Oklahoma Sooners Football: Projections and player freedom

It’s the season for college football preseason predictions, and ESPN analytics guru Bill Connelly has dropped his latest SP+ projections for 2023. In news relevant to this particular audience, his computers spit out the Oklahoma Sooners as the No. 13 team in the country.

For the uninitiated, Connelly designed SP+ as a measure of efficiency that accounts for the impact of tempo and opponent strength. OU finished the 2022 season ranked 20th nationally, with an Offensive SP+ ranking of 10th and a Defensive SP+ ranking of 70th. The 2023 preseason rankings have OU at sixth on offense and 37th on defense.

If you based your level of confidence in OU’s 2023 projection strictly on last year’s numbers, skepticism would be warranted. SP+ had the Sooners at No. 4 overall heading into the season, up from eighth at the end of 2021. Looking at how the 2022 preseason numbers broke down, the projected gains were going to be fueled by a defense slated to improve dramatically from 68th in Defensive SP+ in 2021 to 21st in 2022.

Far cry from where OU ended up, of course. With the benefit of hindsight, though, we can see why SP+ had such an inflated opinion of OU’s defense at this time a year ago.

Connelly’s preseason projections encompass an amalgamation of three factors: returning production, recruiting and recent history. The four inputs for returning production and their individual weighting at the time included percentage of returning tackles (59%), percentage of returning passes defensed (28%), percentage of returning tackles for loss (8%) and percentage of returning sacks (5%).

*Note that while the inputs for returning production on defense have remained the same in the new round of projections, Connelly appears to have tweaked the weighting.

The returning production variable weighed heavily in OU’s favor in 2022. According to Connelly’s numbers, the Sooners had 78% of their production returning on defense, ranking 21st in the country.

Notably, Connelly’s approach to calculating returning production involved one-to-one additions of stats for transfers between FBS schools. In other words, he added the relevant stats from the prior season to both the numerator and the denominator of the metrics that go into returning production – a technique that he acknowledges makes for a crude solution to the problem. For example, if a player with 100 tackles the prior season transferred to a new team, that team would have 100 tackles added to the amount returning from the year before and 100 added to the total number of tackles from the prior season.

OU’s transfer players were nothing if not productive at their previous stops. As an example, linebacker T.D. Roof and defensive back C.J. Coldon each had more tackles in 2021 than OU’s returning leader, LB DaShaun White. In fact, five players who transferred in were among OU’s top 10 in returning tacklers:

*Transfers in italics.

You could draw similar parallels in the other categories. Coldon had 10 passes defensed (the sum of passes broken up and interceptions) at Wyoming in 2021, which was nearly half of the total of all of OU’s own returnees from the 2021 team at 21. Likewise, Roof (10) and Laulu (eight) each had more tackles for loss in 2021 than any of OU’s returning players.

But OU’s transfers ultimately did not impact – for lack of a better word – the defense in 2022 to the same degree as they did with their prior teams the year before. According to data from PFF, all saw their snap counts drop dramatically from one season to the next:

Ergo, all that added production likely led to an overly sunny forecast for OU’s defense in 2022. But we can’t blame forecasters for lacking perfect knowledge of the future.

The real lesson here has nothing to do with playing college football and everything to do with building mathematical models for it. Just as the transfer portal has opened up opportunities to swing teams’ fortunes dramatically, relaxing restrictions on player movement has created more volatility in data-driven projections, too. Until the stat heads come up with more effective solutions, tread lightly with forecasts for squads featuring heavy influxes of transfers.

***

Postscript:

If you’re wondering about the likely impact of the latest group of transfers on OU’s defense in the coming season, bet on at least three of them playing a more prominent role than any players from the 2022 crop did last year. Defensive end Rondell Bothroyd (Wake Forest), linebacker Dasan McCullough (Indiana) and defensive back Reggie Pearson (Texas Tech) all played at least 468 snaps for teams in Power Five conferences last season. The coaching staff assuredly added them to the roster with the intention that they would compete for spots as full-time starters.