Attempting to model NBA over/unders

[ManBearPig]

Yup. Definitely cut down on the number of plays... As I tracked my results I noticed that when I was under 2 pts in my predictions it was just flipping a coin there. IE losing whatever edge I get from modeling. For instances todays CHI@MEM game I come up with between 187-189... line is 187.5 @ JoAnna... I'll give you the SAC@ATL game as an under BOL to you guys

This is exactly what I noticed as well and why I put in a filter for that lower bound - 3pts seems to be a good number here from what I've seen. I also noticed that a number margin too large can be a bad thing too and account for it but I haven't settled on anything yet.

[Buried_PIRATE]

Talk about an under day... looks that damn MLK curse is true!

Either way, I wonder if CHAR@PHI can even cover the over with OT!!!

[ManBearPig]

Talk about an under day... looks that damn MLK curse is true! Either way, I wonder if CHAR@PHI can even cover the over with OT!!!

Dam, Nope not even OT could help us...188 and needed 189.5+...bummer. How did I have this going over.

[CNega]

Thank you! 3/4 today with your picks. Keep up the great work1

[Buried_PIRATE]

TOR@NO / Line: 193 **Playing Over**
5Day: 204
YTD: 197
GRADE: LOSS

MIL@HOU / Line: 193 **Playing Over**
5Day: 205
YTD: 195
GRADE: LOSS

SAC@ATL / Line: 201 **Playing Under**
5Day: 207
YTD: 196
GRADE: WIN

CHAR@PHI / Line: 189 **Playing Over**
5Day: 200
YTD: 193
GRADE: LOSS

NJN@GS / Line: 205 [UNDER though very close to a no play... will be a game time decision based on line]
5Day: 211
YTD: 200
GRADE: LOSS

Curse of MLK... all of my overs went under 1-4 on the day. Will keep on grinding the two games tomorrow at prob around 6 PMish..

53-41 on the year

[CNega]

wow I totally read your initial 1/17 totals wrong. I ended up betting under for Phila/Cha and ended up choosing a totally different game (Utah/Washington) and lost on your Houston/Milwaukee game, so I ended up with 2 extra wins instead of 2 more losses. Purely luck on my part.

[Buried_PIRATE]

wow I totally read your initial 1/17 totals wrong. I ended up betting under for Phila/Cha and ended up choosing a totally different game (Utah/Washington) and lost on your Houston/Milwaukee game, so I ended up with 2 extra wins instead of 2 more losses. Purely luck on my part.

Haha ... I saw your post and was like damn you faded me!

I actually thought you might have gotten a better line on CHAR game, which we lost by 1 pt and my initial lean on one of the other games I changed at the last minute but went from right to wrong... dammit.

Either way, glad it worked out for you and I'm looking forward to getting back in the black

[Dark Horse]

You hit the ground running at 21-11 (page 2, where you and I both felt a correction was in the works), and are now 53-41. That's 32-30 since the opening salvo, suggesting model may already have leveled out. Hopefully not, but possible.

[Buried_PIRATE]

You hit the ground running at 21-11 (page 2, where you and I both felt a correction was in the works), and are now 53-41. That's 32-30
since the opening salvo, suggesting model may already have leveled out. Hopefully not, but possible.

Sadly I agree. The following day after 21-11, I hit a 1-5 day which allowed me to realize I needed filters (most of those games had point totals above 209 which almost always come up UNDER). Following yesterday after the slightly unusual afternoon games I came up 1-4. If those days are anomalies I am on a respectable 30-21.

However, I won't make excuses for bad days. I will keep running this model (with slight corrections).

I am also working on new ideas, but sadly I've been too busy to really get into it. Right now I do have a lot of data to back test (over 90 games) and I wonder if I could back test my data to fit a model that does better than where I am at (57ish %)

Either way I hope u keep contributing, keeps me thinking. Sadly I need to follow bball much much closer... I'm a baseball guy

EDIT: For instance I am not even looking at btb games yet, which obviously has SOME effect. Other ideas are also always appreciated

[Buried_PIRATE]

CHAR@CHI / Line: 183 **Playing Over**
YTD: 187

Trying a new model where I predict the score for each team...

