When I talk about bias, I mean the mostly unintentional favoritism that we humans as a species engage in in nearly all aspects of our lives. Even in a situation where you are 100% unfamiliar with the options, where you should be completely neutral, your biases will guide you. How about the new XFL teams? If you're not from one of the 8 cities hosting a team, and I asked you to pick a favorite, who would you choose and why? I like the Renegades and the Battlehawks for different reasons. The Renegades have a pretty cool name and logo, and I have a bias towards fighting authority and westerns. The Battlehawks have a great name even if it makes no sense, and I'm biased towards them because I used to play WoW as a hunter and my favorite pet was a flying snake thing and that's the picture that forms in my head when I read "Battlehawk" (I didn't say biases had to be rational).
The point is that there's always some basis for a bias. Sometimes that basis is rock solid, other times not so much. The thing is, unlike the @KirkHerstreit implication above, even if the basis for your bias is founded on fact, the bias still exists. That means that when a result that contradicts your bias occurs, you resist. And the bias influences your thinking when trying to make relative comparisons.
I say all that to get to this: you are probably wrong about your conference strength beliefs. I know I was.
Keep in mind that the following discussion is entirely internal. When I compare strengths across conferences I am using only my system's rankings. No bias, 100% transparent. No voters, no committees.
The perception currently is, and for quite a few years has been, that the SEC is the "best" conference in college football. Many SEC fans have stats memorized about not just simple stuff (number of nattys and such) but how many drafted, what rounds, what they're doing in the NFL. It's kinda scary at times.
I think we need a definition of "best." And since we're talking about college football, I think we need to ignore anything to do with the NFL: Ryan Leaf and Tom Brady are why.
So, what is the "best?" In nearly every instance, W-L records are a reliable predictor of future performance. In the pros, this is an easy metric to work with as there are many head to head games and best records get into the playoffs (though the NFL needs to fix their seeding issue). In college there are 130 D-I teams, and teams play 12 or 13 games, with some teams playing D-II teams and so only getting maybe 11 D-I games in. W-L doesn't get it done for college football.
The above is an example of 1) thinking with your bias firmly in place, and 2) moving the goalposts.
The point is that with 130 teams all over the country, we have few ways to measure teams against each other. The AP used to be useful with their voters, but even before instant info and spreadsheets they were extremely biased.
Now that we have the internet, cable TV, ESPN and Fox Sports, personal computers, and Bill James we can actually start to compare across teams who almost never play each other (each year CFB teams only play around 9% of all teams). W-L doesn't do it because of the lack of head to head matchups. Voters are biased. There is no transitive property of college football:
We will take the ranking system I have developed and say that the "best" conference is the conference with the highest average total score, summed from 1st, 2nd, and 3rd level points, and SRS numbers (see here for a breakdown). We use average because it allows us to normalize across conferences with varying numbers of teams. Then we will take those averages over the period from 2013-2019 and make a claim that is at least based in reality and math.
2013 (before bowl games)
Bet that surprised you. Wait until I release my No Season Left Behind for 2013! Anyway, here's what we're looking at:
- Rank - self explanatory, this is the strongest, best, whatever conference
- Conf
- Score - the average summed score
- % of - each conference compared to #1 as a percentage
This last one needs some explanation: let's assume two conferences tied in their score. One way to think of this is if one average team from each conference played each other, there would be no clear favorite and it becomes a coin toss. Now toss the coin 1,000 times. We should see 500/500 wins/losses. In 2013 the SEC was 96.4% as "strong" as the Pac12. If an average team from each conference played each other 1000 times, the SEC would be expected to win 96.4% of 500 times, or 482 games. The Pac12 should win 103.6% of 500 games, or 518 times.
Please note the sample sizes. If the Pac12 and SEC play each other in 3 bowl games, then the sample size is entirely too small to draw any sort of reasonable conclusion about conference strength. We need to look at the entire body of work.
2014
Less of a shock considering Oregon's run.
