# Weekly Calibration #14

**WEEKLY · CALIBRATION** · November 11, 2026 · MarketMania Research · Weekly series

Window: **Nov 2-8, 2026 UTC** · Cutoff: **Mon Nov 9, 2026, 16:00 UTC** · Methodology **v1.1 (2026-08-10)**, hash `e66c7e8c864a2233`

PDF: https://marketmania.ai/research/reports/weekly-calibration-2026-11-02.pdf · Open data (JSON): https://marketmania.ai/research/reports/weekly-calibration-2026-11-02.json

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

## Research Snapshot

| RESEARCH SNAPSHOT |  |  |  |
|---|---|---|---|
| **4,865** directional calls scored | **7** models (stable lineup) | of **9,250** mature forecasts | window **Nov 2-8, 2026 UTC** |
| hit = direction rule: tp1/tp2->hit, sl->miss, expiry->sign of gross pnl | FH gate **1h / 4h / 1d** | base field hit **44.8%** vs **61.1** stated | cutoff **Mon Nov 9, 2026, 16:00 UTC** |
| market: BTC net **-4.47%** (prior -2.25%) | ann vol **55.3%** (was 34.7%) | TOP5 volume **$14.26B**, -4% w/w | pairwise corr **0.94** (was 0.93) |

## Key finding

> **KEY FINDING · OBSERVATION (one weekly window)**
>
> ### The field's overconfidence gap moved +19.6pp -> +16.3pp week over week, 7 of 7 model lines narrowed, and field Brier went 0.2858 -> 0.2772.

## TL;DR

- **OBSERVATION -- The gap read +16.3pp this week.** Directional calls hit 44.8% against 61.1 stated mean confidence (issue #13: +19.6pp on 41.6% vs 61.2). Field Brier is 0.2772 (was 0.2858) and 0 of 7 model lines scored below 0.25, the score of an uninformative always-50% predictor. One window, descriptive.
- **7 of 7 models narrowed their gap.** Gap moves ran from gemini-3.1-pro's -4.4pp (+26.7pp -> +22.3pp) to claude-fable-5's -2.5pp (+20.1pp -> +17.6pp); 7 of 7 model lines narrowed week over week. claude-opus-5.5 is the lowest-Brier line (Brier 0.2609); grok-4.7 leads on hit-rate (45.9%).
- **The high-confidence buckets, N>=10 only.** gpt-6.1-sol 70-80 29.6% (n=27), gemini-3.1-pro 70-80 41.8% (n=297). 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 -$46.85 (gpt-6.1-sol, best) to -$86.17 (grok-4.7, worst), field -$464.89 against issue #13's -$869.65.

### 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 #13 for the back-to-back window Oct 26-Nov 1.

## 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 #14 the series has fourteen 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 | 516 | 38.8% | 45.0% | 56.3 | +11.3 | 0.2609 | 40.7%-49.3% |
| grok-4.7 | 771 | 58.1% | 45.9% | 59.4 | +13.5 | 0.2687 | 42.4%-49.4% |
| qwen-3.8-max | 663 | 51.6% | 44.5% | 59.2 | +14.7 | 0.2703 | 40.8%-48.3% |
| deepseek-v4-pro | 754 | 57.1% | 44.6% | 60.4 | +15.8 | 0.2744 | 41.0%-48.1% |
| claude-fable-5 | 602 | 45.3% | 42.5% | 60.1 | +17.6 | 0.2769 | 38.6%-46.5% |
| gpt-6.1-sol | 807 | 60.7% | 45.6% | 63.0 | +17.4 | 0.2804 | 42.2%-49.0% |
| gemini-3.1-pro | 752 | 56.6% | 44.8% | 67.1 | +22.3 | 0.3031 | 41.3%-48.4% |
| **Field (all models)** | **4,865** | - | **44.8%** | **61.1** | **+16.3** | **0.2772** | - |

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 | 41.4% (222) | 43.2% (380) | n/a (0) | n/a (0) |
| claude-opus-5.5 | 45.6% (506) | 10.0% (10) | n/a (0) | n/a (0) |
| deepseek-v4-pro | 46.3% (257) | 43.9% (492) | 25.0% (4) * | n/a (0) |
| gemini-3.1-pro | 40.0% (5) * | 47.2% (443) | 41.8% (297) | 33.3% (6) * |
| gpt-6.1-sol | 46.0% (113) | 46.2% (667) | 29.6% (27) | n/a (0) |
| grok-4.7 | 48.3% (379) | 43.6% (392) | n/a (0) | n/a (0) |
| qwen-3.8-max | 45.3% (349) | 43.2% (308) | 33.3% (3) * | 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 100.0% (n=3) *; deepseek-v4-pro 0.0% (n=1) *; gemini-3.1-pro 0.0% (n=1) *. Mid-scale ranking: 2 of the 6 models with N>=10 in both 50-60 and 60-70 out-hit 50-60 from 60-70 this week (issue #13: 1 of 6).

