Does anyone use/follow his power rating system? Using the Waybackmachine I was able to find parts of prior years weeks but am missing about half of them. Thanks in advance fellas.
Printable View
Does anyone use/follow his power rating system? Using the Waybackmachine I was able to find parts of prior years weeks but am missing about half of them. Thanks in advance fellas.
nope. it's all public knowledge and very basic starting points.
I like using Sagarin as a starting tool, especially in football. It has helped me identify some valuable underdogs over the years.
i use him also!!!
Looking into it recently
i check his stuff out it has some vaule
Ya I know it is very basic but his rankings are only part of my system. Do you guys know if it is actually published in the paper or is it just online?
It used to be published on Tuesdays.
I find it more useful for college hoops -- especially in some of the smaller leagues like the one with all of the Tennessee teams......
His rankings, btw, are pretty much the definition of vanilla. That isn't to say that they're bad. They're better than what you'd come up with your first try. Just bear in mind that whenever there is a difference between Sagarin and the line, there's a reason for it.
http://www.thepredictiontracker.com/
Somewhere on this site you can find his record ATS. Also, that of all the other public rankings. There is an excel-friendly (comma delimited) file you can use to manipulate the data a bit to check for results by the amount of estimated line value.
That's what i figured. I need to get copies of old newspaper, any suggestions?Quote:
Originally Posted by Pokerjoe
Don't hold your breath. Mostly a throwback to another era. I researched them some years ago, and did come up with some edges. But any number grid put together with any resemblence of intelligence will do that.
Dodger, if you have a db of scores over time, Sagarin's PREDICTOR/BALLANTINE/RHEINGOLD/WHITE OWL algorithm and the ELO can be built to generate his ratings. A book named "The Hidden Game of Football" (used at Amazon for under $2) has predictor documented, and ELO is a well known Chess rating algorithm (uses wins and losses only) that you can find by Googling.
Now you can save a bunch of cash on old USA Today newspaper on Ebay.
I just told you: go to the website I posted above. Look around and you'll find it, Sagarin ratings, weekly, with the spread and score for the game, too.Quote:
Originally Posted by dodger33
Wow i didn't look at it hard enough the first time. Thank you very much!Quote:
Originally Posted by Pokerjoe
I will have to check that out. Thanks!Quote:
Originally Posted by Flying Dutchman
Site is great...thanks
if you want to get previous end of year rankings, easy enough to just switch the date in the url at the end
weekly is tougher to come by
I used Sagarins ratings a little bit but only in college hoops. Useful information but nothing groundbreaking.
Quote:
Originally Posted by mtneer1212
Yeah for college hoops he is valuable to a degree. Ohio Valley Conference is a good example.
It's always worth a look during the season as to how the Sagarin ratings shape up. Typically I feel that it takes at least a month of games for the ratings to really mean anything. Have to also be careful in how it stacks teams up since they might have gotten a strong rating thanks to a few blowout wins at home, but did not do as well on the road, etc.
In all the years I have followed these numbers, I have found home dog plays with them seem to do best but every game is different.
sagarin is all math......
I used the Sagarins ratings in betting NBA games last season. It is a very useful tool. I am going to try the same method I use in NBA in the coming CFB and NFL season. These numbers are not all maths.
It can be useful for some things, but not as good the more well-known the team is.
He is somewhat helpful in NCAA football especially if there is a large discrepency between his projected line and the actual line.
So what are they? All ice cream?Quote:
Originally Posted by Augustus
...do you guys even think about what you are saying?
I appreciate how Sagarin does a good job pointing out the difference in points between the #7 team and the #83 team overall for NCAAF. His power index values are solid. However, I like the Massey comparison for ranking the teams for those values. Massey does a best fit/consensus of all the rankings out there. Add in a turnover adjustment and you pretty much have the lines LVSC release.
sagrin ratings work more than they dont
what does this mean?Quote:
Originally Posted by dinaro7
you cannot blindly bet the difference between sagarin and the market line and make money.
why does he use the same adjustment factor for all home teams when obvioulsy some teams are better or worse at home? (rhetorical)
Quote:
Originally Posted by Peregrine Stoop
Really. Sag ratings are vanilla stats. You can not find a good sized sample in which they beat the line. And of course not: even if they could beat the line, everyone would be betting them (it is public information, after all); and if everyone was betting them, the line would move and they wouldn't win any more.
Why do you say it's obvious? There are some extra HFAs in some spots, yes. But most of the differences between h/a results are just noise that don't hold up over time. HFA samples are sick small.Quote:
Originally Posted by arwar
But this is a can of worms. I mean, there are still people who think giant SEC stadiums create extra HFA, and they get very emotional in their insistence, especially on the forums.
If anyone was going to use Sagarin effectively, I'd say use it as a starting point, and then adjust from there for injuries, motivation, specific HFA, as you see fit. It can give you a numerical anchor, anyway.
A question ...the idea behind something like say Sagarin is to estimate how many points Team X would score against Team Y. So from this you can figure out that say Team X should beat Team Y by say 20 points 50% of the time.
Let's assume temporarily (for the sake of an example) that you believe Sagarin is a better predictor than the Vegas line (this probably isn't true, but I'm just trying to understand something here.)
If Sagarin says the line should be -20 and Vegas has it at -5, then I'd feel pretty confident about betting on the favorites. But what if Sagarin says -20 and Vegas says -10? Or -20 and -19?
Like, as the gap between Sagarin (or whatever method you use) and the Vegas line shrinks, you have less and less confidence in your bet, right?
So how do you set cutoffs for discrepancies between your favorite handicapping method (which you believe is accurate) and the Vegas line?
One simple way that makes sense to me is to view the Sagarin ratings as random, say normally distributed estimates of whatever the final margin of victory of the game is. So you could take the Sagarin number and then tack on a standard deviation number to it.
So for example, if Sagarin says -20, then I'd also need to generate a standard deviation...let's say the number 4. The smaller my standard deviation, the more confidence I have in each Sagarin number, and the more confidence I have in my bet.
Anyway, I'm just wondering how you guys deal with this issue (effectively confidence intervals from statistics: http://en.wikipedia.org/wiki/Confidence_interval).
The only cutoff you're looking for is the point at which betting isn't profitable (that is, the point at which you don't have an edge over the line and juice).
There are two continuums involved in modeled edge estimates:
A continuum of modeled line discrepancy.
A continuum of confidence in your model.
The first continuum might not be linear (a 10 pt line discrepancy might not be 2x the value of a 5 pt line discrepancy), but it should be progressive (a 10 pt line discrepancy should be worth more than a 5 pt line discrepancy). You'll need to develop an algorithm to translate your line discrepancy into an edge estimate.
The second continuum is based mostly on the (non-data-mined) sample size behind it. This is something that you should basically be thinking of in terms of Kelly Fraction; that is, as your confidence in your model increases, so should your Kelly Fraction. You might want to write an equation translating z-scores into Kelly fractions (search this sub-forum for more info on z-scores or Kelly).
Your actual results will lie somewhere between a system an actually-existing God gives you (warranting a Kelly "fraction" of 1) and the equation you buy off a guy in the parkinglot outside a racetrack (warranting a Kelly "fraction" of 0).
BTW, Sagarin rates a 0.
Have fun.
use sagarin
adjust for turnovers
adjust for special teams
adjust for 4th down and 3rd down luck
profit
do i sleep with an old chick tonight? hurrry
Is she fat?Quote:
Originally Posted by aggieshawn
This is something I'm wrestling with at the moment. How do you quantify 'confidence' quantitatively? I'm sure each person has their own way, but I just want to know how other people come up with their 'confidence continuum'.Quote:
Originally Posted by Pokerjoe
What do you mean by, "adjust for special teams?" Obviously you need to neutralize anything you think is more luck than skill and has influenced past scoring (turnovers and 3rd down rates, like you say). But I do not think special teams (other than excessive returns for TD) is mostly luck, and thus should not be adjusted for directly when tweaking a score-based power rating system. Do you disagree? If so, what kind of adjustments do you make?Quote:
Originally Posted by Peregrine Stoop