FanNation - The Republic of Sport

Thanks for the input .. but my system just puts numbers on what I hope is based on common sense. As I hope is clear - most of Round 1 is based on simple W-L records. No "style points" and no point differentials are involved. Utah rates at the top because of their undefeated season. But ... their position might change in the next round. A hint of things to come: when you account for "SoS", things are gonna get pretty strange. This first round is straightforward, but if you're serious about adjusting for SoS, the rankings can get really interesting.

I think the system is still subjective, and when you have Rice at #20, the "improved system" is suspect. I do not believe for a second Utah could contend with Florida or Texas, and I am sure that is an assumption shared by most AP sports writers.

I'm interested to see what your future rounds will do, I think this is excellent work. "Putting numbers on common sense stuff" (paraphrasing - sorry) is not easy, and the old adage that common sense isn't common will rear it's head for certain down the road.

Good luck, will be watching to see what turns up.

QUOTE(#3):

Originally posted 10:15 AM ET  03.02 by gatorlonghorn

I think the system is still subjective, and when you have Rice at #20, the "improved system" is suspect. I do not believe for a second Utah could contend with Florida or Texas, and I am sure that is an assumption shared by most AP sports writers.

What do you think is "subjective" about my scheme? I think you may be using the term "subjective" incorrectly. The opposite of "subjective" is "objective" - i.e., based on fact, not opinion. My choices for how to combine terms and weight them are subjective, but once the choices are made, the calculations are objective. If you want different weighting, or a different combination, develop your own system and we can talk.

You're evidently misunderstanding what the 'accomplishment points' mean. Let me remind you of what I said at the end of post: "...accomplishment points will not necessarily produce accurate predictions of the outcome of any actual game."

Improved? Who made any claims about improvement and over what? I was curious to see what would happen and this is step one of a multi-step process. I'm NOT making any claims that this will be an infallible ranking system or a perfectly accurate predictor of game outcomes - any such interpretation of these results is simply a misinterpretation. It's based on a team's "body of work" during the 2008 CFB season.

I'm willing to entertain suggestions for changes and if time permits, I can explore what those changes might produce.

I couldn't care less about what AP sports writers think.

Keep developing your system, TB, and keep posting the results. I find this type of analysis intriguing, especially since there is no way in Hades I could even attempt it.

QUOTE(#4):

Originally posted 01:18 PM ET  03.02 by GatorRobster

I'm interested to see what your future rounds will do, I think this is excellent work. "Putting numbers on common sense stuff" (paraphrasing - sorry) is not easy, and the old adage that common sense isn't common will rear it's head for certain down the road.Good luck, will be watching to see what turns up.

Thanks, GR. Indeed, "putting numbers on common sense" is a fair description of what the scheme does. Evidently, there's no way to keep folks from misinterpreting the results, which I expected, anyway. I've got a system pretty well worked out for calculating SoS for each team using: the sum of accomplishment points for its opponents (relative to the average accomplishment points for ALL teams) and the percentage of road games on its schedule (I noticed some wide variation in the number of road games amongst the Div 1-A teams and I wanted to boost a team's SoS if it had more road games.

I think my choice of variables conforms to "common sense" and although my weightings are mostly a WAG, I hope most people would find them at least plausible.

QUOTE(#1):

Originally posted 11:51 PM ET  03.01 by tnerb57

Seems to me that your exhaustive study has shown that the current system is not quite the demon that we all make it to be. Perfect? By no means. I would still prefer a playoff, but outside of Utah and Texas, most of us could not argue with the results of the last year.

I should have responded to this earlier (I almost said "Sooner"!) ... I believe that the existing ranking system tries to do MORE than what I'm hoping to accomplish. The relatively good agreement between my Round-1 accomplishment points and the polls is explained by the significance of W-L records. My version of things will eventually result in accomplishment points based on the SoS, as well as the W-L record, but be prepared for some surprising things - my preliminaries suggest that - given the relative strength of the schedule - some teams are considerably better than their W-L record would suggest.

This shouldn't be used to infer that such teams are BETTER than others in a ranking sense, but rather indicates their positive accomplishments were achieved against a tough schedule, overall. I THINK this is what we hope to achieve with a scheme to estimate SoS.

A teaser in this vein - OU played a considerably tougher schedule (according to the Round-1 accomplishment points) than UTx, so the adjustment to their accomplishment points rates them ABOVE the 'horns. But then this implies that - since UTx won their head-to-head game - that victory is a bigger accomplishment than it might otherwise seem to be. So UTx's Round-2 SoS adjustment will need to be increased ...

I'm just following the numbers to see where they lead.

The hard thing to quantify here is how deep into a season do you need to go to get enough valid data? My guess is about 8 games at least.

A very good question. If I had nothing but spare time, it might be feasible to see how stable the results are with respect to the number of games played. Keep in mind that bowls and conference championships are typically against opponents of reasonable quality, whereas the opponent's strength during the regular season of 12 games is all over the map (but depends on what conference the team is in, of course).

Unfortunately, an analysis of the sensitivity of the results to the number of games played is well beyond my current capabilities. Unless I can write computer code to do all the analysis ... someday, maybe.

But what I'd like to see is how sensitive the results are to the various subjective choices I had to make in designing the scheme. Also beyond my present capabilities, sadly.