Wolverine247 Board

Very, very interesting - thanks for sharing!

I like the direction you're headed in here. The one thing I'd say is that turnovers are often created by a team because they're better than the other team. So if a team is +5 on turnovers in a game, it's possible that that's due to luck, but there's also a good chance that it's because they're the better team.

That being said, no formula will ever be perfect, as there are holes in every rankings formula.

I completely agree. That's the problem - some element of turnovers are luck (as a Bills fan, I particularly think of Jairus Byrd's 9 INT's his rookie year, then 5 INT's the next 36 games).

That's why I'd give "bonuses" to teams that do tend to create a lot of turnovers. Some of it is luck, some isn't. e.g. forcing fumbles is usually not luck, but having the fumble bounce to one of your players is.

LSU created a lot of turnovers last year, which is one reason for their run. However, I think that helped them outplay their talent level - if they don't get those turnovers, they're obviously not quite as good. JMO.

No problem, Todd, I enjoy discussing it. And certainly enjoy feedback/questions.

The garbage time thing is certainly an issue. I was thinking of that myself. It would have to be manually included in some way (i.e. actually paying attention to what happened in the games, not just the stats). Possibly triggered when the 2nd-string QB enters the game? Possibly use a formula that accounts for scores with a certain amount of time left on the clock. For example, if Team A is up by 21 points in the 2nd quarter and score a TD, that obviously count because there's tons of time left in the game. However, if Team B is up by 21 points with 1:00 to go in the 4th quarter...should that really count? It's something I'd have to take a look at.

Regarding SOS...this is the beauty of the system. I don't believe SOS even needs to be included. If this kind of formula can work, the actual strength of the team itself would be calculated, rather than the strength of their opponents. So we wouldn't need to compare SOS's plus performance to try and gauge how good teams are...we would know outright that Team A is better than Team B (ideally, of course).

As to your example of beating the 3rd ranked team in the country...I'd try to use stats that are as up to date as possible. Kind of like final rankings. If Team X looks great now but proves to be terrible later, then it's just not going to be as good of a win no matter how you look at it. One issue that might also need to be addressed is injuries due to key players. If Robert Griffin had gotten hurt last year, would Baylor have been as good? No. But should that take away from teams who beat Baylor with Griffin in? I don't believe so. That's a human element that could be a problem, though not a ubiquitous one, thankfully.

In the end, what I really want to see is a formula that is inherently predictive in who will beat who, with the caveat that turnovers can provide the key swings. So if we know that ND is better than MSU and Boise is not as good as MSU, then a game between the two would favor ND if turnovers are equivalent.

I have thought about it, but putting it into an actual ranking is another thing.

It's true that including point differential encourages teams to run up the score. That's why the BCS disallowed it. Without having calculated anything...I think I'd say that running up the score MAY not make as much of a difference as people believe. In the end, comparing each team directly to their opponent is going to end up with a range of "strength" for each team. Increasing the point differential in one game out of 12 will affect the range, but by how much? Not sure at this point.

At the least, I'm hopeful it could clearly delineate tiers of teams (e.g. Tier 1 is elite teams, Tier 2 is teams that would need a positive turnover margin to defeat the Tier 1 teams, etc.). Sometimes ranking college football teams feels like ranking college football prospects. Nkemdiche is ranked #1, but he's clearly the top prospect. Similarly, Alabama is clearly the top team. So an integer scale (1, 2, 3, 4...) for team rankings doesn't show the things that the ratings do for college prospects (Nkemdiche being rated a 103 vs. a bunch of players clearly behind him being ranked 100 - still elite, but not on the level of Nkemdiche). My rankings would hopefully provide those ratings like 247 provides them for CFB prospects.

Back to your comment - if running up the score would end up being an issue (which, frankly, I'd say it's sometimes an issue now anyway), then I'd have to find some way to adjust for it. But it's one of those issues that shouldn't be an issue between teams of relatively equal strength. If you don't want them to run up the score, then stop them.

Regarding the highly ranked team tanking...it's possible we just can't account for it. My feeling is that that doesn't happen all that often, so it may just be an outlier. Maybe one way to account for it would be to incorporate "streak" bonuses (probably small ones). If you're playing a team that's on a winning streak of X games, then you get X times a certain number of points if you beat them. That's just off the top of my head.

Regarding the SOS issue with no common opponents...I think in the end it will tie together enough that comparisons can be made, even if it's through opponents of opponents. Like I said above in this post...I'm thinking each team will end up with a "range" for their strength that will narrow with more and more data (i.e. playing more games). The question I'm struggling with is if CFB plays enough games to truly narrow that range enough to make decisive conclusions.

Also is looking like I can't avoid ranking D1AA teams with this approach. Blech.

Haha it's on my list of "eventually" projects, so hopefully at some point. Working on a Ph.D isn't exactly conducive to spending time on these projects, unfortunately.

I gotcha. Please do keep us posted - very interesting stuff.

Todd - don't know if you saw the PSU game, but that introduced another problem. PSU up by 11 with about 40 second left to play. Game isn't in doubt, but NW moves the ball then fumbles. Before that, PSU had a -1 turnover margin, after that it was even. But it didn't change anything, really.

I think I'm going to definitely have to adjust for time left in a game when certain events occur (TD's, turnovers, etc.).

Todd, you're are a class act.

Thank you and ditto to Michigan. I liked Purdue this year, and Michigan did a terrific job on defense. Sure hope PSU can do that well against the Boilermakers.

Gotta remember something about being in the SEC. Doesn't matter if you don't beat anyone, being in the SEC is alone worth 5 spots in the polls

The ND loss is looking better and better every day, but I think one thing that hurts Michigan is how badly Alabama won that game. Especially with two losses, both games would have had to have been last-minute nailbiters in order for Michigan to be ahead of LSU at this point. Not saying I necessarily agree that LSU deserves to be ahead of Michigan right now - just explaining the common perception that's out there.

I'll be realistic too. There is a very slim chance LSU would crush us this year. They are not that good. For LsU to only drop 5 places is very suspect to me. To see Florida jump K-State, ND, and WV is also very confusing.

Wait another two weeks when Florida and South Carolina will have both lost and Mississippi State is moved to the 3 spot.