Comparison of runs-to-wins converters against actual data

By Tangotiger

All team data from 1919-2000 was "binned" into one of 25 bins (5 classes of RS+RA x 5 classes of RS-RA).

The closer the run differential is to zero, the higher the best-fit "x" value is to .29. The higher the run differential, the more the best-fit value approaches .27. Since as the run differential approaches zero, it doesn't matter what value you use (there's little difference in the exponent being 1.03 ^ .29 or 1.03 ^ .27), then I would lean towards minimizing the win% values from the extreme teams just as much as the middle teams. A best-fit value of .28 will do that.

 
Quality ENV RSG RAG Actual Win% Tango win% ActualExp TangoExp ClayExp PatdavidExp (.28) ActualBestFit TangoBestFit
1 1 3.96 3.13 0.5906 0.5983 1.56 1.7 1.73 1.73 0.228 0.271
1 2 4.58 3.44 0.6245 0.6246 1.76 1.77 1.81 1.79 0.273 0.273
1 3 4.99 3.83 0.6214 0.6187 1.87 1.82 1.87 1.84 0.286 0.276
1 4 5.63 4.18 0.6373 0.6367 1.88 1.88 1.94 1.9 0.278 0.276
1 5 6.36 4.72 0.6384 0.6413 1.91 1.95 2.02 1.96 0.268 0.277
 
2 1 3.79 3.39 0.5499 0.5473 1.8 1.71 1.73 1.74 0.298 0.271
2 2 4.28 3.68 0.5671 0.5666 1.78 1.77 1.8 1.79 0.278 0.274
2 3 4.74 4.11 0.5678 0.5653 1.9 1.82 1.87 1.84 0.293 0.276
2 4 5.3 4.54 0.5754 0.5723 1.97 1.88 1.94 1.9 0.296 0.277
2 5 5.92 5.05 0.5724 0.5771 1.82 1.94 2.01 1.96 0.251 0.278
 
4 1 3.33 3.94 0.4275 0.4284 1.73 1.71 1.74 1.74 0.278 0.272
4 2 3.74 4.28 0.4378 0.4411 1.87 1.77 1.81 1.79 0.3 0.274
4 3 4.11 4.75 0.4328 0.4344 1.87 1.83 1.87 1.84 0.287 0.276
4 4 4.55 5.28 0.4257 0.4301 2 1.88 1.94 1.9 0.304 0.277
4 5 4.98 5.94 0.4191 0.4158 1.86 1.94 2.01 1.95 0.26 0.277
 
5 1 3.09 4.23 0.376 0.3686 1.61 1.72 1.75 1.75 0.24 0.271
5 2 3.48 4.55 0.3786 0.3834 1.85 1.77 1.81 1.79 0.294 0.274
5 3 3.75 5.12 0.3558 0.3617 1.9 1.82 1.87 1.84 0.295 0.275
5 4 4.21 5.7 0.3608 0.3613 1.89 1.89 1.94 1.9 0.279 0.277
5 5 4.68 6.39 0.3492 0.3527 1.99 1.95 2.02 1.96 0.287 0.277