Hollingers Playoff odds

[ShoogarBear]

Actually, this was basically the conclusion of my project.

Welcome to the So Clueless It's Funny Club.

It was on if draft position was a statistically significant variable when regressed against changes in win percentage. I wanted to do this to see if tanking is even an advisable option. I included other variables for comparison in my model (HOFers, All-Stars, Team Payroll, etc.) My model only explained about 45% of the change in win percentage, which is fine, it wasn't about being a perfect model, more about creating a stats project and understanding what I was doing. My conclusion/recommendations section basically said that future models would require a certain amount of trial and error regarding addition of quan ative variables, but maybe just as importantly, it would need to seek a way to measure qualitative variables like team chemistry or coaching strength.

Also I basically said that this is the reason that the statistical analysis which prevails in sports that are team sports in name only like baseball can't make it (at least at the moment) as well in sports like basketball, or football where each person's performance is largely responsible for the next person's ability to perform.

No big revelation or anything, just kinda fun to do.

This has some things in common with a project did a few years ago. I wanted to which of the principle team statistics that related to "style of play" projected to winning a championship. By "style of play" I meant things like offensive/defensive FGA, FGP, 3PP, FTM, ST, TO, Rebounds, etc. I excluded statistics like PPG and point differential under the assumption that they were secondary characteristics of the style of play.

The underlying hypothesis was Pop's assertion defensive FGP was going to be the most important characteristic for a championship team, and that things like steals and blocked shots were not going to be very important.

Anyway, I initially did a logistic regression approach to determine the statistically significant factors for "champion" versus "non-champion". After I finished, I realized that logisitic regression wasn't the correct method, because "champion" was not an independent outcome for each team (i.e., if one team is champion, by definition all the other teams cannot be). I actually posted the results here; I'll have to see if I can find that.

I had planned to relook at the data using a principal component analysis; but I haven't really done much of that before and never really got around to it. Principal components might be another way to look at what you were trying to do regarding draft position.

[Drachen]

@ Shoog: Sounds pretty interesting, in fact I read a similar (not same) project for my lit review. Anyway, maybe I will get around to posting mine. I may refine it a bit first since this project was actually due one week before I thought it was due (tomorrow vs. next thursday like I thought), basically meaning I had one week less to do it than I had originally assumed so it may be a little rushed, and it was no where near as ambitions as I originally conceptualized (but it is probably ok for a leveling course). Anyway, if I post it I am sure that the sharks will be circling (as they do here). However, it is likely that I will never get around to refining it, and therefore it never gets posted. LOL.

[ShoogarBear]

Found my post from 2007, if you're interested:

http://www.spurstalk.com/forums/showthread.php?t=61146

[sabar]

There's 5 minutes of my life I'll never get back. If the computer "plays out" the remaining season 5,000 times, it means he is randomizing something. (Otherwise, the simulation would come out the same every time.)

He's way out on a limb there. The Spurs have the best record of any team in any major sport over the last decade. They're off to the best start in franchise history. They are 20% through with the season, and already so far ahead of the record curve that they would have to have a collapse of Biblical proportions to miss the playoffs.

So Hollinger has decided that the two teams with the best records in the East and West will make the playoffs. I'm sure the Vegas bookies are trembling at this revelation.

Obviously the randomness is from the match-ups. All he does is replay each game of the season out. Each team match-up for the rest of the season favors a certain team to win by some percentage points. He predicts future wins and losses by how well teams are doing against each other right now. All it is then is tossing a biased coin for the rest of the games in the season 5000 times based on the mystery Hollinger stats and compiling the min/max/median.

Since he hides his math, we will never know if he just makes things up or actually did research on what stats correlate to what. ESPN has predictions on who will win for every game. So do a bunch of bookies. I have no idea how they are determining these percentages.

Anyone that has read Hollinger for a while will know that his stats are flawed. Last year's predictions were a pretty big joke.

Until someone makes a competing model (very difficult without access to a sports database which costs $$$) then there isn't much credibility to the whole thing.

[Drachen]

Found my post from 2007, if you're interested:

http://www.spurstalk.com/forums/showthread.php?t=61146

I will, but not tonight. LOL

[SouthTexasRancher]

Only memory I have of you this off season is your love fest for Roger Mason and how disappointed you were that he left.

[SouthTexasRancher]

Touche.

Also wrote this though: http://spurstalk.com/forums/showthread.php?t=161882

chazley, rein in that enormous false ego of yours because it is getting old to a lot of folks around these here parts. Spurs Talk has a big bunch of excellent posters....you just ain't one of them. Work on learning how to be humble and let others do the judging.