Came up with:
Bulls: 102
Char: 86

That total comes out to 188... will be rolling with that tonight

BOL

[ManBearPig]

The best thing would be to get a database going back 5 yrs or so and start using that to find some correlations do some regressions and back test. These typically cost money but they are worth it and should pay for themselves in the long run. I have some sites saved off at home but a quick google search will yield the same results probably too. The benefit of having everything in a database to query what your heart desires is very powerful. Also I think if you had some more data to run through, even if it is past data, you could still get a better feel.

I would suspect each season isn't always the same so what works this season may not work next or did not work last. Getting a better understanding of these things is something I want to spend some more time on because there could be value there. I don't want to get so wrapped up on the short term results that I lose the ability to adapt long term.

However, you can't travel a road that isn't built so I'm looking at this season as building a foundation and then working on improving it from there. As I understand things better and get more and more defined parameters I'll be able to make better improvements.

[Buried_PIRATE]

CHAR@CHI / Line: 183 **Playing Over**
YTD: 187
GRADE: Loss

Updated

[Buried_PIRATE]

Been running bad lately 1 Win in past 6... signs of fading... I'm not sure. Plays I'm on tonight:

UTA@NJN / Line: 192
Model: 177 / Under

PHI@ORL / Line: 200
Model: 195 / Under

DET@BOS / Line: 191
Model: 197 / Over

MEM@NO / Line: 184.5
Model: 189 / Over

WAS@MIL / Line: 189
Model: 182 / Under

BOL on your plays

53-42 from 12/28

[sickpicks]

Been running bad lately 1 Win in past 6... signs of fading... I'm not sure. Plays I'm on tonight:

UTA@NJN / Line: 192
Model: 177 / Under

PHI@ORL / Line: 200
Model: 195 / Under

DET@BOS / Line: 191
Model: 197 / Over

MEM@NO / Line: 184.5
Model: 189 / Over

WAS@MIL / Line: 189
Model: 182 / Under

BOL on your plays

53-42 from 12/28

I like three plays and good luck with others two plays

[Buried_PIRATE]

Not chasing... but I also took:

POR@SAC / Line: 195.5
Model: 191 / Under

[ManBearPig]

I think you have something that will win in the long-term as long as you stay with it and constantly...I did have a question about the regression modeling. I don't develop my lines through regression testing but I'm playing around with it.

I just have a question about the dependent variable and what values your using. I've ran it using the set line, but don't know if that's the best way to do it. I've back tested it for a couple days and it's not very good for predicting a line but it is close to the set number. I'm pretty sure I'm using it incorrectly so just wondering if you had any suggestions there.

[madmiek]

good job.

[Buried_PIRATE]

I like three plays and good luck with others two plays

Wow tough past couple days. Dammit.

WAS@MIL was a tough loss going down by 1...

UTA@NJN was going to hit but then we got crushed with fouls in garbage time... either way... day could end okay we'll see :/

[Buried_PIRATE]

I think you have something that will win in the long-term as long as you stay with it and constantly...I did have a question about the regression modeling. I don't develop my lines through regression testing but I'm playing around with it.

I just have a question about the dependent variable and what values your using. I've ran it using the set line, but don't know if that's the best way to do it. I've back tested it for a couple days and it's not very good for predicting a line but it is close to the set number. I'm pretty sure I'm using it incorrectly so just wondering if you had any suggestions there.

I created a variable, total team points, which I calculated as the average game score for each team. I then regress that versus other variables of interest.

If you come up with a formula that you use lmk and we can discuss.

[ManBearPig]

Thanks...I generate quite a few numbers based off different stats so I'll play around with it and see what I come up with.

Also, if you want a little more accuracy or something else to play around with may I suggest calculating each teams PF/PA at home and on the road. Those numbers are very helpful to have along with the average total as well, I actually don't use the average total but probably should.

[Buried_PIRATE]

We go 3-3 last night, will post the grades a bit later. 2 of the wins I pegged as unders also went into OT and held up... so maybe a sign of positive things to come.

