Analysis — Direction
Recomputed from btc.db on every request · 2026-11-18 15:12:02 UTC
The phase filter applies to all sections. Only evaluated UP/DOWN rows are included. Strength = max(p_up, p_down). Accuracy uses stored correct (missing values are recomputed; ties predict UP). Brier = mean((p_up − outcome)²), where UP = 1 and DOWN = 0; lower is better (50/50 baseline: 0.25). Empty groups show —. Invalid probability rows excluded: 0; correct values recomputed: 0.
Phase comparison
| Phase | Model | N | Correct | Accuracy | Brier score | Avg prediction strength |
|---|---|---|---|---|---|---|
| Phase 3 | GPT | 48 | 20 | 41.7% | 0.2683 | 57.8% |
| Phase 3 | Jev | 48 | 23 | 47.9% | 0.3569 | 77.0% |
Direction accuracy by realized volatility
Retrospective / conditional on realized volatility: subsequent-hour RV is known only afterward. This is exploratory and not tradable prospectively. Phase 4 thresholds were calibrated after the direction experiments. Observations from different models for the same hour are not independent.
RV = 100 × sqrt(sum(log(C_i / C_(i−1))²)), exactly twelve consecutive 5m returns including the preceding close. QUIET: RV < 0.227425610%; NORMAL: 0.227425610% ≤ RV < 0.370664156%; ACTIVE: RV ≥ 0.370664156%.
GPT / OpenAI: ACTIVE 43.8% (N=16), non-ACTIVE 41.9% (N=31); difference +1.8 percentage points. Jev: ACTIVE 37.5% (N=16), non-ACTIVE 51.6% (N=31); difference -14.1 percentage points.
Missing RV excluded: 2; invalid probability rows excluded: 0. ALL includes only RV-assigned rows. Stored actual_return is in percent; missing/invalid returns are excluded only from magnitude metrics (Return N). No interpolation or network fetching occurs in this page.
| Phase | Model | Eligible N | RV assigned | Missing RV |
|---|---|---|---|---|
| phase3 | openai | 48 | 47 | 1 |
| phase3 | jev | 48 | 47 | 1 |
| Phase | Model | Actual regime | N | Correct | Accuracy | Brier | Return N | Mean |return| | Median |return| | Mean RV |
|---|---|---|---|---|---|---|---|---|---|---|
| phase3 | GPT / OpenAI | ALL | 47 | 20 | 42.6% | 0.2683 | 47 | 0.2985% | 0.2033% | 0.3658% |
| phase3 | GPT / OpenAI | QUIET | 9 | 5 | 55.6% | 0.2494 | 9 | 0.1536% | 0.1301% | 0.1958% |
| phase3 | GPT / OpenAI | NORMAL | 22 | 8 | 36.4% | 0.2696 | 22 | 0.1854% | 0.1537% | 0.2784% |
| phase3 | GPT / OpenAI | ACTIVE | 16 | 7 | 43.8% | 0.2771 | 16 | 0.5353% | 0.4955% | 0.5816% |
| phase3 | GPT / OpenAI | NON-ACTIVE | 31 | 13 | 41.9% | 0.2637 | 31 | 0.1762% | 0.1308% | 0.2544% |
| phase3 | Jev | ALL | 47 | 22 | 46.8% | 0.3625 | 47 | 0.2985% | 0.2033% | 0.3658% |
| phase3 | Jev | QUIET | 9 | 5 | 55.6% | 0.3237 | 9 | 0.1536% | 0.1301% | 0.1958% |
| phase3 | Jev | NORMAL | 22 | 11 | 50.0% | 0.3435 | 22 | 0.1854% | 0.1537% | 0.2784% |
| phase3 | Jev | ACTIVE | 16 | 6 | 37.5% | 0.4104 | 16 | 0.5353% | 0.4955% | 0.5816% |
| phase3 | Jev | NON-ACTIVE | 31 | 16 | 51.6% | 0.3377 | 31 | 0.1762% | 0.1308% | 0.2544% |
Calibration analysis
Blue bar: actual accuracy. Black marker: ideal accuracy, equal to the mean prediction strength in that bucket. Buckets include their lower bound and exclude their upper bound, except 100% is included in the last bucket.
GPT
| Strength bucket | N | Accuracy | Ideal (mean strength) | Actual / ideal · 0–100% |
|---|---|---|---|---|
| 50%–55% | 16 | 31.2% | 53.1% | |
| 55%–60% | 17 | 47.1% | 57.1% | |
| 60%–70% | 15 | 46.7% | 63.5% | |
| 70%–80% | 0 | - | - | — |
| 80%–90% | 0 | - | - | — |
| 90%–100% | 0 | - | - | — |
Jev
| Strength bucket | N | Accuracy | Ideal (mean strength) | Actual / ideal · 0–100% |
|---|---|---|---|---|
| 50%–55% | 5 | 60.0% | 51.8% | |
| 55%–60% | 3 | 33.3% | 58.7% | |
| 60%–70% | 9 | 33.3% | 64.8% | |
| 70%–80% | 7 | 57.1% | 71.7% | |
| 80%–90% | 10 | 70.0% | 85.5% | |
| 90%–100% | 14 | 35.7% | 94.3% |
Threshold analysis
Keep predictions whose strength is at least the threshold (not Jev's separate confidence field). Compare accuracy and Brier together with N; higher thresholds leave fewer observations. These are historical subsets, not evidence of future trading profit.
GPT
| Minimum strength | N | Accuracy | Brier score |
|---|---|---|---|
| ≥ 50% | 48 | 41.7% | 0.2683 |
| ≥ 55% | 32 | 46.9% | 0.2704 |
| ≥ 60% | 15 | 46.7% | 0.2856 |
| ≥ 70% | 0 | - | - |
| ≥ 80% | 0 | - | - |
| ≥ 90% | 0 | - | - |
Jev
| Minimum strength | N | Accuracy | Brier score |
|---|---|---|---|
| ≥ 50% | 48 | 47.9% | 0.3569 |
| ≥ 55% | 43 | 46.5% | 0.3691 |
| ≥ 60% | 40 | 47.5% | 0.3754 |
| ≥ 70% | 31 | 51.6% | 0.3940 |
| ≥ 80% | 24 | 50.0% | 0.4331 |
| ≥ 90% | 14 | 35.7% | 0.5608 |
Jev confidence investigation
Probability margin = abs(p_up − p_down). Evaluated Jev rows with valid confidence only. Missing or invalid confidence excluded: 0.
If confidence repeats the probability margin, r approaches 1 and the absolute difference approaches 0. The table groups pairs by margin (lower bound included; 100% included in the last group).
| Probability margin | N | Mean margin | Mean Jev confidence |
|---|---|---|---|
| 0–10% | 5 | 3.6% | 3.8% |
| 10–20% | 3 | 17.3% | 17.3% |
| 20–30% | 5 | 25.2% | 25.4% |
| 30–40% | 6 | 36.7% | 36.0% |
| 40–50% | 5 | 44.8% | 45.2% |
| 50–60% | 0 | - | - |
| 60–70% | 4 | 63.5% | 63.5% |
| 70–80% | 6 | 76.0% | 76.0% |
| 80–90% | 11 | 86.9% | 86.9% |
| 90–100% | 3 | 94.7% | 94.7% |
BTC Predictor