View New Posts
1. ## Basketball Math - Off/Def - Net Efficiency

Anyone ever use this concept to cap basketball ??

I stumbled upon this a few weeks ago and was able to get some decent outcomes over the weekend applying it to NCAA.

Basically I am taking team's offensive efficiency and subtracting it from their defensive efficiency

OFF/DEF efficiency is the amount of points scored for (OFF) or against (DEF) per 100 possessions.

My formula goes like this...

TEAM's OFF EFF x 100
(MINUS)
TEAM'S DEF EFF x 100

= POWER RATING

Then I am taking the power rating of both teams - seeing which team is favored and seeing what the margin is between those numbers and applying it to either the money or the spread.

-

I hit at about 59% rating over the weekend using this method.

Not the biggest sample size - but I feel like there is some value in this formula and will try and hammer it over the course of March.

-

Do you guys have any tips of tricks as per formulas or models to apply to NCAA basketball ??

I am kind of borrowing a concept that helped me quite a bit during the NFL season (Net YPP) and changing the numbers a bit to fit a basketball model.

Let me know if you have any thoughts or additional ideas on this or anything else related to this.

-

Here is an example of my work...

2. too much work, if you can get it automated it might help, manually this is horrendous

3. Originally Posted by nash13
too much work, if you can get it automated it might help, manually this is horrendous
Only took me about an hour to cap 32 games.

Once my spreadsheet is created it's basically just plug and play - need to update the numbers manually every day but that's not a huge deal to me. If you do it manually you don't miss out on the little things like those secret home court advantages that arise from the venues and such.

If it's going to be hitting at a rate of +55%, an hour worth of work is totally fine with me.

4. Go to SBR odds, click on a team, go to matchups. Scroll down, in bold it gives the power ratings.
Do they match up with yours? It will save quite a bit of time. GL

\$30
Charity
donation 4/29/2019

5. Interesting format here...are you going to implement this for the Tourney? Will it work for NBA?

\$20
Angelman
donation 02/18/2019

6. In general, this model is way too simple to have an edge, unless there is something proprietary about the efficiency numbers you are using.

7. Originally Posted by rkelly110
Go to SBR odds, click on a team, go to matchups. Scroll down, in bold it gives the power ratings.
Do they match up with yours? It will save quite a bit of time. GL
It seems their power rating is based on PPG over PACE - mine is a little bit different. But thanks for the heads up there.

8. Originally Posted by SBR drew
Interesting format here...are you going to implement this for the Tourney? Will it work for NBA?
Yes. I will apply it to NBA but I don’t really have time to plug both sets of numbers in - I have had success with this a small bit on NBA - but there is more value in NCAA in my experience.

9. Originally Posted by Barrakuda
In general, this model is way too simple to have an edge, unless there is something proprietary about the efficiency numbers you are using.
If I can hit at a rate of +55% - I still call it an edge. Even if I am winning 11/20 - I am happy with the outcome.

10. Originally Posted by aljack
If I can hit at a rate of +55% - I still call it an edge. Even if I am winning 11/20 - I am happy with the outcome.
If you hit 55% long term you have more than a slight edge my friend.

11. I am usually skeptical of any "systems" but this one passes the smell test for me good luck bro, I am gonna test drive it too probably

12. It's not really a system - more of a model.

I took it a beating using it on Wednesday - moved to my totals model for Thursday - will go back to it on Friday and see how it does.

13. If you add pace to your approach, you'll have a real chance.

14. The kenpom formula turns out to be right on the line, mostly but not always (90%+ including totals). Of course, pace is considered.

15. Originally Posted by aljack
If I can hit at a rate of +55% - I still call it an edge. Even if I am winning 11/20 - I am happy with the outcome.
What makes you think you can hit 55%? Do you really think a model using 2 publicly available variables is going to beat anything? Try logging 1,000 picks at pickmonitor.com, and then you'll know if you have anything.

16. Jack...keep bangin out the research. Harder u work...the luckier u get.

Malinsky referenced this in his NCAA writeup. Villanova apparently has one of best Offensive Efficiencies of Last Decade. Have to figure the Wildcats go deep in tourney.

\$20
Angelman
donation 02/18/2019

17. Originally Posted by ChuckyTheGoat
Jack...keep bangin out the research. Harder u work...the luckier u get.

Malinsky referenced this in his NCAA writeup. Villanova apparently has one of best Offensive Efficiencies of Last Decade. Have to figure the Wildcats go deep in tourney.
Unreal Chucky; they should be winning the East but it's tough to maintain that type of efficiency all the way through to the end. Perhaps they regress a bit, but even then they should walk through their bracket.

