Editor’s Note: Had a small bug in the original code which caused the simulations to include 18 teams instead of 17. Numbers have been updated.

As soon as the calendar flips from December to January, the same thought always occurs to me: “hey, it’s just one month until lacrosse starts.” Perhaps it’s the holiday chaos, but this thought never bubbles up in my brain until the new year starts. And yet again, I flip the calendar, excited to pursue new LacrosseReference projects. There will come a year when I no longer feel that way, but it is not going to be 2020.

And what of the teams, the players, the coaches? I imagine they are feeling the same excitement, although given the 52-week a year commitment that is college sports, I suspect the realization did not sneak up on them in the same way it did for me.

What’s more, we have two new division I programs this year (LIU and Merrimack) to add some additional spice to the stew that is men’s college lacrosse. Two programs, both with their virgin 1500 LaxELO ratings and (at the time of this writing) blank team pages. The tender green shoots of a new lacrosse season are just about to poke themselves up out of the winter ground. (Except in upstate New York, New England, and lacrosse’s Midwest outposts; please allow an additional 4-6 weeks.)

Yes, this season is full of possibilities. Rivalries renewed, goals scored, games won, players developed, postseason glory to be written. Every team is 0-0, in control of their own destiny. Right? 

Pre-Season Probabilities

Not so fast…

To test out this idea, I went back to the simulation model that I built last spring for our Bracketology feature. If you recall, the goal of the model was to run through the remaining games thousands of times and see how often various things came to pass. After a few assumptions, we could project how often each team would make the NCAA tournament, win their conference title, and receive a first-round home game. Since the committee seemed to use selection criteria that closely matched my assumptions (namely RPI), the model nailed every at-large selection.

Anyway, now that I’ve got the model, there is no reason to run it on the remainder of the 2020 season’s games. Of course, since the season hasn’t started, that is all the games, but the simulator works the same either way. So then, it’s just a matter of collecting the schedules for each team, telling the model to run through a significant number of season simulations and sitting back with a cold beer to wait.

Like the computer in Hitchhikers’ Guide to the Galaxy, after what seemed like an interminably long time, a number was produced. Well, a large set of numbers actually. And the numbers told a very interesting story.

The Starting Line

Here is every team, along with the results of the simulations. For each team, we have the average RPI, Strength-of-Record, and Strength of Schedule; this is an average of the results of each individual simulation. We also show the proportion of the simulations in which each team was selected for the NCAA tournament and the breakdown of those selections between auto-bids and at-large selections (I’m counting the ACC as an auto-bid even though I know it is not technically one). Note that the simulations take into account the fact that each team’s starting Lax-ELO rating is shifted toward the average to start each new season.

We saw last season that the selection committee largely went by RPI in choosing at-large teams. As a result, Cornell, who had an impressive body of work, was left out and Maryland, who had a higher RPI, ended up dancing. Because of the importance of RPI, the list of opponents and the conference each team plays in is as (some might argue more) important as the results on the field.

One thing to keep an eye on in the table above is instances where a team’s RPI and SOR ranking differ. Because RPI weights your schedule strength more than SOR (which is more concerned with the difficulty of achieving a given record), if the RPI ranking is better (i.e. lower), that indicates the team will benefit from their opponents’ strength and their opponents’ opponents’ strength. If SOR is better, then regardless of the quality of the resume, RPI is going to dock the team a few spots.

Because we think SOR is a better reflection of a team’s resume than RPI, better RPI rankings than SOR rankings in the table indicate a team that we think could get short-shrift come Selection Sunday.

Digging into the individual conferences, I’ve also included the average wins and losses projected by the Lax-ELO model. Here are the specific results of the simulations for each ACC team.

The model expects Duke and Virginia to be the cream of the ACC crop this year. Aside from giving Duke a slight edge in Strength-of-Schedule, the model expects them to have very comparable resumes when all is said and done.

The ACC gets interesting in the lower reaches, where Syracuse, Notre Dame and North Carolina collectively account for a little over 1.2 tournament slots. Syracuse and Notre Dame specifically provide some insight into the interplay between RPI, SOS and SOR. As noted many times on this site, Strength-of-Record (SOR) does what RPI does (combine opponent strength and wins/losses) in a more nuanced way that doesn’t rely as heavily on the opponent’s opponent’s strength. RPI expects Syracuse to be 3 slots (15 vs 18) better than ND this year. SOR thinks the gap will be six slots. As mentioned above, this would suggest that come selection time, Syracuse could be at a disadvantage relative to the Irish.

This is largely the result of the Irish’s strength-of-schedule, which projects as 3rd nationally. Not that Syracuse’s schedule is a slouch (10th nationally), but it’s clear that RPI likes the Irish’s schedule better than it does Syracuse’s. As a result, expect Notre Dame to have an edge come selection time if they are on the bubble or up for a home-game against a comparable Syracuse team.

And here we’ve got the results for each Big East competitor.

The Big East actually provides an even starker example of this principle. Take a look at Denver vs Georgetown. The Hoyas project with nearly 2 more wins then the Pios. Their SOR (12 vs 15) is three spots better. They have a 46% chance to win the conference tournament vs 36% for Denver. Lax-ELO, which was quite successful as a prediction tool last season, thinks that Georgetown is going to be the best team in the conference.

But Denver’s chances to make the NCAA tournament are a full 6 percentage points higher than Georgetown’s. Why? RPI.

