# Weekly Calibration #13

**WEEKLY · CALIBRATION** · November 4, 2026 · MarketMania Research · Weekly series

Window: **Oct 26-Nov 1, 2026 UTC** · Cutoff: **Mon Nov 2, 2026, 16:00 UTC** · Methodology **v1.1 (2026-08-10)**, hash `e66c7e8c864a2233`

PDF: https://marketmania.ai/research/reports/weekly-calibration-2026-10-26.pdf · Open data (JSON): https://marketmania.ai/research/reports/weekly-calibration-2026-10-26.json

> Research question: *"when a model says 70, does it hit 70% of the time?"*

## Research Snapshot

| RESEARCH SNAPSHOT |  |  |  |
|---|---|---|---|
| **5,094** directional calls scored | **7** models (stable lineup) | of **9,304** mature forecasts | window **Oct 26-Nov 1, 2026 UTC** |
| hit = direction rule: tp1/tp2->hit, sl->miss, expiry->sign of gross pnl | FH gate **1h / 4h / 1d** | base field hit **41.6%** vs **61.2** stated | cutoff **Mon Nov 2, 2026, 16:00 UTC** |
| market: BTC net **-2.25%** (prior -2.39%) | ann vol **34.7%** (was 29.3%) | TOP5 volume **$14.86B**, +6% w/w | pairwise corr **0.93** (was 0.93) |

## Key finding

> **KEY FINDING · OBSERVATION (one weekly window)**
>
> ### The field's overconfidence gap moved +4.5pp -> +19.6pp week over week, 0 of 7 model lines narrowed, and field Brier went 0.2484 -> 0.2858.

## TL;DR

- **OBSERVATION -- The gap read +19.6pp this week.** Directional calls hit 41.6% against 61.2 stated mean confidence (issue #12: +4.5pp on 57.0% vs 61.5). Field Brier is 0.2858 (was 0.2484) and 0 of 7 model lines scored below 0.25, the score of an uninformative always-50% predictor. One window, descriptive.
- **No model narrowed its gap.** Gap moves ran from grok-4.7's +12.8pp (+4.3pp -> +17.1pp) to gemini-3.1-pro's +17.4pp (+9.3pp -> +26.7pp); 0 of 7 model lines narrowed week over week. claude-opus-5.5 is the lowest-Brier line (Brier 0.2676); grok-4.7 leads on hit-rate (42.5%). Every line is over-confident this window (positive gap); in issue #12 claude-opus-5.5 was the only line whose gap was negative (-1.5pp).
- **The high-confidence buckets, N>=10 only.** deepseek-v4-pro 70-80 30.0% (n=10), gpt-6.1-sol 70-80 26.9% (n=52), gemini-3.1-pro 70-80 41.4% (n=280), gemini-3.1-pro 80-100 36.4% (n=55). Cells below N=10 are marked insufficient and rank nothing.
- **Trading, kept separate from every calibration table.** 0 of 7 lines finished net-positive; net PnL ran -$93.22 (claude-opus-5.5, best) to -$161.50 (gemini-3.1-pro, worst), field -$869.65 against issue #12's -$43.00.

### Week-over-week chart (see PDF for the grouped bar chart)

Overconfidence gap in percentage points (mean stated confidence minus hit-rate), week over week, sorted by this week's gap. Short names: opus-5.5 = claude-opus-5.5, fable-5 = claude-fable-5, gemini-3.1 = gemini-3.1-pro, gpt-6.1 = gpt-6.1-sol, qwen-3.8 = qwen-3.8-max, deepseek = deepseek-v4-pro. "Was" values are the numbers published in issue #12 for the back-to-back window Oct 19-25.

## Why it matters

Every MarketMania forecast carries a model-stated confidence from 0 to 100 alongside its direction call. This report checks whether that number tracks reality: on a well-calibrated forecaster, calls made at 70% confidence should hit their direction about 70% of the time. With issue #13 the series has thirteen points on every model's gap -- enough to see movement, not enough to claim a trend: calibration and market regime move together, and durability claims belong in the Monthly series, not here.

