Attempting to model NBA over/unders

monkeyking · 12-28-10, 06:10 PM

Good luck...I would suggest as a start to not bet every game, especially those that you think is only 1-2 points off. Not only could your numbers be wrong, but there is a lot of variance in a typical NBA game.

Buried_PIRATE · 12-28-10, 09:30 PM

Originally posted by Buried_PIRATE

Either way, predictions for tonight:

Orl @ Cle
Prediction: 192.25
Line: 191
Play: Over
Grade: Win

Bos @ IND
Prediction: 186.7181
Line: 190
Play: Under
Grade: Win

NYK @ MIA
Prediction: 197.9417
Line: 204.5
Play: Under

MIL @ CHI
Prediction: 183.0715
Line: 178.5
Play: Over

TOR @ DAL
Prediction: 200.1264
Line: 208
Play: Under

LAL @ SAS
Prediction: 200.7372
Line: 199
Play: Over

POR @ DEN
Prediction: 202.4372
Line: 204
Play: Under

Possibly some beginners luck, start the night 2-0, will update the rest later

Buried_PIRATE · 12-28-10, 09:32 PM

Orl @ Cle
Prediction: 192.25
Line: 191
Play: Over
Grade: Win

Bos @ IND
Prediction: 186.7181
Line: 190
Play: Under
Grade: Win

Some nice beginners luck, start the night 2-0

Buried_PIRATE · 12-28-10, 10:13 PM

Hmmm damn can't edit my posts. And a double post

Either way looks like it'll be a 4-1 start. I'll just try to keep it down to one post till I figure out how to edit my previous.

Buried_PIRATE · 12-28-10, 11:34 PM

Originally posted by Buried_PIRATE

Orl @ Cle
Play: Over
Grade: Win

Bos @ IND
Play: Under
Grade: Win

NYK @ MIA
Play: Under
Grade: Win

MIL @ CHI
Play: Over
Grade: Loss

TOR @ DAL
Play: Under
Grade: Win

LAL @ SAS
Play: Over
Grade: Loss

POR @ DEN
Play: Under
Grade: Win

5-2 not bad for a first day.

Will post my predictions for tomorrow soon

Inspirited · 12-29-10, 12:06 AM

nice first day there!

LiarGame · 12-29-10, 12:20 AM

Congrats on a winning day!

OnlyMayo · 12-29-10, 12:38 AM

Share the system?

yurnero · 12-29-10, 12:43 AM

nice first day, what kinda system / numbers are you using?

Buried_PIRATE · 12-29-10, 11:38 AM

Model went 5-2 for day 1. Today is a big slate with 11 games so we'll see how it holds up.

My model is based off of offensive rating (points per 100 possessions), defensive rating (points allowed per 100), a variable for home court advantage, and a variable for tempo. I've been using YTD stats but make minor hypothetical adjustments in new developments (trades/injuries). From these variables I weight them based on importance (found from a basic multi-variate regression) and plug in the team input for each game and match up.

Slate for today:

GWS @ ATL / Line: 202.5 / Prediction: 199.23 / Play: Over
IND @ WAS / Line: 196.5 / Prediction: 189.9 / Play: Under
CLE @ CHA / Line: 200 / Prediction: 190.9 / Play: Under
BOS @ DET / Line: 186.5 / Prediction: 193.6 / Play: Over
LAL @ NOH / Line: 190.5 / Prediction: 192.95 / Play: Over
DEN @ MIN / Line: 221 / Prediction: 204 / Play: Under
NJN @ OKC / Line: 197 / Prediction: 196.3 / Play: Under
MIA @ HOU / Line: 198 / Prediction: 196.97 / Play: Under
PHI @ PHX / Line: 207 / Prediction: 200 / Play: Under
UTA @ LAC / Line: 194.5 / Prediction: 197 / Play: Over
MEM @ SAC / Line: 195 / Prediction: 193.18 / Play: Under

One thing to notice, is some games the line and the prediction line up almost exactly +/- 1, I am following these results "strictly" to test the model. However, knowing the variance in the NBA and as I get more workable results, hopefully I can hone in on a more safer zone of which bets to play and which to pass.