2015
OK, I swear I'm not manipulating the numbers. This is only weird if you ignore the FCS issue: The Pac-12 plays few if any FCS schools, and so they are not penalized.
2016
Feel better?
2017
Still kind of weird?
2018
SEC and B1G very close!
2019
Remember - all of these scores are before bowl or playoff games, mainly for ease of use - I can add them in later and probably will for a follow-up post.
When we look at the sequence chronologically, we get this:
When we look at the sequence chronologically, we get this:
| Conf | 2013 | 2014 | 2015 | 2016 | 2017 | 2018 | 2019 | AVG |
| Pac12 | 2.7646 | 3.0025 | 2.8922 | 2.6629 | 2.7807 | 2.5167 | 2.8244 | 2.7777 |
| SEC | 2.6642 | 2.8567 | 2.7944 | 2.8616 | 2.6352 | 2.7114 | 2.2258 | 2.6785 |
| B1G | 2.1906 | 2.6308 | 2.6655 | 2.7552 | 2.8836 | 2.7051 | 2.5081 | 2.6198 |
| Big12 | 2.4058 | 2.4974 | 2.576 | 2.3517 | 2.5728 | 2.6216 | 2.8256 | 2.5501 |
| ACC | 2.4324 | 2.7279 | 2.1945 | 2.8284 | 2.7673 | 2.255 | 2.1126 | 2.4740 |
| IND | 1.3621 | 1.1785 | 1.5266 | 1.0168 | 1.7065 | 1.8375 | 1.4224 | 1.4358 |
| AAC | 1.2669 | 1.082 | 1.6266 | 1.6609 | 1.4325 | 1.2116 | 1.6618 | 1.4203 |
| MTNW | 0.9556 | 1.383 | 1.1279 | 1.3926 | 1.2078 | 1.2313 | 1.6082 | 1.2723 |
| MAC | 0.9879 | 1.0498 | 1.2194 | 1.3159 | 1.1326 | 0.9726 | 1.138 | 1.1166 |
| C-USA | 0.9009 | 1.1966 | 0.9919 | 1.0935 | 1.1174 | 0.9335 | 0.9231 | 1.0224 |
| Sun Belt | 1.051 | 0.8347 | 0.91 | 1.0249 | 0.8275 | 1.0026 | 1.3794 | 1.0043 |
So on average over the past 7 seasons, the Pac 12 has been the strongest conference. No, I was not expecting that. But remember: we're dealing with 1) body of work, 2) sample size, 3) no points for playing FCS schools, and 4) no bowl games.
I expect reactions to be similar to this: "You're crazy if you think [my school's conference] isn't the bestest! [other school conference] sucks!" And if you only look at individual results when drawing this conclusion, you may be right. It's easy right now to point at LSU and say the SEC is the baddest of the bad. They did go 8-2 in bowls/playoffs. But if we examine that claim more closely, then we will find that of the ten SEC bowl games:
- They were expected to win 6 of them (they won 5), and
- Of the 4 they were expected to lose they won 3
- So they were 60% against expectations, but much better when they were supposed to lose.
Another aspect of this bias expresses itself in the cannibalism mantra: my conference is the best, so when we beat up on each other we shouldn't be punished as much as other teams. As if your pet conference is the only one that plays in-conference games. But you'll hear this from fans of every conference, usually followed up by "[Your conference] best team would be the 5th best team in [my conference]." They have no way to back up that claim other than to look at the rankings, find the top 4 teams and then see if the team they're attacking is ranked lower. Easy enough, but you know they don't do it. They "just know."
To wrap this all up: using a 100% unbiased and transparent method of comparing teams that do not play each other hardly ever (except within their own conferences), we assume that it is fair to take the score that this method generates for each team within the conference and average them to generate the conference score. The claim is that the "average" team from each conference can be expected to perform to that score. The higher the score, the better the average team and so the better the conference. Simple, yes?
No comments:
Post a Comment