IBM created "Deep Blue" which beat Gary Kasparov (the world chess champion) in 1997. Watson, also from IBM, dismantled Ken Jennings and Brad Rutter, the two previous Jeopardy game show grand champions. Can we utilize years worth of data on plays, player personnel, weather conditions, formations, and all other pertinent information to effectively outwit the best and brightest NFL coaches? Can science effectively outwit humans in the most popular sport in America today?
http://www.nytimes.com/2012/11/25/sports/football/more-nfl-teams-hire-statisticians-but-their-use-remains-mostly-guarded.html?pagewanted=all&_r=2&
Hi Mat,
Due to a continuous interaction between players, individual, dyadic and collective behaviors in team sports is characterized with non-linearity. This feature is grounded on players contextual dependency to the others in the neighborhood, this means that each player behavior is constrained and constrains the behavior of the others, (i.e., teammates or/and opponents). Therefore I believe that with a large history of data we can predict future behaviors but only on probabilistic basis, not with 100% accuracy.
Cheers,
Pedro.
Hi Mat!
I agree with Pedro about the effect of non-linearity.
If you wish to read what is done with data analysis in basketball, here is a reference.
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0047445
Pedro and Christophe,
Thanks for the responses, I definitely agree about the effect of non-linearity, however I believe that American football is the most linear, or scripted and chess-like, of all sports (see attachment below).
While basketball, football/soccer, baseball, and rugby are mostly fluid and improvisational, teams in American football depend heavily on their play structure to execute all plays efficiently.
Question for you guys...
What discipline would be the most effective for trying to accurately predict future play calls based on compiled data? Would it be regression analysis...something else?
Thanks again for the replies...
Mat Ball
http://www.google.com/imgres?imgurl=http://www.compusports.com/images/fbsample.gif&imgrefurl=http://www.compusports.com/playbook.asp&h=410&w=541&sz=7&tbnid=E7ahvncd_Iw4mM:&tbnh=90&tbnw=119&prev=/search%3Fq%3Dexample%2Bfootball%2Bplay%26tbm%3Disch%26tbo%3Du&zoom=1&q=example+football+play&usg=__SBw_OdDgilrKhhD8xaTpAKT3QII=&docid=taCEFRZEw6jlWM&sa=X&ei=8MuoUZ2xMaOziwLEtIGYAQ&ved=0CC8Q9QEwAQ&dur=1179
Mat,
I agree about the chess-like nature of American football, but with conscious pieces!
Your question about the analysis is a big question! I am not a statistician but I would say that Bayesian approaches and agent-based modelling with rules derived from the real player stats would be two possibilities.
Christophe
Thanks Christophe,
I'm beginning my hunt, and I'll dive deeper into those two disciplines, although I will most definitely need to solicit the help of some professors from the decision sciences sector (I'm just an MBA student and a novice with statistics).
Appreciate the advice.
Best regards,
Mat
well, if am right you meant a coachs' success right? I may not differ from my colleagues, when they all gree that huge datasets could give some prediction in the coachs' behavour, but all this said I want all of you to note this "if we could be able to predict say exchange rates then wealth would be for all" am saying this because we have lots of available intra-day data for excahnge rates, but why are losses recorded on money markets?
In other words, irrespective of the data available on the coach's history all statistics computed from it would not be trusted without a grain of salt. A real life example is Scholari who failed at Chelsea FC yet he had been very successful with other teams, say.
So use the statistical measures with alot of caution, though its wise to have an insight based on empirical methods than mere guess work. Good luck!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics