What The Sagarin PREDICTOR Tells Us About The BGSU Game
Although many folks quote the Sagarin “Rankings”, this is actually the least insightful of the Sagarin data. Let’s take a look at the “RATINGS” for Michigan and Bowling Green to see what we should expect on Saturday.
Expect another barn burner for BGSU! Michigan is predicted to be just under a 14 point favorite. Do not expect a blow-out.
Sagarin uses two basic ratings: PREDICTOR (in which the score MARGIN is the only thing that matters) and ELO-CHESS (in which winning and losing only matters, the score margin is of no consequence). The overall rating is a synthesis of the two diametrical opposites, ELO-CHESS and PREDICTOR.
Per Sagarin: ELO-CHESS is “very politically correct. However, it is less accurate in its predictions for upcoming games than is PREDICTOR” which is the best single predictor of future games.
To make predictions for upcoming games, simply compare the RATINGS of the teams in question and allow an ADDITIONAL 3 points for the home team.
|
|
Predictor |
Rank |
Sch Rank |
EloChess |
Rank |
Rating |
Rank |
|
Michigan |
77.83 |
36 |
61 |
83.27 |
19 |
80.55 |
24 |
|
Bowling Green |
67.13 |
83 |
69 |
63.2 |
103 |
65.37 |
90 |
|
Difference |
10.7 |
|
20.07 |
15.18 |
|
||
|
Home Field |
3.01 |
3.01 |
3.01 |
|
|||
|
Total |
13.71 |
23.08 |
18.19 |
|
I've included all the Sagarin ratings and rankings data but the PREDICTOR is supposed to be the most accurate.
September 20th, 2010 at 2:31 PM ^
as I heard a rumor that the entire defense is getting Denard Shoes and trashing the concrete shoes they tried last week.
September 21st, 2010 at 9:13 AM ^
Is that what happened to the Defense? *smacks head* Well that explains SO MUCH!
September 20th, 2010 at 2:37 PM ^
Seems great...what did it say about the James Madison- V. Tech game?
I actually like the way Sagarin goes about it but football is just a nuts game. My buddy made all his bets based on the largest variations between the actual lines and Sagarin for 1 year and cleaned up. He swore by it. Then he lost with it the last 2...now not so much. Not sure if it's any better than the AP poll. That would be a good analysis for anyone with access to the data. Sagarin predictor success vs Vegas predictor success....take all the games and see who was closer to the outcome.
September 20th, 2010 at 2:37 PM ^
sounds about right to me. Defense will come out strong. Offense will get to use back-ups.
September 20th, 2010 at 2:40 PM ^
Sagarin would therefore predict a two-TD win? Is that a barn-burner? And if we factor loss of their QB?
I think we should expect a win, and an easier game than last week--with a chastened D and an O continuing to get untracked. But it's an IMPROVED D that we gotta really hope for. What kind of offense does BGSU run?
September 20th, 2010 at 2:45 PM ^
Apparently, around these parts a couple touchdowns *is* a barn burner.
Anything less is UNACCEPTABLE and must be in all caps.
September 20th, 2010 at 3:40 PM ^
Clarkiefromcanada: I thought you of all people would have a good grasp of the king's english.
Barnburner: "A barnburner is a very high scoring athletic contest."
If there are not lots of points scored on Sat. I will be shocked and dismayed.
September 21st, 2010 at 12:20 PM ^
Another factor is the two Pick 6's that BG had last game. Sagarin uses score margin in his analysis and those 2 TD's are extremely unlikely and most likely skew the data so early in the season.
September 20th, 2010 at 2:43 PM ^
I don't know what the stats are underlying this "elochess" thingee, but if any of you followed nate silver at fivethirtyeight.com during the elections, you saw a better solution - a monte carlo simulation.
Which is basically:
Estimate a distribution of points for Michigan (i.e. we think they will get xx plus or minus y). This can include any effect of home-field advantage, injuries, whatnot.
Estimate a distribution of points for BGSU (we think they will get ww polus or minux z).
Get the computer to draw from the distribution of points for each team (no more complicated than drawing from a deck of cards); the team with the higher # points "wins" round 1.
Do this 1000 times (get a beer while the computer does it for you). The % of wins over the entire set will be an unbiased, "non-politically-correct" version of whatevertheheck "elochess" is supposed to be.
September 21st, 2010 at 9:11 AM ^
and the real answer is probably doing this 1000000 times (not just 1000)...after all, your computer is fast, and you have a fridge full of beer.
But the real question is how do you get your estimates? Is it record/resume? How do you include that? Do you just use RPI score (or a similarly calculated thingy)? And (if you want to be a BCS computer ranking) how do you do this without looking at scoring margin.
