Over the past few weeks, I released my math-based projections for the 2018 college football season as well as a retrospective of the 2017 season. As a part of my retrospective analysis, I found a way to crunch the numbers is a new way where I was able to decouple any given team's win total into 4 contributions: the baseline preseason win prediction, the effect of actual (vs. projected) schedule, the effect of actual team ability (vs. projected), and the effect of luck / execution (which is essentially if the team won or lost close games). I then visualized this data for selected teams using a water fall chart, an example of which is shown here:
The first three teams of interest come from three different conferences: Purdue, Wake Forest, and Iowa State. All three teams were much better than expected last year. Ironically, according to my math, they were all in reality even better than we thought that they were, as all three teams had about one fewer win than expected due to luck and schedule. While each team did get a bit of a boost in the preseason rankings this year, it likely is not enough. Any of the three might be good enough to be in the Top 25 this year.
In the case of Iowa and Virginia Tech, both of those teams essentially performed as expected last year, but my data suggests that they both, in fact, were better than we thought by 1-2 games. Both teams had a significant luck / schedule penalty that offset their higher ability score. Both teams did see their preseason rankings rise over last year's value, but I wonder if it is enough. I projected both of these teams to be surprise divisional champs this year, and this analysis makes my feel a little better about those picks.
Indiana and Navy's profiles suggest that they are teams that also suffered significantly from poor luck and harder-than-expected schedules last year. Both teams appeared to underachieve in 2017, but their ability scores suggest that they were both better than we thought. Their 2018 preseason rankings are both in the 60s, but this seems a bit low to me on both counts.
Finally, Georgia Tech and South Florida also made my list as possible Hidden Treasures. In all honesty, both teams did appear to underachieve last year, winning at least 2 fewer games than expected. While some of this was due to lower than expected ability, the luck / schedule contribution was actually much higher. Both teams did take a hit in preseason rankings this year, but I suspect both teams will be just a little better than we think this year.
As I mentioned above, this is a pretty simplistic analysis that does not take into effect the very important factors of roster turn-over, recruiting, coaching, injuries, and player development. All of these factors could also explain some of the preseason ranking curiosities that I point out above. I am honestly not sure. But, I am certain curious to see how this will play out. In less than a month, we will start to have our answer.
As I have continued to think about this new set of data, I wondered if this data might reveal something about the upcoming 2018 season. For example, in the chart above on South Carolina, they began 2017 with fairly low expectations, and I pegged them to win only about 6 games based on their preseason ranking and schedule. In reality, they were a bit of a surprise and actually won 8 regular season games. I heard that they won their bowl game as well. However, the waterfall chart tells a slightly different story. Based on my power rankings, South Carolina had a negative "ability" score, which essentially means that they were actually a bit worse, in reality, than the preseason projections. But, their schedule was easier than expected and they were a bit lucky, which resulted in a boost of over 2 wins.
This got me to thinking. If my math is correct that South Carolina was, in fact, no better at the end of the 2017 season than we thought that they were at the beginning, then it might lull some of the prognosticators to rank them higher this off-season than they deserve. Sure enough, the Gamecocks consensus ranking rose from 40 in the summer of 2017 to 26 in the summer of 2018. Interesting. While it is true that a lot changes on a roster from year to year, I am very curious to watch South Carolina to see if they are for real, or if they are just Fool's Gold. Considering I picked the Gamecocks as the surprise winner of the SEC East, this is a relevant question for the entire college football landscape.
To investigate this effect further, I went through my numbers and looked for teams in two general categories. First, I looked for teams whose win totals in 2017 were boosted by at least 1.0 games due to a combination of "luck" and their schedule and whose preseason ranking improved by more than 10 slots over my 2017 post-season power rankings. Teams with this profile I would consider to be possible Fool's Gold.
On the flip side, I also considered teams whose win total was hindered by more than 1.0 game due to luck and schedule in 2017 and whose preseason ranking was depressed by more than 10 slots over my 2017 post-season power rankings. I would consider all of these teams to be possible Hidden Treasures. In total, I identified 19 total notable (mostly Power 5) teams in one of those two categories. Let's start with the Fool's Gold:
In general, I identified 10 total Power 5 teams that perhaps are over-valued this summer. Within this group, there are a few different situations. I already mentioned South Carolina, a team that in reality was no better than expected in 2017, but seems to have gotten a ratings boost for 2018 anyway. A similar and more mild version of this effect is apparent with Missouri and West Virginia. They both seemed to over-achieve a bit in 2017, but their actual ability was as expected. Nevertheless, both teams got a decent boost in ranking in 2018. Will they "regress to the mean" in 2018?
The 2nd category of Fool's Gold are for teams that actually were worse than expected in 2017, but through luck and schedule actually more-or-less met win total expectations. These two opposing factors effectively offset. In this category we have several PAC 12 teams including USC, Oregon, and Stanford, as well as Texas A&M. My post-season power rankings had all four teams well below their 2017 preseason placement, but in general they all popped right back up into almost the same preseason ranking in 2018. Time will tell if this is justified or not, but this would seem to be good news for Washington, who based on a similar analysis was a good as advertised last year, but just had a little more bad luck.
Arizona is also an interesting story. My math suggests that they actually were better than expected last year by over a full game, but they also did benefit by over a full game from schedule / luck. In this case, the preseason prognosticators seems to have gone a bit over-board as the Wildcats jumped all the way to 35 in the preseason magazines. I think this might be a bit optimistic.
Finally, there are the teams that definitely under-achieved in 2017, but that luck and schedule might have actually made things look a bit better than they actually were. UCLA and Tennessee fall into this category. While both teams were bad last year, my math suggests that they were actually worse than we thought by a game or two. Both teams were downgraded in the preseason this year, but my math suggests that maybe the prognosticators did not go far enough.
Now, let's turn to the teams that might be pleasant surprises this year. Where should we look for Hidden Treasure? The chart below shows the 9 teams that I think might surprise in 2018.
In the case of Iowa and Virginia Tech, both of those teams essentially performed as expected last year, but my data suggests that they both, in fact, were better than we thought by 1-2 games. Both teams had a significant luck / schedule penalty that offset their higher ability score. Both teams did see their preseason rankings rise over last year's value, but I wonder if it is enough. I projected both of these teams to be surprise divisional champs this year, and this analysis makes my feel a little better about those picks.
Indiana and Navy's profiles suggest that they are teams that also suffered significantly from poor luck and harder-than-expected schedules last year. Both teams appeared to underachieve in 2017, but their ability scores suggest that they were both better than we thought. Their 2018 preseason rankings are both in the 60s, but this seems a bit low to me on both counts.
Finally, Georgia Tech and South Florida also made my list as possible Hidden Treasures. In all honesty, both teams did appear to underachieve last year, winning at least 2 fewer games than expected. While some of this was due to lower than expected ability, the luck / schedule contribution was actually much higher. Both teams did take a hit in preseason rankings this year, but I suspect both teams will be just a little better than we think this year.
As I mentioned above, this is a pretty simplistic analysis that does not take into effect the very important factors of roster turn-over, recruiting, coaching, injuries, and player development. All of these factors could also explain some of the preseason ranking curiosities that I point out above. I am honestly not sure. But, I am certain curious to see how this will play out. In less than a month, we will start to have our answer.
Comments
Post a Comment