## Hit-rate by forecast horizon (FH)

| Model | 1h | 4h | 1d |
|---|---|---|---|
| claude-fable-5 | 42.8% (353) | 41.6% (202) | 44.7% (47) |
| claude-opus-5.5 | 45.7% (311) | 43.9% (164) | 43.9% (41) |
| deepseek-v4-pro | 43.3% (443) | 45.6% (261) | 50.0% (50) |
| gemini-3.1-pro | 43.7% (458) | 47.6% (248) | 41.3% (46) |
| gpt-6.1-sol | 43.2% (472) | 46.2% (275) | 61.7% (60) |
| grok-4.7 | 45.5% (462) | 46.5% (258) | 47.1% (51) |
| qwen-3.8-max | 43.7% (389) | 45.1% (233) | 48.8% (41) |

n in parentheses; * = N<10 (insufficient). 5 of 7 model lines hit better on 1d than on 1h this window (issue #13: 3 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 |
|---|---|---|---|---|---|
| gpt-6.1-sol | 807 | 39.6% | -$46.85 | +$33.85 | -$73.85 |
| claude-opus-5.5 | 516 | 37.4% | -$54.73 | -$3.13 | -$71.01 |
| qwen-3.8-max | 663 | 37.4% | -$62.57 | +$3.73 | -$79.82 |
| gemini-3.1-pro | 752 | 35.8% | -$69.29 | +$5.91 | -$92.96 |
| deepseek-v4-pro | 754 | 38.1% | -$69.60 | +$5.80 | -$79.57 |
| claude-fable-5 | 602 | 34.7% | -$75.69 | -$15.49 | -$94.40 |
| grok-4.7 | 771 | 38.1% | -$86.17 | -$9.07 | -$99.35 |

0 of 7 model lines finished net-positive and 4 gross-positive this week (issue #13: 0 of 7 net-positive, 0 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 lower volume

> **MARKET CHECK · OBSERVATION (two adjacent calendar weeks)**
>
> OBSERVATION — mean |1d move| 1.63% -> 2.16% (+33% rel), BTC realized vol 37.0% -> 41.0% (ann., hourly); field directional accuracy 41.6% -> 44.8% (+3.1pp), 7 of 7 models improved, sim win-rate up for 7 of 7, field sim PnL -$869.65 -> -$464.89 (adjacent calendar weeks Oct 26-Nov 1 vs Nov 2-8; descriptive, one pair of weeks, not a claim).
>
> Robustness: raw price-sign accuracy 42.8% -> 48.9% — the same direction as the trade-based hit rule.

| Measure | Oct 26-Nov 1 (week A) | Nov 2-8 (week B) | Change |
|---|---|---|---|
| Mean \|1d move\| (5 assets) | 1.63% | 2.16% | +33% rel |
| BTC realized vol (ann., hourly) | 37.0% | 41.0% | +4.0pp |
| Field directional accuracy | 41.6% | 44.8% | +3.1pp |
| Models improving hit-rate | -- | 7 of 7 | -- |
| Raw price-sign accuracy | 42.8% | 48.9% | +6.0pp |
| Field sim win-rate | 33.6% | 37.4% | +3.8pp |
| Field sim net PnL | -$869.65 | -$464.89 | -- |

