Seven days, 1,184 races, and a first-pick win rate of 26.4% — that is the headline from the past week at RTD, and it is a number that holds up well against the sport's natural randomness. Picking randomly from a six-dog field gives you 16.7% by chance. The model is producing first-pick winners at more than one-and-a-half times that rate, over a volume that makes it meaningful rather than coincidental.
The week was not uniform. June 13 was the standout: 66 first-pick winners from 246 races — 26.8% — with 155 of those races producing a top-three selection as the actual winner. That is 63% coverage on a heavy Saturday card. June 14 was the week's toughest day at 23.5% first-pick wins from 119 races, while yesterday (June 18) dipped to 20.5% from 146 races, pulling the weekly average down slightly.
The any-top-three metric is where calibration shows most clearly. Over the seven days, at least one of the three rated selections finished first in 64.4% of races, and placed in the top three in 89.2% of races. That last figure tells you the model's rankings are reading fields accurately — the gap between ranked correctly and won is variance from individual races, not systematic blind spots.
One pattern worth noting: four of the seven days stayed above 24% first-pick win rate. The dips to the low 20s are real, but they are the kind of daily fluctuation you would expect from any system operating over a large number of races across multiple grades and venues. No single grade or track type looks like an obvious weak point from this week's data.
For anyone using RTD for form study, the model is working as it should — identifying likely winners above chance rates and grouping competitive fields accurately enough that the top three selections place in nearly nine of every ten races.