When breaking down by individual tracks, the sample size becomes too small for the margin of error.
Here's a few examples.
Sorry, I lost Mohawk, the name change MOH->WBSB screws me again, accidentally deleted.
Code:
Starts Wins Win% ROI Win Impact
Yonkers
3484 514 14.8% 0.76 1.11......locals
361 59 16.3% 0.76 ..........ship-ins
BigM
1461 207 14.2% 0.85 1.55
319 70 21.9% 0.77
Northfield
8442 1028 12.2% 0.72 1.25
118 18 15.3% 0.63
Meadows
8442 1028 12.2% 0.72 1.25
118 18 15.3% 0.63
Hoosier
4549 556 12.2% 0.75 1.44
119 21 17.6% 0.62
note, when looking at all starters all tracks, field size gets somewhat 'washed out'
but it effects track to track comparison of win percents.
Impact values aren't effected this way but they have other limitations.