I can’t think of a better example of the difference between RPI and SOR than this. Yes, Georgetown’s schedule is not nearly as strong as Denver’s. Removing common opponents, Georgetown plays Lafayette, UMBC, Fairfield, Mount St Marys, Bellarmine, Loyola MD, and Utah. Denver plays Delaware, Air Force, Duke, St. Bonaventure, Cleveland State, Notre Dame, and Ohio State. Clearly, Denver has the better schedule.

But the fact that Denver’s RPI is 4 spots ahead of Georgetown, despite the Hoyas being projected as the better team, is the result of RPI’s reliance on opponent’s opponent’s records. By playing teams from the ACC and Big Ten, Denver assures themselves that the 3rd factor of RPI (opponent’s opponents) is a huge plus for them. Georgetown, by scheduling the way they did, assures themselves that RPI is going to ding them. And remember, opponents’ opponents’ strength is 25% of the RPI calculation.

The question for those interested in tournament selection philosophy is: do we want Georgetown to get docked so severely for scheduling lower-tier teams? Or, put another way, do we want to reward Denver for simply scheduling so many top-conference teams, regardless of how they fare in those games?

Frankly, in a Georgetown vs Denver debate, I’m not sure it’s as critical an issue. They are both Big East teams with a robust lacrosse pedigree. If Georgetown had wanted a stronger schedule, they probably could have gone out and gotten one.

But consider the way that this affects lower-tier teams who may not be able to simply build a stronger schedule.

With that, we turn to the SoCon.

Not that there was a lot of suspense with last year’s selections, but High Point was a team that generated some controversy. If you recall, the Panthers had some early season upset wins over Virginia and Duke (both on the road). High Point also had a bad loss to St. John’s and they lost in the SoCon tournament. They made the full journey from pre-season also-ran to early season darling, to potentially scary tournament team, to spectator.

The St. John’s loss was probably disqualifying for them as an at-large, but by digging into their resume, it really highlighted for me the challenges faced by a strong program from a mid- or low-tier conference. This year’s crop of SoCon teams face the same situation. To see why, just compare Notre Dame and Richmond.

If we exclude non-conference opponents, their schedules look like this. Notre Dame has a game against Duke, North Carolina, Virginia, Syracuse plus potentially 3 more tournament games against that group. Richmond’s conference slate will include Air Force, High Point, Jacksonville, Bellarmine, Furman, VMI, and Mercer, plus any tournament games as well. The Spiders’ non-conference schedule includes Notre Dame, Virginia, Duke, Loyola and Maryland; a veritable murderer’s row.

The difference here is that Notre Dame’s conference slate allows them to boost their resume by scheduling tough non-conference teams. Richmond’s conference slate means that no matter what they do in the non-conference, the best they can hope for is a middling SOS and consequently, RPI.

The simulation results bear that out. Notre Dame’s schedule is projected to be the 3rd toughest in the nation; Richmond’s is 38th. Notre Dame projects a record of 7-6. Richmond projects a record of 11-5. Of course, according to Strength-of-Record, which we think is the best measure of a team’s quality, they are equivalent (20th nationally). But the key, if the selection committee uses the same approach as last year, is that Richmond’s RPI falls 5 slots behind Notre Dame’s.

As a result, Notre Dame gets an at-large in 35% of our simulations. In the instances where they do not win their conference title, Richmond is selected in just 8.3% of the simulations. I think that gets to the heart of it; having such a big RPI boost from a top-tier conference slate gives those teams a much much larger margin for error than the teams from lower-tier conferences.

Why does this matter?

The “so what” of all this is open for debate. As I mentioned, for a top-tier conference team, a poor schedule could be a reason to dock them come tournament time. Or, it could be a somewhat altruistic attempt to get some exposure for the teams at the bottom of the lacrosse barrel. On the other end of the spectrum, if you are an up and coming team, you should probably just focus on winning your conference tournament at the expense of all else because that is the only way you are getting into the NCAAs.

As it stands right now though, your schedule is your destiny to a large extent. Even the herculean attempts of a lower-conference team to play difficult opponents gets diluted by the conference slate they are forced to play. As long as RPI is the main measure by which the committee selects their at-large teams, it’s going to be harder for teams to piece together the season-by-season progress needed to launch themselves into the upper echelon of lacrosse.

And the simple reason for that is a team’s progress in Division I lacrosse is (at least in my view) tied to advancing to the quarterfinals and Championship Weekend of the tournament. If you aren’t playing then, you aren’t going to really be part of the conversation when the most eyes are tuned in. But if a lower-tier team’s chances of making the tournament are so tied to the more random outcomes of their conference tournament than to the full season’s resume, they are going to be less likely to be in that position year after year.

A system that is stagnant is not good. We want teams cycling in and out of the top echelon. Schumpeter’s creative destruction needs to be let loose on college lacrosse!

A message to the committee

Fortunately, there is a simple solution to the issue we’ve discussed here: stop relying so heavily on RPI to select your at-large teams.

There is nothing inherently wrong with the RPI formula. The problem becomes when RPI is the end-all-be-all for at-large selections. I’m not saying that the committee just gets the RPI rankings, draws a line under the top x teams and then hits the golf course. But that’s how it worked out last year. If that is how it’s going to be going forward, then I think there is a problem.

And I’m not saying that they should just use something like SOR blindly either. But if they are going to use a single metric, I’d rather it be SOR than RPI. Select teams on the basis of how impressive their season was, not how impressive their schedule was.

And with that, happy lacrosse season 2020.