What this is
Recompute 's per-year OLS (reproduced exactly at point values), then leave-one-out and a 20,000-draw Monte Carlo over the counter's printed-precision bounds. The r≈0.97 is the floors-as-points case; within TPC's own bounds the correlation is usually below 0.9 and the slope spans 0.3–0.8.
Where it comes from
Our calculation from other figures in the trainer’s notes:
- EstimateTail against rows proportionality by window, Nov 2017 – Mar 2026
- ReportedHow much TPC's cumulative game count could have grown in each year from March 2020 to March 2026, given how precisely it was printed. Some years were printed only to the nearest 10 million, so each annual rise is a range
- EstimateNov2017 Mar2020 window tail half open, 24 November 2017 to 31 March 2020
Show the working
- tails drawn uniformly in / bounds (FY3/21 (5.96,16.96), FY3/22 (15.9,35.9) with joint (71,82), FY3/23 (5.7,21.01), FY3/24+25 [0,5.31), FY3/26 (8.31,10.31), W1 delta (67,69) less rows-inside (30.6,39.27))
- per-year OLS of tail on rows: r median 0.83, 5–95% 0.58–0.95
- slope 0.32–0.84
- intercept −0.5 to 6.8