56.2 | 43.8 | 50.0 | 0.2501 | 0.0 |
| Majority | 2024 | 251 | 56.2 | 43.8 | — | — | — |
| Persistence | 2024 | 251 | 56.2 | 55.0 | 54.5 | 0.2486 | 0.6 |
| Model (LogReg) | 2024 | 251 | 56.2 | 61.4 | 69.4 | 0.2207 | 11.8 |
| Base Rate | 2025 | 251 | 47.8 | 47.8 | 50.0 | 0.2508 | 0.0 |
| Majority | 2025 | 251 | 47.8 | 47.8 | — | — | — |
| Persistence | 2025 | 251 | 47.8 | 63.3 | 63.1 | 0.2365 | 5.7 |
| Model (LogReg) | 2025 | 251 | 47.8 | 70.5 | 75.8 | 0.2018 | 19.5 |
| Base Rate | 2026 | 155 | 54.8 | 54.8 | 50.0 | 0.2494 | 0.0 |
| Majority | 2026 | 155 | 54.8 | 54.8 | — | — | — |
| Persistence | 2026 | 155 | 54.8 | 57.4 | 57.6 | 0.2435 | 2.3 |
| Model (LogReg) | 2026 | 155 | 54.8 | 69.0 | 75.2 | 0.2032 | 18.5 |
| Base Rate | pooled | 905 | 51.8 | 48.4 | 47.1 | 0.2501 | 0.0 |
| Majority | pooled | 905 | 51.8 | 48.4 | — | — | — |
| Persistence | pooled | 905 | 51.8 | 57.8 | 56.2 | 0.2446 | 2.2 |
| Model (LogReg) | pooled | 905 | 51.8 | 65.0 | 71.1 | 0.2157 | 13.8 |
Block bootstrap (20-day blocks, 1000 runs, pooled forward): BSS 95 % interval [9.9 ; 17.7].
Confusion matrix pooled: true 0 correct 292 of 436, true 1 correct 296 of 469 (recall 0 = 67.0 %, recall 1 = 63.1 %).
Regularisation chosen per refit — 2023: C = 0.01, n_train = 581 · 2024: C = 3, n_train = 829 · 2025: C = 0.03, n_train = 1080 · 2026: C = 0.3, n_train = 1331.
| Decile | n | mean p % | observed % | difference |
|---|---|---|---|---|
| 1 | 91 | 24.0 | 24.2 | 0.1 |
| 2 | 91 | 34.3 | 35.2 | 0.8 |
| 3 | 91 | 39.4 | 29.7 | -9.7 |
| 4 | 91 | 43.8 | 49.5 | 5.6 |
| 5 | 91 | 47.4 | 47.3 | -0.2 |
| 6 | 90 | 51.9 | 53.3 | 1.4 |
| 7 | 90 | 57.1 | 51.1 | -6.0 |
| 8 | 90 | 62.0 | 68.9 | 6.9 |
| 9 | 90 | 69.1 | 72.2 | 3.1 |
| 10 | 90 | 82.4 | 87.8 | 5.4 |
Sharpness: p ≥ 0.7 on 14.0 % of the days (n = 127), hitting 84.3 % there · p ≤ 0.3 on 9.0 % (n = 81), hitting 75.3 % there.
| Base Rate | the unconditional frequency in the training window, and at the same time the constant comparison forecast. |
| Baseline | the number a model has to beat. Three of them here: base rate, majority (always the more frequent class) and persistence. |
| Persistence | the forecast “today like yesterday” — the value of the same target on the previous day, turned into a rate. |
| BSS | Brier skill score: what percentage of the base rate constant's Brier score the model saves. 0 means equally good, negative means worse. |
| AUC | the probability that a random positive day is scored above a random negative one. 50 is a coin flip. |
| Brier | the mean squared error of the probability. Smaller is better. |
| Sharpness | on what percentage of the days the model says something clear (p ≥ 0.7 or ≤ 0.3) and how often it hits there. A calibrated model without sharpness is useless. |
| Value Area | the price band [rval, rvah] in which 70 % of the RTH volume traded. Always the previous day's here. |
| Trend Day | an RTH with a body ratio ≥ 0.6 — the body of the daily candle fills at least 60 % of its range. |
| Body Ratio | br = |rc − ro| / (rh − rl), computed on the RTH session. |
| relr | RTH range divided by the median of the 20 previous RTH ranges. relr ≥ 1 means: today is bigger than the typical one of the last four weeks. |
| Walk-Forward | every test year is computed from a model that has only seen data before it; the training window grows with every year. No fit sees its own test. |
The numbers in this report are recomputed on every publish — from data/daily-bars-*.json and data/open-features-*.json, which the extractor builds from the tick store.
Run: 1607 daily bars, 0.8 s.