paper: Midwest Regional Strength of Schedule and RPI

Thread created automatically to discuss a document in CD-Media.

Midwest Regional Strength of Schedule and RPI
by: CJV648

Rates teams’ strength of schedules in qualifying and uses this to develop RPI to rank teams.

Strength of schedule (SOS) is based on alliance partners’ and opponents’ records. Negative SOS denotes an easy schedule, positive a hard schedule. RPI combines winning pct. and SOS.

Chicago07.xls (555 KB)

Care to outline some of the theory behind this?

It is inspired by the RPI used to rank NCAA football/basketball teams that tries to correct for the differing schedules of teams. Their approach is to use a combination of own winning pct, opponents winning pct. and opponents’ opponents’ winning pct. FRC’s alliance setup means that we should take into account the strength of alliance partners as well.

To implement this, the spreadsheet calculated a strength of match (SOM) for each team in each match. This is based on the won-loss record of alliance partners less the won loss record of alliance opponents normalized to zero for a nuetral strength match.

Technical considerations: The 1/6 in the cell arithmetic amounts to a dummy 50 pct winning term to replace the team’s own record in calculating the SOM. Appropriate subtractions are made from each alliance’s win-loss records to avoid bias based on the effect on the teams’ records of the outcome of the match being studied.

A team’s strength of schedule (SOS) is just their average SOM.

I got lazy with my RPI and just used a weighted sum of winning pct. and SOS

Ok.

Did you do all this by hand? Redoing the sheet for the West Michigan Regional is taking a while.

I’m a wiz at cut & paste, but formatting the results to fit the formulas is a pain. Give me a little bit and I’ll put one up for WMR

Seems like you ought to be able to do almost all of it in a script. (Then you could compile numbers using results from every regional!)

Last night, I did all but the last 20 matches in OpenOffice (the last 20 I put in 0’s for everything). It’s a little nicer, but I lost the sorting (since it can’t be done automatically).

This looks like a useful tool for scouting, with some work.

It was pretty interesting to see where some of the teams at MWR were compared to their actual ranking. Just proves how luck prevails.