Final Sagarin FWIW

[katss07]

Dude (lady?)... These are "ratings," not "rankings." You appear to be using the two words interchangeably. They are not interchangeable.

Learn what they are, what they represent, and how to interpret them before spouting off. There is actually a fair bit of value contained in these ratings if you balance them with some human judgment.

Sent from my Pixel XL using Tapatalk

Well, I’m sorry Lady (dude?), but these ratings suck.

[thebootfitter]

Well, I’m sorry Lady (dude?), but these ratings suck.

That's a fair opinion to hold if you try to make them mean something they are not. But the formula is relatively simple in principle (even if the specifics of it are complex). And once you understand the basics, you'll see that they don't "suck" at all.

They simply use actual scores from actual games between actual teams to develop a rating that is the statistical "best fit" based on those past results. When this rating is applied to future games, it is generally about 75-80% accurate in predicting W/L outcomes. With that level of accuracy, it's not really valid to say it "sucks."

Football is a game with many variables, so we wouldn't expect any rating system to be perfect. Results always fall somewhere within the deviations predicted by the variance in actual results. For context, the statistical standard deviation of results using the Sagarin ratings is generally around 14 points. That means that if two teams met on a neutral field with equal ratings, a win or loss by 14 points would still be within one standard deviation. And statistically, we generally expect roughly 2/3 of observations to fall within one standard deviation of the calculated means.

So even if a team that is a 28 point underdog according to Sagarin ends up winning, it's not valid to say that the ratings "suck." It just means that this happened to be an observation that was more than 2 standard deviations from the mean -- which is expected to happen a couple times out of a hundred.

If future predictions based on Sagarin ratings was less than 50%, or even under 60%, I think I'd listen to arguments suggesting that it "sucks." (And poor Jeff would probably be out of work.) But just because it makes implications you don't like, it is not valid to say that it "sucks."

(The lady comment wasn't a dis, BTW. Just a nod to some recent conversation where Ursus indicated there were plenty of female posters on here. Didn't want to presume either way.)

Sent from my Pixel XL using Tapatalk

[katss07]

That's a fair opinion to hold if you try to make them mean something they are not. But the formula is relatively simple in principle (even if the specifics of it are complex). And once you understand the basics, you'll see that they don't "suck" at all.

They simply use actual scores from actual games between actual teams to develop a rating that is the statistical "best fit" based on those past results. When this rating is applied to future games, it is generally about 75-80% accurate in predicting W/L outcomes. With that level of accuracy, it's not really valid to say it "sucks."

Football is a game with many variables, so we wouldn't expect any rating system to be perfect. Results always fall somewhere within the deviations predicted by the variance in actual results. For context, the statistical standard deviation of results using the Sagarin ratings is generally around 14 points. That means that if two teams met on a neutral field with equal ratings, a win or loss by 14 points would still be within one standard deviation. And statistically, we generally expect roughly 2/3 of observations to fall within one standard deviation of the calculated means.

So even if a team that is a 28 point underdog according to Sagarin ends up winning, it's not valid to say that the ratings "suck." It just means that this happened to be an observation that was more than 2 standard deviations from the mean -- which is expected to happen a couple times out of a hundred.

If future predictions based on Sagarin ratings was less than 50%, or even under 60%, I think I'd listen to arguments suggesting that it "sucks." (And poor Jeff would probably be out of work.) But just because it makes implications you don't like, it is not valid to say that it "sucks."

(The lady comment wasn't a dis, BTW. Just a nod to some recent conversation where Ursus indicated there were plenty of female posters on here. Didn't want to presume either way.)

Sent from my Pixel XL using Tapatalk

I have formed an opinion over the past few years. Sagarin is an okay predictor, but I think looking at the AGS poll and polls on Game Threads are better. Just my personal opinion. I favor the “eye test” over some computer. The ratings are decent predictors like I stated earlier, but I think the game polls are better. I might go back and compair the polls to the Sagarin prediction.

In my rant (I typed that all using the “Must protect SHSU football!” reaction), I was simply looking at the ratings. Is that fair? Maybe not. I get that Home Field would push SHSU up a bit, but if we are looking at this as a ranking (that’s basically what it is, a cpu ranking), then it puts Sam behind a majority of teams they shouldn’t be behind, such as Kennesaw, Jax St and USD. The Kats are a top 8 team at least. And that is why I find these rankings or ratings BS. Just my opinion.

Again, that was just my initial reaction. I haven’t seen those conversations with ursus. I apologize if I offended anyone.

[Daytripper]

Besides the top two, it's hot garbage.

[Go Green]

How can Princeton be 165 and Columbia 181?

http://www.ncaa.com/standings/football/fcs/ivy-league

[ElCid]

Without the algorythym he uses and the actual weighting it would be hard to calculate each teams deviation. But it would be easy for him. He could call it a confidence factor or whatever. Write a bit of code and he’s done.

Ok, but I was thinking a simple deviation from their predicted point spread for any given team. Anyone could do that.

[Noryan34]

Ok, but I was thinking a simple deviation from their predicted point spread for any given team. Anyone could do that.

That would be fairly easy if you had historical data (his rankings each week) throughout the year. Where it gets murky is the week to week weighting of results. But ya your right one could gather some reasonable data from each week to week performance and run some simple analysis.