UVA I agree with you whole hardly. The above example is a great start. All you have to do is come up with some filters and breakdown your records by HF/HD/AF/AD and by lines. I guarantee you will come up with some good results. To give you an ex:, I have a simple system in which I only use win/loss records which I will not devulge. The system comes up with a pretty good line. I then apply filters and I am hitting well over 60% in just about every sport. It takes work and from year to year you may have to make adjustments because teams do change.
-
UVA I agree with you whole hardly. The above example is a great start. All you have to do is come up with some filters and breakdown your records by HF/HD/AF/AD and by lines. I guarantee you will come up with some good results. To give you an ex:, I have a simple system in which I only use win/loss records which I will not devulge. The system comes up with a pretty good line. I then apply filters and I am hitting well over 60% in just about every sport. It takes work and from year to year you may have to make adjustments because teams do change.Comment -
Where do you suggest you plug in the league median in UVA3021's formula?Comment -
What is the possession formula for NHL?
And what is the offensive and defensive efficiencies for NHL?
Any formula and stats for this ?
MilwaukeeComment -
I think you have this backwards. It's watching games that leads to deception. Math leads to clarity. It's like saying that, after driving from Vegas to LA, guessing the distance is better than checking your odometer.
As to the rest of the discussion, I refuse to contribute anything. I will not so much as nudge a single one of you the right direction, LOL. The game's too tough as it is. But I will say that everything presented so far, to me, is just the new vanilla, and thus will not beat the spread going forward. Not that you should take my word on it or miss out on the fun of playing with numbers.
Wait, is that a nudge?Comment -
Literally, this means that you you do not need the league average points per possession if you already accounting for the league average points per possession. How do you account for something if you do not know it? Well, you don't, but you get it from somebody else instead! So, while Justin7 made no sense he probably meant obtaining adjusted efficiency numbers from an outside source like kenpom.
Since uva3021 is not rushing to redeem himself, here is a hint. Kenpom is one of the places where you can find the proper methodology/formula.Comment -
i agree with you about averages, and all the numbers should be standardized to their schedule (the average efficiency of the teams they played as a value of 1)
no need to redeem myself, and since I don't really use the basic formula anymore, there is no need for me to adjust itComment -
Hey guys, I like where this thread is going. I have been testing a system like this and been hitting around 60%.
One thing that I know I need add in how certain paced teams react to others. I think I would be able to get a better handle on this if there was somewhere that listed the amount of possessions in individual games. Obviously, this can be calculated, but it is quite tedious to do for every game every day.
Does anyone know somewhere that would list this?Comment -
Wow good stuff but gambling is still hard as heckComment -
has anyone actually verified that the OP's formula is better then the basic pace rating formulaComment -
The opening line has already done the math for you and your time would be better spent finding valid up to date reasons on how best to beat it rather than time wasting formulas that in all probability are missing essential latest news.
Comment -
But how often have we seen a sharp looking line be completely off?
The Vegas linesmakers are human and they make errors. Human error needs to be accounted for and I don't give much credit to line movement as an effective betting strategy. Yes, it definitely helps with money management and value, but not with handicapping.
The problem (which all of us should know at this point) is that there are ALOT of outside circumstances that affect how a game is played. From personal player issues to officials to current league wide trends (remember the NHL under streak?) a model will never be completely accurate, but we're aiming for a 54% hit rate.
My personal model has tonight's best play to be to be Clev/Miami Over 190.5. However, this is a perfect example of outside circumstance, notably LeBron's return. Is he going to put up a lot of points? Maybe in trying to he'll force a lot of shots and score much lower than average. Maybe Cleve's defense will focus on shutting him down and Wade throws up 45+. Or maybe both teams will play this just like an average game.
Meanwhile for the GS/PHX game, it's handicapped at the Under, but barely. The better play would be GS -2.5.
Filters are definitely needed (for instance, I wouldn't take the PHX/GS Under due to filters same with Cleveland +5.5) but I feel that there has to be a better way.