Either way, total record goes to 56-45... not quite where I want to be, but still positive which is always good

[ready-set-tilt]

This is a great thread, I am a long-time lurker who felt obligated to post a thank you for putting in all the hard work. It seems like the model has slowed down a bit lately, but it may just be a cold streak. As long as you keep improving it, you should be able to stay positive for the long term. I think the idea of inserting a variable for referees may be a way to gain a hidden advantage. Good luck!

[ManBearPig]

So I've been playing around with regression in my model and I think I've found a correlation that has relevance, well actually two I just can't decide which to use but they are close so I don't think it matters.

Looking at tonight I've calculated. These are not just the regression numbers but does include them. Looks like POR/LAC is the best play so of course it will go way under. I don't want to hi-jack your thread but just thought I'd share what I came up with tonight. I will probably be incorporating a B2B and a 5/10 day number to try and incorporate as well, but I want to give this a run first.

1/20/2011 Charlotte/Philadelphia
Predicted: 186.78

1/20/2011 Chicago/Dallas
Predicted: 180.55

1/20/2011 Portland/L A Clippers
Predicted: 203.44

[Buried_PIRATE]

Hey, definitely not hijacking.

I forgot I won't be home till after 8 tonight, so I will have a prediction for the late game but not the first 2. BOL on the plays

[Buried_PIRATE]

POR@LAC / Line: 194.5
Model: 197 / Over

56-45 YTD

[joanapoker]

nice call yesterday!

[Buried_PIRATE]

POR@LAC / Line: 194.5
Model: 197 / Over
Total: 201
Grade: WIN

57-45

nice call yesterday!

Haha after the slump I was in, I will enjoy that winner!

More games to cap today

BOL to you guys

[Buried_PIRATE]

POR@SAC / Line: 195.5
Model: 191 / Under
GRADE: Win

UTA@NJN / Line: 192
Model: 177 / Under
GRADE: Loss

PHI@ORL / Line: 200
Model: 195 / Under
GRADE: Win

DET@BOS / Line: 191
Model: 197 / Over
GRADE: Loss

MEM@NO / Line: 184.5
Model: 189 / Over
GRADE: Win

WAS@MIL / Line: 189
Model: 182 / Under
GRADE: Loss

Updated the play grades from 2 nights ago

Record is 57-45

[Buried_PIRATE]

Plays for tonight boys, hot off the presses. Record going in is 57-45

DET@NJN / Line: 190.5
Model: 186 / Under

MIL@CLE / Line: 189
Model: 182 / Under

NO@ATL / Line: 184.5
Model: 193 / Over

UTA@BOS / Line: 194
Model: 200 / Over

HOU@MEM / Line: 205.5
Model: 203 / Under

[joanapoker]

Also playing the Under at Cleveland

I'll probably tail that Under at Memphis!

Gl tonight pirate

[sickpicks]

i like this one

DET@NJN / Line: 190.5
Model: 186 / Under

[Buried_PIRATE]

59-48 going into today

BOS@WAS / Line: 191
Model: 195.6 / Over

ATL@CHAR / Line: 185
Model: 190.5 / Over

CLE@CHI / Line: 191.5
Model: 196.7 / Over

SA@NO / Line: 185.5
Model: 194.8 / Over

[joanapoker]

I'm also on the Over at Wash.

Careful with Overs at Chicago! Powerful defense!

[chino_09]

BOS@WAS / Line: 191
Model: 195.6 / Over


Loving this pick. 7 out of last 9 games for Boston have gone Over. Washington's Nick Young has been red hot too. Great matchups in this one: Rondo/Wall, Blatche/KG.....

BOL to us!!

Not so sure about the Over for the CLE/CHI game, Pirate. But gl to u regardless!!

[Buried_PIRATE]

I'm also on the Over at Wash. Careful with Overs at Chicago! Powerful defense!

BOS@WAS / Line: 191 Model: 195.6 / Over Loving this pick. 7 out of last 9 games for Boston have gone Over. Washington's Nick Young has been red hot too. Great matchups in this one: Rondo/Wall, Blatche/KG..... BOL to us!! Not so sure about the Over for the CLE/CHI game, Pirate. But gl to u regardless!!

BOL to us