\$20
Angelman
donation 02/18/2019

18. Originally Posted by aljack
Anyone ever use this concept to cap basketball ??

I stumbled upon this a few weeks ago and was able to get some decent outcomes over the weekend applying it to NCAA.

Basically I am taking team's offensive efficiency and subtracting it from their defensive efficiency

OFF/DEF efficiency is the amount of points scored for (OFF) or against (DEF) per 100 possessions.

My formula goes like this...

TEAM's OFF EFF x 100
(MINUS)
TEAM'S DEF EFF x 100

= POWER RATING

Then I am taking the power rating of both teams - seeing which team is favored and seeing what the margin is between those numbers and applying it to either the money or the spread.

-

I hit at about 59% rating over the weekend using this method.

Not the biggest sample size - but I feel like there is some value in this formula and will try and hammer it over the course of March.

-

Do you guys have any tips of tricks as per formulas or models to apply to NCAA basketball ??

I am kind of borrowing a concept that helped me quite a bit during the NFL season (Net YPP) and changing the numbers a bit to fit a basketball model.

Let me know if you have any thoughts or additional ideas on this or anything else related to this.

-

Here is an example of my work...

Yes aljack. I use a similar logic to my NBA "strategy". However I don't use for sides but only for total points.

It's not a bad way to look at games/picks however what I have noticed, having a sample of aprox. 2500 analyzed games is that it tends to lose at aprox. 48.8 clip. Again, we might not use the exact same formulas, etc but the concept is the same.

I have found other such indicators I am currently tracking (PACE, Away/Home pts for and against, etc) and that I use in conjunction with off/deff indicator.

In short, I keep track of 7 indicators ( total grind) and use just two of them in opposition:

This is working so far in this NBA season at currently 57.3% clip.

Sample size is vital. Must have thousands of games analyzed so you can take out the outliers.

Cheers

19. Originally Posted by allnighter
Yes aljack. I use a similar logic to my NBA "strategy". However I don't use for sides but only for total points.

It's not a bad way to look at games/picks however what I have noticed, having a sample of aprox. 2500 analyzed games is that it tends to lose at aprox. 48.8 clip. Again, we might not use the exact same formulas, etc but the concept is the same.

I have found other such indicators I am currently tracking (PACE, Away/Home pts for and against, etc) and that I use in conjunction with off/deff indicator.

In short, I keep track of 7 indicators ( total grind) and use just two of them in opposition:

This is working so far in this NBA season at currently 57.3% clip.

Sample size is vital. Must have thousands of games analyzed so you can take out the outliers.

Cheers
Yea.

Usually what I'll do is I'll use a combination of stats and if there is a consensus throughout all of them - that will be my pick.

If you look here --> https://www.sportsbookreview.com/for...h-madness.html

I have been using this combination of numbers along with a few other statistical models to get a decent read on games.

I am hitting at about 56.5 %

I use:

NET EFF RATING:
Offensive Efficiency (x 100) (minus) Defensive Efficiency (x 100) = Rating

MOE RATING:
The Margin Of Expectation Rating is the average number in which a team is against the spread all season up until the tournament. The spread is fundamental measurement of expectation for a team.

PPG Rate:
Average Points Per Game for each team - as a straight up number, against the spread and as a game total.

Net PPG Rate:
Average Scoring Differential Per Game for and against for each team - as a straight up number, against the spread and as a game total.

-

I also used to use PXSP Rating (Avg Possessions Per Game (X) Average Shooting Percentage) - however I realized the number I was getting was almost exact to the PPG - so I scrapped that idea.

Next year (unfortunately) I will be focusing more on more derivative stats.

"Assists to Turn Over Ratio" is a really sharp stat.

1H Net PPG & 2H Net PPG - also both really sharp stats.

Hoping to clean up next year. This season has been a learning experience.

20. I think one question you need to address is how much data should you include for a team. Season YTD data or last x number of games? Do you also segregate statistics accumulated at home or on the road?

21. Originally Posted by Bsims
I think one question you need to address is how much data should you include for a team. Season YTD data or last x number of games? Do you also segregate statistics accumulated at home or on the road?
Season YTD data.

I include the MOE rating to get a better idea of how a team has been fairing in recent times.

If they have a positive MOE rating then I know they have been beating expectation as of late.