My question is should league average be accounted in the handicapping formula or should it rather supplement it? For instance, two high scoring teams with weak defenses are going to have their expected points brought down by accounting for the league average. I can see why most games should definitely account for it, but there are certain games which seem like they shouldn't.
Any thoughts or insight?Comment -
Would this be because of the assumption that their previous opponents would fit the standard deviation of the league average?
Also, are filters a good idea? I don't want to spread out too thin without them, but I'm worried I might be cutting more into winnings then anything else.Comment -
well lets say the team averages 105 ppg , the league average is 100, and the opponent average points allowed per game is 102
then 105*(102/100) = 107.1Comment -
So both totals from my model hit, but both ATS missed.
The GS/PHX game was much lower than it predicted, but hit the under regardless. I'm going to keep with this for totals only, but watch the spreads and see how I can tweak this.
Here's for today:
Portland/Washington: Over 194.5 (Strong Play)
Charlotte/NJ: Over 185 (Strong Play)
Orlando Portland: Over 186(Strong Play)
Philly/Atlanta: Over 192.5 (Strong Play)
NY/NO: Over 204.5 (Weak Play)
Houston/Memphis: Over206 (Strong Play)
Note: The fact all of these are coming out strongly of the over is worrying me...
Chicago/Boston: Over 190.5 (Weak Play)
Minnesota/SA: Over 213 (VERY Weak Play)
Indiana/Phoenix: Over 213 (Strong Play)
LAC/Denver: Over 211.5 (Weak Play)
Sac/LAL: Over 200.5 (Strong Play)
Dallas/Utah: Over 192.5 (Strong Play)
Whether or not the play is strong or weak is what the model is telling me, not my analysis.
All overs...
Damn.... Have to see what happens and make some adjustments...Comment -
How are you calculating expected pace?So both totals from my model hit, but both ATS missed.
The GS/PHX game was much lower than it predicted, but hit the under regardless. I'm going to keep with this for totals only, but watch the spreads and see how I can tweak this.
Here's for today:
Portland/Washington: Over 194.5 (Strong Play)
Charlotte/NJ: Over 185 (Strong Play)
Orlando Portland: Over 186(Strong Play)
Philly/Atlanta: Over 192.5 (Strong Play)
NY/NO: Over 204.5 (Weak Play)
Houston/Memphis: Over206 (Strong Play)
Note: The fact all of these are coming out strongly of the over is worrying me...
Chicago/Boston: Over 190.5 (Weak Play)
Minnesota/SA: Over 213 (VERY Weak Play)
Indiana/Phoenix: Over 213 (Strong Play)
LAC/Denver: Over 211.5 (Weak Play)
Sac/LAL: Over 200.5 (Strong Play)
Dallas/Utah: Over 192.5 (Strong Play)
Whether or not the play is strong or weak is what the model is telling me, not my analysis.
All overs...
Damn.... Have to see what happens and make some adjustments...Comment -
Hello everyone!
Since I was about to post some questions related to team efficiencies and the thread is fairly 'fresh', I will post it here.
I tried to model team's points scored in a game based on possessions and both team's and its opponent's efficiencies. I know this is only the first step to modelling anything, but I wanted to make sure I am calculating things properly before I add other variables. I tired a few approaches to came up with a very basic estimate of home team pts scored in a game (Hpts) based on home team past offensive efficiency (HOE) and its opponent's past defensive eff (ODE), as well as projected pace of the game measured in possessions (Poss). Basic formula would be (following Justin's book):
Hpts=HOE*Poss (and divided by 100 if HOE is measure in points per 100 possessions)
Then I modelled expected possessions for the game using regression, where y was number of actual poss in a game (actually it is also an estimate, based on formula provided in the first post of the thread or even a bit simpler formula - no significant difference between them).