Either way, let's see how DAY 2 goes.

As a disclaimer: I am just a kid that enjoys statistics and this challenge. I am just a small time bettor and this is just plain fun for me. Also if you have any input into my methodology lmk!!

Buried_PIRATE · 12-29-10, 12:28 PM

Argh can't edit posts still... GWS @ ATL should have the play as under. Either way, looking forward to the games

OnlyMayo · 12-29-10, 12:33 PM

GL...

LatinPimp · 12-29-10, 12:40 PM

Great first day... But remember for every great day there is a Bad day right around the corner, I'll be following along and see how the formula does for ya...

BOL

OnlyMayo · 12-29-10, 10:51 PM

Looking pretty descent still...GL

dlew2k · 12-29-10, 11:37 PM

I think you made a mistake on your first game..
GWS @ ATL / Line: 202.5 / Prediction: 199.23 / Play: Over
shouldn't the play be UNDER ??

oops.. Just saw your post. Sorry.

(Just my opinion) You may want to adjust your numbers based on Home , road, last 7 games. Or even consider a head to head ranking for each game played versus your numbers.

Buried_PIRATE · 12-30-10, 10:10 AM

Went out and got smashed last night, ill update my rec and add my numbers today when my head stops killing me

I am definitely trying to improve the model though and I think ranking which games are the best to play is the next step

Buried_PIRATE · 12-30-10, 11:28 AM

Originally posted by Buried_PIRATE

GWS @ ATL / Line: 202.5 / Prediction: 199.23 / Play: Under (corrected in post below)
TOTAL: 196
GRADE: WIN
IND @ WAS / Line: 196.5 / Prediction: 189.9 / Play: Under
TOTAL: 194
GRADE: WIN
CLE @ CHA / Line: 200 / Prediction: 190.9 / Play: Under
TOTAL: 193
GRADE: WIN
BOS @ DET / Line: 186.5 / Prediction: 193.6 / Play: Over
TOTAL: 196
GRADE: WIN
LAL @ NOH / Line: 190.5 / Prediction: 192.95 / Play: Over
TOTAL: 191
GRADE: WIN
DEN @ MIN / Line: 221 / Prediction: 204 / Play: Under
TOTAL: 233
GRADE: LOSS
NJN @ OKC / Line: 197 / Prediction: 196.3 / Play: Under
TOTAL: 197
GRADE: LOSS (predicted 196.3 total ~ 197 lol!)
MIA @ HOU / Line: 198 / Prediction: 196.97 / Play: Under
TOTAL: 245
GRADE: LOSS (way off the mark... yikes)
PHI @ PHX / Line: 207 / Prediction: 200 / Play: Under
TOTAL: 233
GRADE: LOSS
UTA @ LAC / Line: 194.5 / Prediction: 197 / Play: Over
TOTAL: 198
GRADE: WIN
MEM @ SAC / Line: 195 / Prediction: 193.18 / Play: Under
TOTAL: 198
GRADE: LOSS

Day 2: 6-5

Giving the model a total record of 11-7 over 2 days playing EVERY NBA game "strictly." I will keep posting this exact model in the thread to continue to grow the sample size and in the future start to try to rank games that I think are better plays based on the model.

Either way, I will be back to post the slate for these 3 games a bit later.

Buried_PIRATE · 12-30-10, 01:48 PM

Continuing with my little project. 11-7 so far.

Small slate today, Dirk not playing in the Mavs game... came up with:

NYK @ ORL / Line: 213 / Prediction: 198.17 / Play: Under
SAS @ DAL / Line: 191.5 / Prediction: 195.8 / Play: Over
UTA @ POR / Line: 188 / Prediction: 196 / Play: Over

monkeyking · 12-30-10, 02:58 PM

It seems your numbers always point towards the average (190-200), but theres a reason vegas sets these numbers at 180 or 220 sometimes. Does your model ever predict a 160 or 240 type of game?

