Posted:
#1
As of recently I've grown an interest in Basketball statistics. I'm tinkering with a simple home made mathematical model calculating NBA lines, a bit crude as of now but I'm looking to refine it, hence starting this thread. It takes into account quantifiable factors known before the game; variables in the model are:
*Offensive rating/Defensive Rating/Pace for the last 6/last 12 matches/Season
*Hcourt advantage adjuster (crude average except for DEN and UTA, only two teams with an above average statistically significant Hca in the league).
*B2B road games adjuster
*Referee bias (optional)
On a side note I also have an input where I can get pythagorean expectations (%age probability) for covering the spread and the total based on season to date results.
Here's a picture:
https://dl.dropboxusercontent.com/u/5628798/Screenshot%20from%202015-03-29%2013%3A02%3A44.png
I might get into the details how everything's calculated and weighed together if someone's interested and I have the time to do a write-up. Model is full time+OT only. It gives out lines pretty close to Vegas opening lines most of the times. It actually produces an expected score for each team which is nice since you could use it to judge not only the spread but any betting proposition; totals, team totals etc.
What is interesting of course is when lines diverge. An example was yesterdays matchup GSW@MIL. Books had GSW at -5, model had them at -13.5. Ended 108-95. No team news out that could account for more than a few point give or take. A single game doesn't say anything of course, just brought it up as an example.
What is missing is team news and lineup changes, since the model only works with averages. There's no way it can take into account an event like yesterday's when ATL decided to rest their usual starting 5. Primarily I think I'll be posting interesting diverging lines so anyone interested can give input as to why they might differ; what info is out there? If there's no reasonable explanation that could be a value scenario.
If anyone's got suggestions on simple quantifiable variables that are easily found online and that could be incorporated that is also welcome. Please have some underpinning to your reasoning in this case: your own data or someone else's analysis (give source). For example I adjust the score with one point if B2B road; I know this is the average effect on the point diff since I have a database with 8,000 NBA games where I have looked this up. General statements without evidence like "I think four road games in five days have an impact on the total going under" are pretty useless. Also please don't suggest any kind of streaks. They are random, period.
Grateful for your feedback! I won't be posting everyday but every now and then.
DISCLAIMER. THIS IS A HOBBY PROJECT. I HAVEN'T MADE ANY MONEY MYSELF BETTING THE NBA. I CAN BARELY TELL A POINT GUARD FROM A CENTER AND IF YOU'D ASK ME ABOUT LEBRONS CURRENT FORM I WOULDN'T HAVE A CLUE. MY INTERESTS ARE PRIMARILY STATISTICS, PROBABILITY THEORY AND PREDICTIVE MODELING, WHICH ALL COME TOGETHER NICELY IN THE PROCESS OF VEGAS LINE MAKING AND ANALYSING. THIS IS A WORK IN PROGRESS, I'M CALIBRATING THIS TOOL AND NOT WAGERING ANY MONEY MYSELF AS OF YET. DON'T EXPECT PROFITABLE RESULTS FOLLOWING ANY LEADS. BE AWARE THAT VEGAS LINES ON MAINSTREAM SPORTS ARE VERY SHARP.
0
)
Hot Forum Topics
