HB Betting Experiment

Overview

Every workflow execution, settled-only performance, and model calibration for the three-year HB betting experiment.

Experimental model. Past results do not guarantee future performance. Current stage is paper validation. All figures below are computed from settled records only; any claim not reproducible from this database is provisional and is not published here.

System status

SYSTEM OK

Last run — · model v2.0.0

Total workflow runs

510

0 run(s) today

Bets recommended

418

1,050 pass / no-bet

No-bet + pass rate

71.5%

Share of candidates the model declined

Cumulative paper P&L

$7,593.71

386 settled paper positions

Verified real-money P&L

$498.05

30 verified settled bets

Settled-only ROI

28.2%

$28,710 risked · $36,802 returned

Win rate

59.6%

235W · 159L · 15P · 7V

Resolved predictions

416

1052 unresolved / declined

Maximum drawdown

$571.21

Current streak

3W

Longest: 7W / 6L

Brier score

0.230

Calibration of no-vig consensus probability

Last successful run

—

Manual run

Data freshness

1127s

Age of odds snapshot on last run

Feed health

Odds APIOK
KalshiOK
StorageOK

API requests remaining

4,460

Reported by Odds API on last run

Cumulative paper P&L

Settled paper positions only, in resolution order.

Results

Settled-only ROI over time

Running return on total amount risked.

Recommendation distribution

Bet / paper / pass / no-bet across all candidates.

Results by sport

Settled wins and losses per sport.

Performance by confidence tier

Win rate and ROI per tier (settled only).

Performance by model version

Net P&L and ROI per released model version.

Monthly run volume

Workflow executions recorded per month.

Calibration buckets

Predicted probability vs. observed win rate. Closer bars mean better calibration.