I do not segregate home/road stats - I only started doing this when conference tournaments started - which are mostly on neutral court.

22. Originally Posted by aljack
Season YTD data.

I include the MOE rating to get a better idea of how a team has been fairing in recent times.

If they have a positive MOE rating then I know they have been beating expectation as of late.

I do not segregate home/road stats - I only started doing this when conference tournaments started - which are mostly on neutral court.
Just sharing my results on YTD stats in general: they tend to lose just nearly under 50%
However the last x (3-5) number of games tend to "win" at aprox. 51 % (obviously this stat alone will lose you money)

23. and a small correction regarding the YTD for Average scored points for and against in the previous NBA season (tracked for 981 games) was 49.9%
This is how tight the lines are...

24. Originally Posted by aljack
Season YTD data.

I include the MOE rating to get a better idea of how a team has been fairing in recent times.

If they have a positive MOE rating then I know they have been beating expectation as of late.

I do not segregate home/road stats - I only started doing this when conference tournaments started - which are mostly on neutral court.
As the season progresses markets tend to catch up to YTD data. It's best to account or at least weight for recent performance.

There are many differences among markets, but there are also similarities. Truthfully, to handicap most sports, you need about for games of data for each team, the last four played. Handicapping can spawn from there to as much or as little detail as necessary.

\$20
Angelman
donation 02/18/2019

25. looking at your numbers for tomorrow. seems like you like the over in both ncaa tournament games and villanova to cover the -6.5. is that correct? aljack.

26. Not gonna bet the overs - but Villanova -6.5 & Duke -3.5 are my plays.

27. a number of thoughts,

i do think it's possible for simple things to work. not highly likely but possible.

what about strength of schedule? i'm assuming the efficiency #'s aren't adjusted... injuries too. past or present

if a team is 3 points better than average vs. a team 3 worse than average, i don't think it follows that they should be six point favourites neutral court. although not sure that different from 6 points.

isn't this basically what kenpom, sagarin, teamrankings.com etc. do? and probably the first places that oddsmakers go when making lines. they may completely disregard these numbers but i bet they look at them

predictiontracker tracks this stuff....... i think you'll find in time that this type of analysis gives you ideas and then you interpret the ideas i.e. you develop expertise

28. Originally Posted by gojetsgomoxies
a number of thoughts,

i do think it's possible for simple things to work. not highly likely but possible.

what about strength of schedule? i'm assuming the efficiency #'s aren't adjusted... injuries too. past or present

if a team is 3 points better than average vs. a team 3 worse than average, i don't think it follows that they should be six point favourites neutral court. although not sure that different from 6 points.

isn't this basically what kenpom, sagarin, teamrankings.com etc. do? and probably the first places that oddsmakers go when making lines. they may completely disregard these numbers but i bet they look at them

predictiontracker tracks this stuff....... i think you'll find in time that this type of analysis gives you ideas and then you interpret the ideas i.e. you develop expertise
Generally I get the most edge when I combine a whole bunch of different models and go with a consensus.

I am going to get into blending stats for different time frames as well - I see that there is room for improvement and I know exactly how to make an adjustment that will improve my models.

29. al, you have a good run in ncaa tournament.

30. Originally Posted by aljack
Generally I get the most edge when I combine a whole bunch of different models and go with a consensus.

31. Originally Posted by Barrakuda
Dunno whats funny here. Ensemble methods (https://en.wikipedia.org/wiki/Ensemble_learning) provide more accurate results than single models in most cases. Stuff like Random Forests (https://en.wikipedia.org/wiki/Random_forest) can even provide very accurate predictions even when each individual model is barely better than chance.

Even if the system mentioned in the OP doesn't bear fruit, the fact aljack has (independently?) worked out ensembling suggests he's at least near the right track.

32. Originally Posted by Barrakuda
In general, this model is way too simple to have an edge, unless there is something proprietary about the efficiency numbers you are using.
This.

33. Jack, u may have something here. Keep working at it.

Note how Villanova did come thru. Such an efficient offensive team. Noted by Dicenzo's showing in Finals.

\$20
Angelman
donation 02/18/2019

34. Originally Posted by aljack
It's not really a system - more of a model.

I took it a beating using it on Wednesday - moved to my totals model for Thursday - will go back to it on Friday and see how it does.
when you mean how it does? do you mean how it does compared to the closing line or the final scores? who cares how it does to the final scores, we only care how it does against the closing line