September 21st, 2010 at 9:21 AM ^
In the case of politics, they use
<br>Polling results. For football, you could use some form of regression analysis based on historical data (although come to think of it you could get a similar result by asking enough people just to guess and collecting their guesses into a distribution).
September 20th, 2010 at 2:58 PM ^
I think even Sagarin would admit that any predictive value is almost worthless until the prior season's results, which simply act as a bias, are no longer part of the computation.
September 20th, 2010 at 3:13 PM ^
I wonder how the predictor is doing ATS this season.
EDIT: Found the answer - Sagarin Predictor 47% ATS (61-68) and Elo 49% (64-65)
September 20th, 2010 at 3:35 PM ^
I am horrible at this stuff but doesn't that mean that the games were closer than predicted? Or do I have that backward?
September 20th, 2010 at 3:43 PM ^
It means that the Sagarin predictors are about as good picking against the spread (ATS) as one flipping a coin would be over time. Basically the predictors picked the correct outcome ATS about 50% of the time.
So basically, what the OP is basing his closer-than-the-experts-think prediction on is a system that is picking correctly 50% of the time. This may improve over time as the Sagarin system gets more data.
September 21st, 2010 at 8:15 AM ^
Absolutely WRONG!
Every single spread in every system is calculated so that it is "correct" precisely 50% of the time. The closer to 50% ATS the better the predictor.
It would be pretty easy to always "win" if your system was always had spreads that were consistently too low. Of course, such a system would be useless.
September 21st, 2010 at 9:03 AM ^
you have to think like the Vegas sports book here. They want to win both ways, and they often adjust the lines so that they have about the same amount of money on each side. So if team A wins, Vegas wins, and if team B wins, Vegas wins. Like they say...the house always wins. So if you want to win at gambling, be the house.
/gambling advice (to be taken purely in jest--HA!)
September 21st, 2010 at 9:04 AM ^
Or rather, it would be right if the spread was the optimal prediction. That's what they're trying to achieve, of course, but the spread is not magical, just another form of predictor, and a superior predictor would beat it more than 50% of the time.
Sweet license plate.
September 21st, 2010 at 9:25 AM ^
You said "I am horrible at this stuff". I'd stick with that. The predictors forecast the difference in score or the actual score of the game and really have nothing to do with ATS when they are formulated. ATS is set by the casinos to achieve a 50-50 split in wagers. The Predictors' performance ATS is what is being tracked by sites like collegefootballpoll.com and thepredictiontracker.com. If they are only 50% ATS, their prediction is akin to flipping a coin ATS and are pretty useless to a gambler.
September 21st, 2010 at 11:03 AM ^
I suspect that part of the problem with the predictor is that early in the season, it incorporates data from the previous season that is no longer fully applicable- it has no or little data from the current season to use to make predictions. On the other hand, you can use logic and your instincts to process information that no algorithm can, and make bets based on your gut.
September 20th, 2010 at 4:18 PM ^
so after all that we should win.
September 20th, 2010 at 5:25 PM ^
by more than two TDs, it's going to be a long season.
September 20th, 2010 at 5:27 PM ^
when the customers from the local Torch & Pitchfork outlet show up at the AD's office after a particularly painful loss.
September 20th, 2010 at 8:00 PM ^
You're back!
You are the "real Don", right? Not that impostor fella...?
September 20th, 2010 at 9:27 PM ^
I'm thinking it's really Dick Whitman.
September 21st, 2010 at 2:46 PM ^
September 20th, 2010 at 10:58 PM ^
It was the correct ugly mug staring back at me in the mirror when I was shaving this morning, so I guess I'm not the impostor.
Speaking of which, I haven't seen Webber's Pimp around today. After he relinquished my avatar he stole somebody else's. Very odd.
September 21st, 2010 at 8:30 AM ^
September 20th, 2010 at 7:35 PM ^
The defense will be much better in this game and we will execute a lot better offensively. I expect a 35-14 victory. GO BLUE
September 21st, 2010 at 12:09 AM ^
In the mgoblog Yahoo! Pick 'Em group, I'm killing The Sagarin Predictor (43-29 OTY)
Michigan is currently a 24.5 favorite over BGSU.
I'm taking Michigan.
I suggest you do the same. Or don't. Because this is basically flipping a coin. Or finding a coin on the ground. Specifically, a lucky penny. Heads up, of course. I flipped a coin and gave Michigan heads, and BGSU tails.
Therefore, Michigan will cover.
See how that all worked out.
September 21st, 2010 at 12:56 AM ^
I think the defense will come out angry and aggressive. BGSU will be held under 20 points. Our offense will score a ton of points. I'm predicting a major blow-out, UM 59 - BGSU 17. Lots of playing time in the 2nd half for Tate and Gardner, but they will be trying to score a lot of points...since they both want to play.
Comments