Buried_PIRATE · 12-30-10, 05:51 PM

Originally posted by monkeyking

It seems your numbers always point towards the average (190-200), but theres a reason vegas sets these numbers at 180 or 220 sometimes. Does your model ever predict a 160 or 240 type of game?

Not really... I'm using averages and season data. The variance is going to be pretty low. The bounds from typical NBA teams I've found from messing around has been around 170-210. I quickly plugged in two high scoring teams and I came out > 210 when the Knicks would play the Suns... so yeah, sometimes.

My model is lacking a lot of things that Vegas probably uses in their lines, I'm missing back to back day information, the latest injuries, other recent events etc. But at the end of the day, when I did my regression based approach I thought I honed in on some key variables. I'm going to stick with this model in this thread for awhile, but I am working on other more complex models. The problem is I have trouble finding past betting lines so I have to stay updated with this model to build my sample size.

Either way, we'll see how it goes for some more slates of games.

ManBearPig · 12-30-10, 06:03 PM

I've been working on something as well so I'll be interested in how you do. I don't use much statistics but do crunch some numbers. Once thing you'll notice is the sweet spot is around 199. If you take the avg of the lines on almost any night they will be right around that mark...never fails.

Another thing is I'm not as interested in trying to predict a number above or below the current line because there are some numbers that you'll find harder to achieve. i.e the DEN/MIN game last night. It's hard to predict a score total of 240 and based on the sweet spot number you probably never will. The way around it is to try and develop a new separate number (obviously using current line) to try to predict the pace that will determine if it goes O/U that number. Easier said than done, but I've had moderate success.

I've been tweaking and re-tweaking as I gain more data so we'll see how it goes. If I can hit between 55%-60% I'll be doing pretty well even though it doesn't sound like it. Since you're a statistics guy I'd be curious what stats you find are most useful because I'm looking at similar things.

Buried_PIRATE · 12-31-10, 09:11 AM

2 for 3 last night... 13-8 thus far. Ill be out of town the next two days so I doubt I will have time to make picks.

Buried_PIRATE · 01-02-11, 04:35 PM

Originally posted by Buried_PIRATE

NYK @ ORL / Line: 213 / Prediction: 198.17 / Play: Under
Total: 215
Grade: LOSS
SAS @ DAL / Line: 191.5 / Prediction: 195.8 / Play: Over
Total: 192
Grade: WIN
UTA @ POR / Line: 188 / Prediction: 196 / Play: Over
TOTAL: 197
GRADE: WIN

Will update more plays for tomorrow

Buried_PIRATE · 01-02-11, 05:28 PM

Yikes. Some pretty close lines tonite

BOS @ TOR / LINE: 197.5 / Prediction: 195.31 / Play: Under
DAL @ CLE / LINE: 188.5 / Prediction: 192.2 / Play: Over
HOU @ POR / LINE: 200.5 / Prediction: 201.3 / Play: Over
PHX @ SAC / LINE: 208.5 / Prediction: 202.7 / Play: Under
MEM @ LAL / LINE: 200.5 / Prediction: 197 / Play: Under

ManBearPig · 01-02-11, 05:55 PM

If your lines are too close to the actual line you may want to incorporate some point overlay. A game that you have predicted to be within a point is still a toss-up. You may want to use like a 3pt overlay and see if that improves your numbers. Line movement alone could alter the play from an over to and under or vice-versa...just a though. GL though.

Buried_PIRATE · 01-03-11, 12:01 AM

Originally posted by Buried_PIRATE

BOS @ TOR / LINE: 197.5 / Prediction: 195.31 / Play: Under
TOTAL: 172
GRADE: WIN
DAL @ CLE / LINE: 188.5 / Prediction: 192.2 / Play: Over
TOTAL: 199
GRADE: WIN
HOU @ POR / LINE: 200.5 / Prediction: 201.3 / Play: Over
TOTAL: 185
GRADE: LOSS
PHX @ SAC / LINE: 208.5 / Prediction: 202.7 / Play: Under
TOTAL: 183
GRADE: WIN
MEM @ LAL / LINE: 200.5 / Prediction: 197 / Play: Under
TOTAL: 189
GRADE: WIN

Solid day... probably still beginners luck. Right now my model is 17-9

I will keep posting my predictions in here, tweaking the model slightly.