## How to read this

Gap pp = mean stated confidence minus hit-rate, in percentage points; positive means overconfident. Brier is the mean squared error of the stated probability against the realised outcome, so lower is better and 0.25 is what an uninformative always-50% predictor scores. Coverage = calls scored divided by the mature calls available to that model. Confidence buckets are the models' own natural breakpoints, not an arbitrary binning; cells below N=10 are marked insufficient and never used to rank anything. Prediction and trading metrics live in separate sections and are never combined.

## Calibration by model (sorted by Brier, best first)

| Model | N | Cov. | Hit rate | Mean conf | Gap pp | Brier | 95% CI |
|---|---|---|---|---|---|---|---|
| claude-opus-5.5 | 591 | 44.4% | 40.9% | 56.4 | +15.4 | 0.2676 | 37.0%-45.0% |
| grok-4.7 | 830 | 62.4% | 42.5% | 59.6 | +17.1 | 0.2754 | 39.2%-45.9% |
| qwen-3.8-max | 711 | 53.7% | 41.3% | 59.0 | +17.7 | 0.2763 | 37.8%-45.0% |
| deepseek-v4-pro | 745 | 56.0% | 41.5% | 60.2 | +18.8 | 0.2805 | 38.0%-45.1% |
| claude-fable-5 | 633 | 47.6% | 40.4% | 60.5 | +20.1 | 0.2835 | 36.7%-44.3% |
| gpt-6.1-sol | 829 | 62.3% | 42.5% | 63.2 | +20.7 | 0.2922 | 39.1%-45.9% |
| gemini-3.1-pro | 755 | 56.8% | 41.7% | 68.4 | +26.7 | 0.3205 | 38.3%-45.3% |
| **Field (all models)** | **5,094** | - | **41.6%** | **61.2** | **+19.6** | **0.2858** | - |

N = directional calls scored this window. Cov. = coverage, scored / mature calls available to that model. Gap pp = mean stated confidence minus hit-rate (positive = overconfident). 95% CI is the descriptive Wilson interval on hit-rate.

## Calibration curve: hit-rate by stated-confidence bucket

| Model | 50-60 | 60-70 | 70-80 | 80-100 |
|---|---|---|---|---|
| claude-fable-5 | 43.0% (207) | 39.2% (426) | n/a (0) | n/a (0) |
| claude-opus-5.5 | 42.2% (559) | 18.8% (32) | n/a (0) | n/a (0) |
| deepseek-v4-pro | 40.4% (287) | 42.4% (448) | 30.0% (10) | n/a (0) |
| gemini-3.1-pro | 50.0% (2) * | 42.6% (418) | 41.4% (280) | 36.4% (55) |
| gpt-6.1-sol | 47.3% (131) | 42.7% (646) | 26.9% (52) | n/a (0) |
| grok-4.7 | 44.5% (402) | 40.8% (426) | 0.0% (2) * | n/a (0) |
| qwen-3.8-max | 43.0% (388) | 39.4% (310) | 50.0% (2) * | n/a (0) |

n in parentheses. * = N<10 (insufficient) -- no conclusions drawn from these cells. "n/a (0)" = no calls landed in that bucket this week. Sub-50 bucket (n greater than 0 only): qwen-3.8-max 36.4% (n=11). Mid-scale ranking: 1 of the 6 models with N>=10 in both 50-60 and 60-70 out-hit 50-60 from 60-70 this week (issue #12: 5 of 6).

