Track record

model history and forecast calibration
8 / 8resolved forecasts correctcurrent resolved sample · live and historical limits below
0.074probability error (Brier)lower is better · 0 is perfect · 0.25 means always guessing 50/50
99open forecastsdated in the register · pending stated resolution criteria

The probability scores above grade the current prediction register; some entries were reconstructed during the historical build and are not a live forecasting claim. The chart below asks a different question: if each Board state had become a standardized hypothetical stock model, how did it perform against the broad semiconductor market (SOX)? The live record contains model states saved since Jul 7, 2026. Older points appear only in an optional historical test, rebuilt after outcomes were known, so they are context — not evidence of skill. See how live and historical records are separated.

Live record · 1 month16 recorded model states

Every point below was recorded before the period it measures had ended.

live: 16 · historical test: 84
+12%0%-1%2026-072026-07
Each dot is one Board state’s hypothetical 1 month result versus SOX.
Book — would-be returnlive (12 holding periods)historical test
vs SOX (chips)+5.69%+7.0%
vs S&P 500-2.18%+36.6%

Live = the model state was recorded before its measurement window ended. “Historical test” is a hindsight backtest, not a performance claim. vs SOX strips the chip sector’s move — the honest skill test; vs S&P 500 includes it, so its historical figure is mostly the sector’s run, not skill.

The free record lets you decide whether the system earns trust. Pro Positioning publishes the current standardized stock-level model output: direction relative to SOX, normalized dollar illustrations, and the contributing drivers.