# Weekly Calibration #10

**WEEKLY · CALIBRATION** · October 14, 2026 · MarketMania Research · Weekly series

Window: **Oct 5-11, 2026 UTC** · Cutoff: **Mon Oct 12, 2026, 16:00 UTC** · Methodology **v1.1 (2026-08-10)**, hash `e66c7e8c864a2233`

PDF: https://marketmania.ai/research/reports/weekly-calibration-2026-10-05.pdf · Open data (JSON): https://marketmania.ai/research/reports/weekly-calibration-2026-10-05.json

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

## Research Snapshot

| RESEARCH SNAPSHOT |  |  |  |
|---|---|---|---|
| **4,601** directional calls scored | **7** models (stable lineup) | of **9,249** mature forecasts | window **Oct 5-11, 2026 UTC** |
| hit = direction rule: tp1/tp2->hit, sl->miss, expiry->sign of gross pnl | FH gate **1h / 4h / 1d** | base field hit **43.8%** vs **61.9** stated | cutoff **Mon Oct 12, 2026, 16:00 UTC** |
| market: BTC net **-3.19%** (prior +2.44%) | ann vol **27.2%** (was 21.1%) | TOP5 volume **$15.97B**, -11% w/w | pairwise corr **0.85** (was 0.57) |

## Key finding

> **KEY FINDING · OBSERVATION (one weekly window)**
>
> ### The field's overconfidence gap moved +12.9pp -> +18.2pp week over week, 0 of 7 model lines narrowed, and field Brier went 0.2698 -> 0.2822.

## TL;DR

- **OBSERVATION -- The gap read +18.2pp this week.** Directional calls hit 43.8% against 61.9 stated mean confidence (issue #9: +12.9pp on 49.6% vs 62.5). Field Brier is 0.2822 (was 0.2698) 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 gpt-5.6-sol's +3.4pp (+21.3pp -> +24.7pp) to deepseek-v4-pro's +8.8pp (+9.3pp -> +18.1pp); 0 of 7 model lines narrowed week over week. claude-opus-5 is both the lowest-Brier line (Brier 0.2685) and the hit-rate leader (47.0%).
- **The high-confidence buckets, N>=10 only.** deepseek-v4-pro 70-80 41.7% (n=12), gemini-3.1-pro 70-80 40.8% (n=179), gpt-5.6-sol 70-80 49.6% (n=244). 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 -$40.48 (claude-opus-5, best) to -$98.53 (deepseek-v4-pro, worst), field -$520.82 against issue #9's -$408.24.

### 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 = claude-opus-5, fable-5 = claude-fable-5, gemini-3.1 = gemini-3.1-pro, gpt-5.6 = gpt-5.6-sol, qwen-3.8 = qwen-3.8-max, deepseek = deepseek-v4-pro. The grok column compares two pure grok-4.6 weeks: the 4.5 -> 4.6 flip sits six windows back. "Was" values are the numbers published in issue #9 for the back-to-back window Sep 28-Oct 4.

## 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 #10 the series has ten 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. The grok 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits six windows back (issue #4): both weeks of this issue are pure grok-4.6, with no 4.5-era rows in either window, so the grok week-over-week row is the fifth pure-4.6 against pure-4.6 comparison of the series.

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

| Model | N | Cov. | Hit rate | Mean conf | Gap pp | Brier | 95% CI |
|---|---|---|---|---|---|---|---|
| claude-opus-5 | 517 | 39.8% | 47.0% | 60.8 | +13.8 | 0.2685 | 42.7%-51.3% |
| claude-fable-5 | 613 | 47.1% | 45.0% | 59.9 | +14.9 | 0.2714 | 41.1%-49.0% |
| grok-4.6 | 547 | 41.1% | 43.3% | 59.7 | +16.3 | 0.2717 | 39.2%-47.5% |
| qwen-3.8-max | 707 | 53.2% | 43.1% | 58.5 | +15.4 | 0.2724 | 39.5%-46.8% |
| deepseek-v4-pro | 761 | 57.2% | 42.0% | 60.2 | +18.1 | 0.2792 | 38.6%-45.6% |
| gemini-3.1-pro | 722 | 54.3% | 43.4% | 65.2 | +21.8 | 0.2979 | 39.8%-47.0% |
| gpt-5.6-sol | 734 | 55.2% | 43.5% | 68.1 | +24.7 | 0.3059 | 39.9%-47.1% |
| **Field (all models)** | **4,601** | - | **43.8%** | **61.9** | **+18.2** | **0.2822** | - |

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. The grok 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits six windows back (issue #4): both weeks of this issue are pure grok-4.6, with no 4.5-era rows in either window, so the grok week-over-week row is the fifth pure-4.6 against pure-4.6 comparison of the series.