## Hit-rate by forecast horizon (FH)

| Model | 1h | 4h | 1d |
|---|---|---|---|
| claude-fable-5 | 43.8% (395) | 34.5% (197) | 36.6% (41) |
| claude-opus-5.5 | 44.3% (361) | 33.3% (192) | 47.4% (38) |
| deepseek-v4-pro | 45.5% (442) | 34.2% (260) | 44.2% (43) |
| gemini-3.1-pro | 44.5% (456) | 36.0% (250) | 44.9% (49) |
| gpt-6.1-sol | 45.5% (501) | 37.4% (273) | 40.0% (55) |
| grok-4.7 | 45.4% (513) | 36.0% (267) | 48.0% (50) |
| qwen-3.8-max | 45.5% (429) | 33.2% (238) | 45.5% (44) |

n in parentheses; * = N<10 (insufficient). 3 of 7 model lines hit better on 1d than on 1h this window (issue #12: 7 of 7).

## Trading result (a different question)

A different question: is any of this profitable to trade. "Accurate" and "profitable" are not the same thing, and per methodology this section is never mixed into the prediction/calibration metrics above.

| Model | Trades | WR | Net PnL | Gross PnL | Max DD |
|---|---|---|---|---|---|
| claude-opus-5.5 | 591 | 33.2% | -$93.22 | -$34.12 | -$98.12 |
| qwen-3.8-max | 711 | 34.5% | -$105.90 | -$34.80 | -$112.94 |
| grok-4.7 | 830 | 34.5% | -$109.74 | -$26.74 | -$114.44 |
| claude-fable-5 | 633 | 31.3% | -$127.76 | -$64.46 | -$131.13 |
| deepseek-v4-pro | 745 | 34.8% | -$127.93 | -$53.43 | -$133.18 |
| gpt-6.1-sol | 829 | 33.7% | -$143.60 | -$60.70 | -$149.67 |
| gemini-3.1-pro | 755 | 33.0% | -$161.50 | -$86.00 | -$166.35 |

0 of 7 model lines finished net-positive and 0 gross-positive this week (issue #12: 3 of 7 net-positive, 7 gross-positive). Ranked best-to-worst net PnL. Per methodology this table is never combined with the calibration tables above; a week's PnL is regime as much as skill.

## Platform note: TP/SL calibration

> **Context, not a finding.** Since Aug 13 the public sandbox can trade with model-specific TP/SL multipliers learned from this same weekly history (v1); since Aug 19, v2 adds per-ticker and confidence-bucket (50-70 / 70-100) resolution. That feature consumes calibration history; it does not feed back into any table in this report, which measures stated confidence vs. direction-hit only. No effect size is claimed for v2. The toggle lives at marketmania.ai/indices.

## Market check: bigger daily moves on higher volume

> **MARKET CHECK · OBSERVATION (two adjacent calendar weeks)**
>
> OBSERVATION — mean |1d move| 1.35% -> 1.63% (+20% rel), BTC realized vol 27.5% -> 37.0% (ann., hourly); field directional accuracy 57.0% -> 41.6% (-15.3pp), 0 of 7 models improved, sim win-rate up for 0 of 7, field sim PnL -$43.00 -> -$869.65 (adjacent calendar weeks Oct 19-25 vs Oct 26-Nov 1; descriptive, one pair of weeks, not a claim).
>
> Robustness: raw price-sign accuracy 56.0% -> 42.8% — the same direction as the trade-based hit rule.

| Measure | Oct 19-25 (week A) | Oct 26-Nov 1 (week B) | Change |
|---|---|---|---|
| Mean \|1d move\| (5 assets) | 1.35% | 1.63% | +20% rel |
| BTC realized vol (ann., hourly) | 27.5% | 37.0% | +9.5pp |
| Field directional accuracy | 57.0% | 41.6% | -15.3pp |
| Models improving hit-rate | -- | 0 of 7 | -- |
| Raw price-sign accuracy | 56.0% | 42.8% | -13.2pp |
| Field sim win-rate | 46.6% | 33.6% | -13.0pp |
| Field sim net PnL | -$43.00 | -$869.65 | -- |

