DCI 2015 Predictions Using Machine Learning

[MikeRapp]

Thanks for this.

I also thought the scoring for Finals night was a bigger jump than normal. Considering that scores have been general low overall this year, I think DCI may have felt pressure to bring scores back up. Nobody wants to see someone win with a 96 or less.

Maybe I'm getting off topic now, but...

I think DCI had the judges depress the scoring for this year to raise the bar for 99pts, because of what BD did last Finals. They did need to raise that thresh-hold. However, I do think the winning corps this season was worthy of a 97.65 but no more. Some may say "raw numbers are meaningless, it's the spreads that matter", but according to the judging sheet criteriae, a 98+ is not the same as a 97, and I do think the top 4 shows this season general deserved the raw scores they got, and can be compared to the scores they got last season.

BD this season were pretty much 2pts behind themselves last year. Same for everyone else.

I'm not sure how you would put this kind of info into your program, or if this is even relevant.

What I would REALLY like to see is the predictions for scores based on certain judges. What would Finals scores have been, say, if the same judging panel from Thursday or Friday judged Saturday also? What would the computer predict then?

So, question:

Were this the year Bluecoats unveiled Tilt, do they win the gold?

[BoyWonder1911]

So, question:

Were this the year Bluecoats unveiled Tilt, do they win the gold?

They might have.

And to the "can't have it both ways" reply, what I was trying to get at was that the judges were depressing scores on purpose all season, and at the end they realized that there was no reason to depress them because the shows were about 2 pts less quality than last year, and decided that a corps can't win with a 96 so they have to artificially raise them towards and at the end. Maybe they figured BD would get another 99.65.

I thought the 97.65 was accurate.

Felliniesque would have scored above 99 pts again if it were done this season, even against Ink in direct competition. If you had taken the 2014 BD and brought them through a time machine one year into the future etc etc.

Anyways the point is that I wonder how that would be accounted for in a prediction program. And I'd like to see predictions based on some individual judges' scores.