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

| Model | 50-60 | 60-70 | 70-80 | 80-100 |
|---|---|---|---|---|
| claude-fable-5 | 45.7% (256) | 44.5% (357) | n/a (0) | n/a (0) |
| claude-opus-5 | 50.4% (119) | 46.0% (398) | n/a (0) | n/a (0) |
| deepseek-v4-pro | 42.9% (296) | 41.5% (453) | 41.7% (12) | n/a (0) |
| gemini-3.1-pro | 80.0% (10) | 43.5% (529) | 40.8% (179) | 50.0% (4) * |
| gpt-5.6-sol | 54.5% (11) | 39.6% (472) | 49.6% (244) | 71.4% (7) * |
| grok-4.6 | 38.9% (239) | 46.9% (307) | n/a (0) | n/a (0) |
| qwen-3.8-max | 44.5% (400) | 41.4% (285) | 0.0% (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): grok-4.6 0.0% (n=1) *; qwen-3.8-max 45.0% (n=20). Mid-scale ranking: 1 of the 7 models with N>=10 in both 50-60 and 60-70 out-hit 50-60 from 60-70 this week (issue #9: 4 of 5).

## Hit-rate by forecast horizon (FH)

| Model | 1h | 4h | 1d |
|---|---|---|---|
| claude-fable-5 | 47.1% (380) | 44.1% (202) | 25.8% (31) |
| claude-opus-5 | 49.7% (314) | 45.8% (168) | 28.6% (35) |
| deepseek-v4-pro | 46.1% (432) | 40.1% (277) | 19.2% (52) |
| gemini-3.1-pro | 45.8% (443) | 43.0% (237) | 19.1% (42) |
| gpt-5.6-sol | 46.7% (433) | 42.4% (250) | 21.6% (51) |
| grok-4.6 | 44.3% (314) | 43.9% (198) | 31.4% (35) |
| qwen-3.8-max | 46.6% (423) | 40.7% (243) | 21.9% (41) |

n in parentheses; * = N<10 (insufficient). 0 of 7 model lines hit better on 1d than on 1h this window (issue #9: 5 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 | 517 | 37.7% | -$40.48 | +$11.22 | -$53.90 |
| grok-4.6 | 547 | 35.3% | -$52.78 | +$1.92 | -$79.76 |
| claude-fable-5 | 613 | 35.2% | -$64.62 | -$3.32 | -$69.05 |
| qwen-3.8-max | 707 | 33.7% | -$79.87 | -$9.17 | -$83.77 |
| gpt-5.6-sol | 734 | 33.7% | -$89.16 | -$15.76 | -$94.48 |
| gemini-3.1-pro | 722 | 33.8% | -$95.39 | -$23.19 | -$96.94 |
| deepseek-v4-pro | 761 | 33.5% | -$98.53 | -$22.43 | -$105.03 |

0 of 7 model lines finished net-positive and 2 gross-positive this week (issue #9: 0 of 7 net-positive, 4 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: a down week with bigger daily moves

> **MARKET CHECK · OBSERVATION (two adjacent calendar weeks)**
>
> OBSERVATION — mean |1d move| 0.87% -> 1.54% (+77% rel), BTC realized vol 32.2% -> 30.9% (ann., hourly); field directional accuracy 49.6% -> 43.8% (-5.9pp), 0 of 7 models improved, sim win-rate up for 0 of 7, field sim PnL -$408.24 -> -$520.82 (adjacent calendar weeks Sep 28-Oct 4 vs Oct 5-11; descriptive, one pair of weeks, not a claim).
>
> Robustness: raw price-sign accuracy 49.0% -> 45.5% — the same direction as the trade-based hit rule.

| Measure | Sep 28-Oct 4 (week A) | Oct 5-11 (week B) | Change |
|---|---|---|---|
| Mean \|1d move\| (5 assets) | 0.87% | 1.54% | +77% rel |
| BTC realized vol (ann., hourly) | 32.2% | 30.9% | -1.3pp |
| Field directional accuracy | 49.6% | 43.8% | -5.9pp |
| Models improving hit-rate | -- | 0 of 7 | -- |
| Raw price-sign accuracy | 49.0% | 45.5% | -3.5pp |
| Field sim win-rate | 41.0% | 34.5% | -6.5pp |
| Field sim net PnL | -$408.24 | -$520.82 | -- |

Week A is exactly the issue-#9 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 splice ACTIVE: grok-4.5 rows are aggregated into grok-4.6 (flip 2026-08-24T09:22:00+00:00, six windows back, inside issue #4's window). Week A (2026-09-28..2026-10-04) is pure grok-4.6 (it reproduces the published issue #9); week B is pure grok-4.6.