Week A is exactly the issue-#12 window, so this slice and every week-over-week column in this report compare the same two back-to-back weeks. Descriptive, one pair of weeks; the hit rule is methodology v1.1 (trade-based) and the raw price-sign row is the robustness check (hourly closes at :00 against slots at :01). BTC realized vol here is the hourly-return series from the alive slice; the snapshot row's figure is the audited daily-candle estimate -- different estimators, both reported as measured. Lineage note: the model cutover (2026-10-17T09:01:00+00:00, day 6 of 7 of its Mon-Sun week) sits two windows back, inside the published issue-#11 window, where it spliced three lines -- gpt-5.6-sol -> gpt-6.1-sol, claude-opus-5 -> claude-opus-5.5, grok-4.6 -> grok-4.7. Week A (2026-10-19..2026-10-25, the published issue-#12 window) and week B (2026-10-26..2026-11-01) are both full weeks on the heads (0 and 0 old-id row(s) remapped); week A reproduces the published issue #12. The grok 4.5 -> 4.6 flip (2026-08-24T09:22:00+00:00) sits nine windows back (inside issue #4's window).

## Practical implications

- Stated confidence is still a ranking hint at best, not a probability: the field's gap is +19.6pp this week against +4.5pp in issue #12.
- Gap and regime move together. The field hit 41.6% against 57.0% in issue #12 and stated mean confidence read 61.2 against 61.5, so the gap came out at +19.6pp: a field that hits 41.6% instead of 57.0% closes or opens a confidence gap without a single stated number changing.
- Trading direction this week (0 of 7 lines net-positive) is a regime read, not a strategy result; the three lines that were net-positive one week earlier all finished net-negative (issue #12: 3 of 7 net-positive).

## Limitations

- Prediction metrics (this report) and trading metrics are kept in separate sections per methodology -- they are never combined into a single score.
- 95% Wilson CIs shown are descriptive, not inferential: observations inside one window are dependent, so read them as a range, not a formal coverage guarantee.
- All 5,094 scored calls sit inside one market regime. This window had trend on 1 of 7 days (issue #12: 1 of 7); every comparison with issue #12 is a comparison across regimes as well as across weeks. Thirteen week-over-week points cannot separate drift from regime; no durability claim is made.
- Models report confidence at discrete levels, not a continuous scale; the buckets reflect those natural breakpoints. Cells below N=10 are marked insufficient.
- Underlying price series are reconstructed from trade entry prices (median per symbol-slot), not an independent tick feed. Series density and lineage notes (wave 13): (1) rolling-1w forecasts emit daily since Aug 22, 2026 and rolling-1M daily since Aug 23, 2026, so inside this window (Oct 26-Nov 1) daily coverage is FULL, as it was in issues #4, #5, #6, #7, #8, #9, #10, #11 and #12: the 1w series has slots on 7 of 7 days and the 1M series on 7 of 7 days; both sit outside this report's FH gate (1h/4h/1d) and appear only in the exclusions counter (1w 490, 1M 488). (2) This window is the second full week on the builds that took over at Oct 17, 2026 09:01 UTC: claude-opus-5.5, gpt-6.1-sol and grok-4.7 carry no rows from the old builds in either week, and all 7 lines carry 1,470 rows each in both weeks, so every week-over-week row compares two full weeks on the same builds. (3) No roll-out slot of the flip falls inside either week: the two hourly slots before it (Oct 17, 07:01 and 08:01 UTC) lie two windows back, so this window's invalid counter holds none of their rows. (4) Past lineage flips: the Oct 17, 2026 09:01 UTC flip sits two windows back (inside the issue-#11 window), the grok 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits nine windows back (inside the issue-#4 window), and qwen-3.8-max succeeded qwen-3.7-max on Aug 5, 2026 (fully before this window).
- Market-state row is computed on a single-exchange (binance) daily candle series, as in issue #12 after the audit that found the raw table mixes two exchanges; issue #12 values are as published.
- Research-to-date counter: directional forecasts resolved inside the FH gate (1h/4h/1d) since Jul 11, counted at the issue cutoff (Nov 2 16:00 UTC) -- the definition pinned in issue #3 and carried by issues #4, #5, #6, #7, #8, #9, #10, #11 and #12, which printed 77,801 at the Oct 26 16:00 cutoff. The pack reproduces that pin at the previous cutoff (control OK), so this issue's 82,895 is an additive step under one definition; counter deltas against issue #11 and earlier remain definitional.

