Predicting Games

[marko123]

I have been following this forum for a while now and have learned a ton of metrics based on an individual player level as well as a team level. However, it seems as most metrics cannot predict future performances, only allow me to evaluate PAST performances. I am trying to find a way to predict matchups between different teams but I have no clue what statistics and metrics are useful for forecasting performances and scores. I know people here post projected player stats but it is hard to sort of put that into a model to see which team is more probable to win. If anyone has any suggestions or comments feel free to post.

[J.E.]

Pretty much every metric can be converted to forecast game point differential (or win%).

There are a couple that don't need much/any conversion like (R)APM, ORtg+DRtg (bbref), ASPM, eZpm(?), and those that need it like eWins and WP.

The most recent thread on forecasting/retrodicting is here http://apbr.org/metrics/viewtopic.php?f ... 4&start=15

[bbstats]

A good start:

1) Find the percent of game-minutes each player plays (or estimate/predict). An OK rule-of-thumb is (MinutesPerGame/48)*(Games/66) unless the player is a heavy rotation/starter that was injured, in which case the (Games/66) should = 1 if they are healthy now.

2) Multiply this percentage by their RAPM (which is unfortunately not updated - http://stats-for-the-nba.appspot.com/ranking_rec

Sum of (Team A RAPM*Minutes%) minus (Team B RAPM*Minutes%) is a rough estimate for the team's predicted efficiency differential. If you want point differential, you'll have to predict pace, which is more of an NCAA thing to do

[J.E.]

2) Multiply this percentage by their RAPM (which is unfortunately not updated - http://stats-for-the-nba.appspot.com/ranking_rec

bbv is not updating their matchupfiles. No idea if it will stay that way forever, but I guess not.
Should Aaron never update again, I can fall back to crawling ESPN

[bbstats]

Ah. Huge bummer!

[Crow]

What is helpful in these references?

http://www.cs.toronto.edu/~gdahl/papers/dpmfNBA.pdf

http://t.co/ZZSMEbPrYK

[Crow]

106 pages http://arrow.dit.ie/cgi/viewcontent.cgi ... cschcomdis

[Crow]

http://www.sporttechie.com/2014/04/07/m ... e-madness/ Kaggle competition.

[Crow]

some discussion of play level analysis
http://dspace.mit.edu/handle/1721.1/85464
if anyone buys, maybe summarize impressions?

[Crow]

http://www.vikparuchuri.com/blog/predic ... ls-with-r/

[Crow]

the importance of attributes
http://arxiv.org/pdf/1310.3607.pdf

[Crow]

will this open for u? useful?
http://www.academia.edu/5288042/A_Revie ... _in_Sports

[Crow]

http://www.cgee.org.br/atividades/redirKori/7947

[Crow]

http://www2.isye.gatech.edu/~jsokol/ncaa.pdf NCAA but good discussion of issues

[Crow]

using per possession info
http://dukespace.lib.duke.edu/dspace/bi ... sequence=1