I tried a few regressions (all estimated in excel, using OLS), but results did not seem to differ to much. Explanatory variables were
a) HAP, AAP (average season-to-date average pace for home and away teams, respectively)
b) HAP, AAP, HAP^2, AAP^2, HAP*AAP
c) same as b) but using ln (logarythms), as well as using ln for y (then of course using exponent function for obtaining proper pace no)
I choose c) because R^2 was lowest, but all of the shomed similat results as I mentioned. This step I am quite confident is conducted properly. The next involved calculating home team's offensive eff for this game and here I have serious doubts what should be done.
For this, i have used the same 3 types of regressions as described above, where two explanatory variables were: HOE and VDE. This time I chose function described in b).
Now, for the whole 2009-10 NBA season the market closing line error (derived from closing line total and spread), measured as a mean absolute its distance from the actual pts scored by a home team, is 8,16 (excluding some chunk of the games at the beginning of the season to allow the stats to be more reliable).
The abs mean error for the model I specified above is 8,41.
But as I read this thread, I thought that I made a mistake by not normalizing for league avarages, so I decided to incorporate it.
I am pretty sure that I make some mistake in methodology since every way I tried to incorporte the league averege (in HOE estimates), the error increased. The adjustments for HOE estimates were as follows:
1) dividing HOE by 1,1 (league average for points per poss in general) - error=8,81
2) dividing HOE by 1,115 (league average for points per poss scored by home teams [and allowed by visiting teams]) - error=9,19
3 and further) involved adjusting possessions by the league average, but I assume that if you normalize for points per possessions than you have already normalized for both league averagers in points scored and possessions. - error of 10;11 or even over 12.
So as you can see by normalizing for league average in efficiency I increased the error. Please someone tell me what is wrong with my calculations, because I am pretty sure that adjusted it the wrong way.
Hope everything is clear and thanks for helpComment -
Instead, get familiar with Z-scores, log5 method, and with this:
http://www.rawbw.com/~deano/methdesc.html#corrgaussComment -
Nice stuff here.Comment -
Thank you data for tips, I am familiar with z-scores and heard about methods relating win% to win probability for a given team.
But as far as what I have posted and with reference to what had been written here about poss and eff, are my calculations proper or is there (and where) a miscalculation? I wanted to develop a model which would predict margin of victory and total, and but merely relying on win% you can only derive spread from expected moneyline (the same way moneyline is derived from spread thew other way) and you do not account for other factors.Comment -
Your calculations represent the methods that are inferior to the methods I referred you to. Those methods have more uses than you may realize at this point. Your current methods are no better than coin flipping. Essentially, you are asking if you are flipping your coin the right way.Comment -
Data,
[In general], why always so hard on regression analysis?Comment -
"Secondly sex is way overrated, I have peeped in on people and it seems kind of boring
Porn is an act and not the same as the real thing
Bread I rather throw my 2.5" pole in the water and catch fish and not pussy "
The real question here is what is all this about?Comment -
Well, I guess, in school they tend to talk more about descriptive statistic which seems to be the prevalent branch. The regression analysis are fine for descriptive statistic but frequently are dangerous for a sports bettor who is much more interested in predictive properties of his model. This is why uncovering underlying logic and coming up with meaningful (contrary to regressions) formulas will have a much better chance in predicting future results.Comment -
Is there an alternative to probit/logit models for, let's say, Yes/No props and the related problem of calculating half-point values? The smooth predictions should be more accurate than coming up with a table where the "x-axis" is, let's say, many ranges of moneylines and the "y-axis" is different totals.Comment -
looking for help. I would like to compare my numbers to someone elses. Im working on score poss. , poss, floor %, power %, and play %. Does anyone else do thisComment -
i don' have time for to track the detailsComment -
what is the formula for finding out a teams pace? or is there a site that has team stats such as that?Comment -
how do you find out a teams pace?Comment -
i would think a team's pace could differ in the NBA in a B2B, or 3+ days rest, as well as the pts per possessionComment -
I wish there were a way to win without all this complication...Comment
Geeez!
