Expected wins from SRS?

shawngoat23 · Post #1 » by **shawngoat23** » Mon Feb 18, 2019 9:37 am

I recently started reading El Gee's articles on Backpicks, and I noticed he uses SRS as a measure of a team's strength, followed by the equivalent win-pace. Is there a commonly used formula or table for converting SRS to expected wins or vice versa? Is this dependent on season (for instance, mean and SD of league SRS as parameters)? I hope I'm not missing something super obvious, but I couldn't find anything after about 30 minutes of searching.

shawngoat23 · Post #2 » by **shawngoat23** » Mon Feb 18, 2019 11:13 am

Taking a cursory glance at some of the SRS values and the expected win percentages and using a logistic model, I estimate that the win percentage is roughly 1 / ( 1 + exp( - 0.1476 x )), although the coefficient seems fairly inaccurate since I'm fitting the parameter based on two significant figure data points (integer number of wins and SRS values).

eminence · Post #3 » by **eminence** » Mon Feb 25, 2019 4:19 pm

I'll ask him how he does his on Twitter.

eminence · Post #4 » by **eminence** » Mon Feb 25, 2019 5:03 pm

shawngoat23 wrote:I recently started reading El Gee's articles on Backpicks, and I noticed he uses SRS as a measure of a team's strength, followed by the equivalent win-pace. Is there a commonly used formula or table for converting SRS to expected wins or vice versa? Is this dependent on season (for instance, mean and SD of league SRS as parameters)? I hope I'm not missing something super obvious, but I couldn't find anything after about 30 minutes of searching.

Mmkay, so here's what Ben said he's used:

Expected Wins =((SRS/2+105)^15.3)/(((SRS/2+105)^15.3)+((105-SRS/2)^15.3))*82

I copy pasted that, and didn't double check for any typos, but it looks good at first glance. He said it's just a bit of an update on Oliver's formula.

shawngoat23 · Post #5 » by **shawngoat23** » Sun Mar 3, 2019 8:01 am

Thanks for taking the time to do that!