## Practical implications

- Stated confidence is still a ranking hint at best, not a probability: the field's gap is +18.2pp this week against +12.9pp in issue #9.
- Gap and regime move together. The field hit 43.8% against 49.6% in issue #9 and stated mean confidence read 61.9 against 62.5, so the gap came out at +18.2pp: a field that hits 43.8% instead of 49.6% 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; the same lines printed the opposite sign inside eight weeks.

## 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 4,601 scored calls sit inside one market regime. This window ran 2 trend days of 7 (issue #9: 2 of 7); every comparison with issue #9 is a comparison across regimes as well as across weeks. Ten 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 10): (1) rolling-1w forecasts emit daily since Aug 22, 2026 and rolling-1M daily since Aug 23, 2026, so inside this window (Oct 5-11) daily coverage is FULL, as it was in issues #4, #5, #6, #7, #8 and #9: 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 479). (2) The grok line flipped 4.5 -> 4.6 at Aug 24, 2026 09:22 UTC, six windows back (inside the issue-#4 window): this window carries 1,470 grok rows and none from the 4.5 era, and neither does week A (the issue-#9 window), so every grok number in this issue and in issue #9 is a pure grok-4.6 line -- the grok week-over-week row is the fifth pure-4.6 against pure-4.6 comparison of the series. (3) 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 #9 after the audit that found the raw table mixes two exchanges; issue #9 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 (Oct 12 16:00 UTC) -- the definition pinned in issue #3 and carried by issues #4, #5, #6, #7, #8 and #9, which printed 62,260 at the Oct 5 16:00 cutoff. The pack reproduces that pin at the previous cutoff (control OK), so this issue's 66,861 is an additive step under one definition; counter deltas against issue #8 and earlier remain definitional.

## Counters & lineage

| Counter | Value | Counter | Value |
|---|---|---|---|
| Forecasts total | 10,290 | Mature (scored pool) | 9,249 |
| OK in gate | 9,249 | -- of them directional | 4,601 |
| Out of gate (1w / 1M) | 490 / 479 | -- of them sideways | 4,648 |
| Invalid | 72 | Pending (next issue) | 0 |
| Late closes | 0 | Uptime, all grids | 266 / 266 slots |
| Source file | weekly_metrics_2026-10-05.json | Report cutoff | Mon Oct 12, 2026, 16:00 UTC |
| Generated at (pipeline) | 2026-10-13 15:15 UTC | Methodology | v1.1 (2026-08-10) · hash e66c7e8c864a2233 |

**THIS REPORT: 4,601 scored observations.** MARKETMANIA RESEARCH TO DATE (as of Oct 12, 2026 cutoff): 66,861 directional forecasts resolved since Jul 11 · 7 models tracked (current line-up; earlier versions folded into their successors' lineage) · 5 assets · 5 horizons · hourly · 52 published reports.

> **Issue #10.** Weekly Calibration is a living series. Issue #9's open questions -- "the gap moved -3.2pp while the field base moved +2.8pp -- the fourth issue running in which the two move nearly one-for-one in opposite directions; does that hold for a fifth?" and "The best-Brier line changed hands again, to grok-4.6 -- does it hold for a second issue, or change hands again?" -- read this issue as: field gap +12.9pp -> +18.2pp (0 of 7 lines narrowed) on a field base that moved -5.9pp, a fifth issue running in opposite directions, and the best-Brier line changed hands again: claude-opus-5 at 0.2685 (issue #9: grok-4.6 at 0.2566). Also on the record: gpt-5.6-sol's 70-80 bucket read 49.6% (n=244) against its own 60-70 at 39.6% (n=472). The grok row is pure grok-4.6 this issue and was pure grok-4.6 in issue #9; the 4.5 -> 4.6 flip (Aug 24, 2026 09:22 UTC) sits six 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 +5.3pp while the field base moved -5.9pp -- the fifth issue running in which the two move nearly one-for-one in opposite directions; does that hold for a sixth?
- The best-Brier line changed hands again, to claude-opus-5 -- does it hold for a second issue, or change hands again?
- Monthly series: is the overconfidence gap stable across market regimes (trend vs flat) at monthly n?

## Related research

| Report | Direct PDF link |
|---|---|
| Consensus Watch #10 | https://marketmania.ai/research/reports/consensus-watch-2026-10-05.pdf |
| Weekly Model Watch #10 | https://marketmania.ai/research/reports/model-watch-2026-10-05.pdf |
| Weekly Calibration #9 | https://marketmania.ai/research/reports/weekly-calibration-2026-09-28.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_2026w41,
  title  = {Weekly Calibration #10: confidence vs. direction-hit, Oct 5-11 2026},
  author = {{MarketMania Research}},
  year   = {2026}, month = {October}, day = {14},
  url    = {https://marketmania.ai/research/reports/weekly-calibration-2026-10-05.pdf},
  note   = {Methodology v1.1, hash e66c7e8c864a2233; source weekly_metrics_2026-10-05.json}
}
```