Week A is exactly the issue-#13 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 three 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 (2026-10-26..2026-11-01, the published issue-#13 window) and week B (2026-11-02..2026-11-08) are both full weeks on the heads (0 and 0 old-id row(s) remapped); week A reproduces the published issue #13. The grok 4.5 -> 4.6 flip (2026-08-24T09:22:00+00:00) sits ten 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 +16.3pp this week against +19.6pp in issue #13.
- Gap and regime move together. The field hit 44.8% against 41.6% in issue #13 and stated mean confidence read 61.1 against 61.2, so the gap came out at +16.3pp: a field that hits 44.8% instead of 41.6% 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; no line has finished net-positive for a second issue running (issue #13: 0 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 4,865 scored calls sit inside one market regime. This window had trend on 4 of 7 days (issue #13: 1 of 7); every comparison with issue #13 is a comparison across regimes as well as across weeks. Fourteen 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 14): (1) rolling-1w forecasts emit daily since Aug 22, 2026 and rolling-1M daily since Aug 23, 2026, so inside this window (Nov 2-8) daily coverage is FULL, as it was in issues #4, #5, #6, #7, #8, #9, #10, #11, #12 and #13: 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 487). (2) This window is the third 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 the 7 lines carry 1,461 / 1,470 rows in this window and 1,470 in week A, 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 three 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 three windows back (inside the issue-#11 window), the grok 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits ten 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 #13 after the audit that found the raw table mixes two exchanges; issue #13 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 9 16:00 UTC) -- the definition pinned in issue #3 and carried by issues #4, #5, #6, #7, #8, #9, #10, #11, #12 and #13, which printed 82,895 at the Nov 2 16:00 cutoff. The pack reproduces that pin at the previous cutoff (control OK), so this issue's 87,760 is an additive step under one definition; counter deltas against issue #12 and earlier remain definitional.

## Counters & lineage

| Counter | Value | Counter | Value |
|---|---|---|---|
| Forecasts total | 10,281 | Mature (scored pool) | 9,250 |
| OK in gate | 9,250 | -- of them directional | 4,865 |
| Out of gate (1w / 1M) | 490 / 487 | -- of them sideways | 4,385 |
| Invalid | 54 | Pending (next issue) | 0 |
| Late closes | 0 | Uptime, all grids | 266 / 266 slots |
| Source file | weekly_metrics_2026-11-02.json | Report cutoff | Mon Nov 9, 2026, 16:00 UTC |
| Generated at (pipeline) | 2026-11-10 14:05 UTC | Methodology | v1.1 (2026-08-10) · hash e66c7e8c864a2233 |

Provider incident: on Nov 5 the provider behind qwen-3.8-max rejected 46 of its 210 first-attempt forecasts with an input-moderation error; those rows are counted in Invalid. From Nov 6 the engine retries such a call without the news block; 2 forecasts in this window (qwen-3.8-max: 2 on Nov 7) were produced that way and are flagged `news_stripped` in the open data. Hit rates in this report use the rows that resolved.

**THIS REPORT: 4,865 scored observations.** MARKETMANIA RESEARCH TO DATE (as of Nov 9, 2026 cutoff): 87,760 directional forecasts resolved since Jul 11 · 7 models tracked (current line-up; earlier versions folded into their successors' lineage) · 5 assets · 5 horizons · hourly · 72 published reports.

> **Issue #14.** Weekly Calibration is a living series. Issue #13's open questions -- "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?" and "claude-opus-5.5 kept the lowest Brier score for a second issue -- does it hold for a third issue, or change hands?" -- read this issue as: field gap +19.6pp -> +16.3pp (7 of 7 lines narrowed) on a field base that moved +3.1pp, a ninth issue running in opposite directions, and the lowest Brier score stayed with the same line for a third issue: claude-opus-5.5 at 0.2609 (issue #13: claude-opus-5.5 at 0.2676). Also on the record: gpt-6.1-sol's 70-80 bucket read 29.6% (n=27) against its own 60-70 at 46.2% (n=667). The claude-opus-5.5, gpt-6.1-sol and grok-4.7 rows are the third 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 ten 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 -3.3pp while the field base moved +3.1pp -- the ninth issue running in which the two move nearly one-for-one in opposite directions; does that hold for a tenth?
- claude-opus-5.5 kept the lowest Brier score for a third issue -- does it hold for a fourth 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 #14 | https://marketmania.ai/research/reports/consensus-watch-2026-11-02.pdf |
| Weekly Model Watch #14 | https://marketmania.ai/research/reports/model-watch-2026-11-02.pdf |
| Weekly Calibration #13 | https://marketmania.ai/research/reports/weekly-calibration-2026-10-26.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_2026w45,
  title  = {Weekly Calibration #14: confidence vs. direction-hit, Nov 2-8 2026},
  author = {{MarketMania Research}},
  year   = {2026}, month = {November}, day = {11},
  url    = {https://marketmania.ai/research/reports/weekly-calibration-2026-11-02.pdf},
  note   = {Methodology v1.1, hash e66c7e8c864a2233; source weekly_metrics_2026-11-02.json}
}
```
