APBRmetrics

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

I really need your thoughts to improve my website. I'm making daily score predictions; calculating pace, ranking the teams' efficiencies, player PER's, player match-ups for every single game.

My goal is to produce professional and more visual stats for those who aren't familiar with possession based stats.

www.nbastuffer.com

Thanks,
Serhat

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Thanks a lot who send me their insight via my web site. I have added a new graphical power ranking graph indicating teams' offensive and defensive efficiencies based on last 5 games. So we can see how the teams are performing recently and in the season both.

John Beattie · Joined: 26 Jan 2007 Posts: 44 Location: NYC

Make sure to check your private messages here on the forum. Three days ago I sent you one but perhaps you're not aware of this feature. Meanwhile, just now I tried to send you a message via your website but it froze after I clicked submit.

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Dear John,

Thanks in advance for your comments. I'm having the stats and all that stuff calculated in excel spreadsheets and visuality is limited to excel's visuality features but make sure that I'm working on it. I dream a day that I have some abilities in php or asp programming which will help the web site become more professional and more interactive. I will add more functions such as monthly written articles on NBA statistical topics because I have already had them in my excel workbook as charts and graphs. Also there will be some nice player analysis.

Have a nice day,
Serhat

Mountain · Joined: 13 Mar 2007 Posts: 437

Serhat (hoopseng), I would be interested in hearing more about your methodology and results to date and any adjustments you have made and whether you get better with time. I see the prediction results to date and it appears that the start is modest and finetunng would certainly be sought.

I am not sure if I am on target here but did you use a Kalman filter approach?

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Mountain,

Thanks for the thoughts. My methodology is based on formulas which calculates teams' outings when they have similar situation before.

For example;

if Phoenix is on the road (1)
and playing a 3rd game IN 4 nights game (2)
and in a 4 game winning streak (3)
lineups preferred (4)

same 4 factors for the opponents and league averages also taken into account.

I applied Kalman filter last season and didn't get better results. My model needs to have injury related data more adjusted. So, gametime decisions and questionables, doubtfuls are my number 1 nemesis. If a player is out and it's officially announced I get better results. One thing my model do better is road teams' score predictions are way better than the home teams'.

Briefly, making predictions is tough but fun!

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

I have already changed the season's and last 5 games' efficiency charts. Best teams are located upper right and the worst teams are in lower left section of the graph.

A few people asked me to add "quarters and halfs stats", so it's in progress...

Mountain · Joined: 13 Mar 2007 Posts: 437

The team efficiency charts are helpful to see. Thanks.

So I am not sure if I understand the prediction adjustment fully yet. Do you make a ratings adjustment to what the offensive and defensive efficencies would indicate based on just games when all 4 of your factors are true or do you calculate the general affect of each anytime true and then weight? Finding games where all 4 are true makes for real small samples (just using the current season I assume, so especially small early in season) but captures interactions. Tradeoffs. Or use a hybrid of both the indicators of when the 4 factors are all present and each separately?

Efficiencies by quarters and halves would be very interesting to see. Thanks for working on that and let us know when it is available.

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Mountain,

Yes, I agree, it's sometimes being a small sample, but then I look up for last season stats and current seasons' averages belonging to teams that are in same or similar situations. Unfortunately I only have a 2 year data set and think that, hopefully, after 3 or 4 seasons it will work better. At most of times sample size don't bother my model to work properly. Also, that's why I've started my predictions after a month NBA season has begun.

Today I wrote a new article as my very first basketball article in my life!.

I would like to hear from all of you about what it looked like.

"How fair is NBA schedule?"

http://www.nbastuffer.com/content/blogsection/14/64/

Thanks again for visits and insights.

Mountain · Joined: 13 Mar 2007 Posts: 437

Thanks for the further explanation. Have you compared your predictions to betting lines and seen any trends? I know the 2 aren't exactly the same thing but might be a useful study if you haven't already. Another possibility would be to compile the over / under pattern for each team and maybe try slightly different formulas for the over and under scoring to prediction groups? That may or may not work as teams bounce around but it is a theory that could be tested, especially as your database grows and you see if the departure patterns have much consistency.

Interesting to see the schedule breakout. I imagine folks will check out the treatment of their team and maybe chief rivals.

I had to realize that B2B column was just back to back games that didnt qualify as part of a 3in 4 onr 4in5 but after that I understood.

One thing I checked using your presentation was games played with 1-2 days rest (not part of a 3in 4 or 4in 5). These could be called the normal or even ideal situation. Only 8 teams varied from the mean of near 43 by more than 3. That part isnt bad.

I recall they have a consultant do the schedule. Not sure what criteria they emphasize in doing it beyond getting it done. I'd assume that they'd try to maximize revenue first above all else with fairness a secondary concern. Some favoritism may be involved but you'd need mutli-year data. Geography and amount of air travel time surely affects it some with some cities harder to get to or get from in timely fashion and affecting scheduling. Trying to minimize exposure to fog and snow plane issues might be involved some too to the extent they can without unbalancing things too much.

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Mountain · Joined: 13 Mar 2007 Posts: 437

I'd be particularly interested in what general trends you find related to pace played vs average pace of the 2 teams involved (and how you are predicting that element of the score) and team performance under different actual paces.

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Have fun with my new article...

A break-down analysis about at what positions the teams have strengths and weaknesses.

"Strengths and Weaknesses of NBA Teams"

http://www.nbastuffer.com/content/view/36/64/

Harold Almonte · Joined: 04 Aug 2006 Posts: 412

I'm wondering why Denver doesn't appear in the team strenghts at Center pos. We can't say Camby is overrated at the defensive end. Is he making himself worse, trying to make his SFs better at the defensive end? We can't deny a lot of players are assisted a lot by their "point-defenders" (centers), and most of their deff. strenght would need to share credits with them.

hoopseng · Joined: 13 Oct 2006 Posts: 54 Location: Basketball Research

Only the players matched up in the game are taken into account. Of course, Camby man is good defender, intimidator and he can defend anyone regardless of position.

The chances are; Denver team love to play uptempo, his opponents may be averaging much more points than season averages when playing against Denver. What I have to consider is, adjusting my numbers for the team's pace.