Inspirited · 01-03-11, 12:13 AM

Nice going there.

thedon88 · 01-03-11, 12:14 AM

nice

Buried_PIRATE · 01-03-11, 06:27 PM

Hopefully we continue tonight. I feel like the model is due to let me down at some point or else someone woulda figured this out...

Either way, definitely some good points ManBearPig, I'll keep that in mind.

Predictions for tonight:
MIA @ CHA / Prediction: 191.81 / Line: 192 / Play: Under
(really too close to call, but I have been playing all my plays "strict")
GSW @ ORL / Prediction: 199.35 / Line: 209 / Play: Under
MIN @ BOS / Prediction: 195.27 / Line: 202 / Play: Under
PHI @ NOH / Prediction: 189.3 / Line: 188 / Play: Over
HOU @ DEN / Prediction: 205.7 / Line: 218 / Play: Under
DET @ UTA / Prediction: 199.9 / Line: 192 / Play: Over

17-9 thus far

Buried_PIRATE · 01-03-11, 11:31 PM

Originally posted by Buried_PIRATE

Predictions for tonight:
MIA @ CHA / Prediction: 191.81 / Line: 192 / Play: Under
TOTAL: 178
GRADE: WIN
GSW @ ORL / Prediction: 199.35 / Line: 209 / Play: Under
TOTAL: 200
GRADE: WIN
MIN @ BOS / Prediction: 195.27 / Line: 202 / Play: Under
TOTAL: 189
GRADE: WIN
PHI @ NOH / Prediction: 189.3 / Line: 188 / Play: Over
TOTAL: 161
GRADE: LOSS
HOU @ DEN / Prediction: 205.7 / Line: 218 / Play: Under
TOTAL: > 219
GRADE: LOSS
DET @ UTA / Prediction: 199.9 / Line: 192 / Play: Over
TOTAL: 199
GRADE: WIN

Decent day. Thought the HOU@DEN was going under. Either way, adding 4-2, makes me 21-11. I wonder if I am on to something or if I'm due for a pretty big correction.

pattymayo · 01-03-11, 11:39 PM

Something I noticed-- one of the two you got wrong tonight was PHI/NO, and only wrong by 1.3 points (too high). Now if you followed these teams at all during the season, I'd estimate that 8 out of 10 times it would go Under your predicted total. With your model just 1.3 points above the Line, and then if you combine prior knowledge about the two teams(NO games at home tend to stay low scoring) would you ever go against the model and pick Under despite the 1.3 swing in your prediction? Hope that makes sense,.

Buried_PIRATE · 01-03-11, 11:45 PM

Yeah I definitely would. I'm a huge baseball guy, so this was something I've been doing to kill time.

I started following the NBA more since I got into sports betting. So for instance, I changed the numbers I've been using for the Magic (since the trade) or whenever I notice a key player is out. I don't follow that closely but I'm starting to get into it a lot more and could use that intuition.

I was stupid and lost one of my biggest bets this year when I went against my model tho!!

I think in the back of my head I realize that to keep the system successful is to also follow current team details meticulously to make sure the inputs I'm using are valid

pattymayo · 01-03-11, 11:49 PM

Care to share the factors and variables you use to get your numbers?

ManBearPig · 01-03-11, 11:51 PM

Although our predictive numbers aren't the same we seem to be coming up with similar plays as far as O/U. I have filters in place that filter out games so I don't play them all but I would've went 4-2 missing DET/UTA and HOU/DEN.

People have been successful using these types of models, but you have to continue to monitor the plays and look for corrections that need to be made. I can't speak on the long term because I don't have the data, but if you keep at it you may be surprised what you come up with. I've found filters to be good but too many and you will have a useless system. Don't expect it to fail try to prevent it from failing by keeping up on the as much as possible. You won't have winning nights every night but if you can have more winning than losing your doing well. Good job.