Great job Joe.
In terms of accuracy, I think somebody on CD did a study this year and found that using highest OPR is a little more accurate than other ways this year. I put in the predictions for championship divisions in the spreadsheet that I published using highest OPR.
Team 2834 2011_Scouting_Database Championship v4
http://www.chiefdelphi.com/media/papers/2174
Since JB987 asked for it, I put in a few formula in that spreadsheet and this is what I got for Championship by division during qualifying rounds in terms of percent of time it predicted a win.
Archimedes 73.4%
Curie 74.1%
Galileo 76.9%
Newton 81.0%
Is it perfect? No. Close matches that could have gone either way, mistakes made by teams and better strategy by underdog alliance explained why it was not perfect. What about the actual scores? Some were amazingly close. Knowing the prediction ahead of time that you have an uphill battle allows you to change your strategy and hopefully come out on top.
For IRI, the accuracy may be lower. These are all top notch teams and everyone can score with minibot. However only two minibots can score so that explains why the predicted OPR exceeds the maximum allowed. Also there are only so many tubes to hang unlike the soccer balls last year that can be recycled. Does it matter? No, if an alliance has a higher combined predicted OPR, it still has a higher chance of winning. The extra minibot may be just an insurance if one partner cannot get their's up.
Since this is offseason, there are a lot of other factors. New drivers, upgrade to robots, etc. that will reduce the accuracy.
The predicted score also does not take into account human error: referee that threw a tube right betweeen the tower and your partner's robot when they were about to go to the tower to score their minibot, the extra time to get around the tube costed us a tiebreak match and was eliminated in quarterfinal. These things happen and I am okay with it. It helps to explain why predictions are not always correct.