## Counters & lineage

| Counter | Value | Counter | Value |
|---|---|---|---|
| Forecasts total | 10,290 | Mature (scored pool) | 9,304 |
| OK in gate | 9,304 | -- of them directional | 5,094 |
| Out of gate (1w / 1M) | 490 / 488 | -- of them sideways | 4,210 |
| Invalid | 8 | Pending (next issue) | 0 |
| Late closes | 0 | Uptime, all grids | 266 / 266 slots |
| Source file | weekly_metrics_2026-10-26.json | Report cutoff | Mon Nov 2, 2026, 16:00 UTC |
| Generated at (pipeline) | 2026-11-03 11:35 UTC | Methodology | v1.1 (2026-08-10) · hash e66c7e8c864a2233 |

**THIS REPORT: 5,094 scored observations.** MARKETMANIA RESEARCH TO DATE (as of Nov 2, 2026 cutoff): 82,895 directional forecasts resolved since Jul 11 · 7 models tracked (current line-up; earlier versions folded into their successors' lineage) · 5 assets · 5 horizons · hourly · 64 published reports.

> **Issue #13.** Weekly Calibration is a living series. Issue #12's open questions -- "the gap moved -10.9pp while the field base moved +10.4pp -- the seventh issue running in which the two move nearly one-for-one in opposite directions; does that hold for an eighth?" and "The best-Brier line changed hands again, to claude-opus-5.5 -- does it hold for a second issue, or change hands again?" -- read this issue as: field gap +4.5pp -> +19.6pp (0 of 7 lines narrowed) on a field base that moved -15.3pp, an eighth issue running in opposite directions, and the lowest Brier score stayed with the same line for a second issue: claude-opus-5.5 at 0.2676 (issue #12: claude-opus-5.5 at 0.2443). Also on the record: gpt-6.1-sol's 70-80 bucket read 26.9% (n=52) against its own 60-70 at 42.7% (n=646). The claude-opus-5.5, gpt-6.1-sol and grok-4.7 rows are the second full week on the Oct 17 builds; week-over-week rows compare two full weeks on the same builds. The grok 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits nine windows back. Engine 1.1 has powered the sandbox since Aug 18, i.e. before this window; no cross-engine PnL comparisons are claimed.

## What we're testing next

- Next issue: the gap moved +15.1pp while the field base moved -15.3pp -- the eighth issue running in which the two move nearly one-for-one in opposite directions; does that hold for a ninth?
- claude-opus-5.5 kept the lowest Brier score for a second issue -- does it hold for a third issue, or change hands?
- Monthly series: is the overconfidence gap stable across market regimes (trend vs flat) at monthly n?

## Related research

| Report | Direct PDF link |
|---|---|
| Consensus Watch #13 | https://marketmania.ai/research/reports/consensus-watch-2026-10-26.pdf |
| Weekly Model Watch #13 | https://marketmania.ai/research/reports/model-watch-2026-10-26.pdf |
| Weekly Calibration #12 | https://marketmania.ai/research/reports/weekly-calibration-2026-10-19.pdf |

The three weekly reports publish together as one issue each week; each links straight to the others' PDF and to its own previous issue. Direct links are the posting rule from wave 2 on.

## Cite this report

```bibtex
@misc{mm_calibration_2026w44,
  title  = {Weekly Calibration #13: confidence vs. direction-hit, Oct 26-Nov 1 2026},
  author = {{MarketMania Research}},
  year   = {2026}, month = {November}, day = {4},
  url    = {https://marketmania.ai/research/reports/weekly-calibration-2026-10-26.pdf},
  note   = {Methodology v1.1, hash e66c7e8c864a2233; source weekly_metrics_2026-10-26.json}
}
```
