Re: How many matches are really needed to determine final rankings?
 

Attached is a ZIP file containing partial OPRs for all 109 events in Weeks 1 through 7. Would you like fries with that?

Just kidding.

Let M be the number of qual matches at an event, and T be the number of teams.

The analysis proceeds as follows for each event:

for (k=M; k>T/2;k--){computeOPR(); deleteMostRecentMatch();}


PS - forgot to mention: I can transpose the rows and columns if that would make it easier for you to do your plotting.

Re: How many matches are really needed to determine final rankings?
 

Oooooohhhh.... preettttyyyy.... :)

Had time for just a couple plots. But these really show what I was expecting: lots of movement/changing early on and then a leveling-out. However, the leveling-out really isn't that level. It seems like there are still jumps after teams have matches later in qualifications. Perhaps because of good alliance partners, or perhaps the team had a good match.

The first chart has all teams plotted from the Silicon Valley Regional, and the second plot has just the top ten OPR teams from the North Star Regional.

Attachment 18837 Attachment 18836

Thank you for all the number crunching to get all the OPR scores, Ether. Perhaps later there will be time for more analysis with reference to rankings instead of just graphical. Or, has this analysis gone on long enough? What else would be interesting?

Re: How many matches are really needed to determine final rankings?
 

I'd like to see an analysis of how much different things would be if the old W-L-T structure was still there. Because I know that my team would not have been in the top 10 at our regionals if the old structure was still in place.

Some say this structure is harder to win because you have to out score every team, not just your opponents. But I think it really does ensure that the best teams come out on top, if at the expense of a less exciting rank-watching time during the events.

Re: How many matches are really needed to determine final rankings?
 

You're quite welcome.

The actual elapsed time to get from raw data to those reports for all 109 events (6398 partial OPR reports) was only 17 seconds on a single core of a Pentium D in an 8-year-old machine running XP Pro SP3, using AWK to wrangle the data and Octave to crunch the linear algebra.

Re: How many matches are really needed to determine final rankings?
 

Which regional were you at? Also, what should be used as the 2nd order tie-breaker? Since average score is handy, I think I would prefer to use that. If not that, are coopertition and auto points available for each match? Do you know where one would go to find them?

Nice!