Comparing the Projection Systems

Some of you may prefer Christmas or Thanksgiving, but my favorite national holiday? When Kenpom releases its projections for the next season. It's never announced, it just sort of *happens*. And that happened this past weekend!


I'm still a couple of weeks away from releasing all of my pre-season content, but one of my favorite things to do is take a peak at what the projections systems are saying. We've obviously mentioned Kenpom, but the other website I use daily is BartTorvik.com. Bart is the creator of "T-Rank", and his projections are publicly available and pretty great.


A reminder: these are projections, not predictions. The data is immensely useful, but should not be taken as gospel; there are clear, qualifiable aspects to teams that neither projection system can know. Neither Ken Pomeroy nor Bart Torvik are predicting these outcomes, and are generally open about the blind spots of their systems. So don't bother yelling at either (or me!) on Twitter that they are too low on your favorite team.


Let's dig into some data!


Projected order of finish (in tiers)

I generally could not care less about projected pre-season records; there's a ton of rounding, and the difference between 10-8 and 9-9 is negligible. So instead, I prefer to use the average system rankings (based on Adjusted Efficiency Margin):



Using the average rankings above, breaking the league up into tiers is pretty simple:


Tier 1: Merrimack, Bryant, Wagner, Long Island

Tier 2: Mount St. Mary's, Sacred Heart, St. Francis (PA)

Tier 3: St. Francis (NY), Fairleigh Dickinson

Tier 4: Central Connecticut


Easy enough, right? See you on 11/9?


Where are the Disagreements?

No, not easy enough. Now we need to see where these systems diverge, and at least try to figure out why.


The first thing I noticed is; Bart Torvik's system is much more bullish on the NEC as a whole. The average NEC team sits 251st in Torvik's rankings, which is ~8 spots ahead of where it finished last season. Kenpom? An average ranking of 265, which is still ~14.5 spots better than where it finished last season. Huh? Essentially; Torvik's system likes the NEC more! So, I guess we like Bart more?


Anyway; instead of comparing a team's ranks at Torvik compared to Kenpom, it's better to compare their rankings at each site when compared to the average NEC team. When we do this, there are 4 teams whose rankings diverge significantly between systems.


Favored by Torvik


1) Long Island (66 spot difference when compared to average)

The Sharks were a very good defensive team last season, finishing 3rd in the NEC in Adjusted Defensive Efficiency. While both systems project LIU to be the 2nd best defensive team in the NEC this season, Torvik's system projects a much larger improvement (over 3 points/100 possessions), while Kenpom projects a smaller jump. Perhaps one explanation; when Derek Kellogg was at UMass, his teams were generally better on the defensive side of the ball. Maybe Torvik's system is taking that more into account. Another possibility? Torvik is more bullish on the addition of former Hofstra big man Isaac Kante, who is projected to be the 2nd most valuable player in the league behind teammate Eral Penn. Ultimately Torvik has LIU as a title contender, while Kenpom has them finishing more in the middle of the pack.


2) Mount St. Mary's (41)

The Mount should be one of the more fascinating teams in the NEC this season, so I guess it makes sense that the projection systems can't figure them out. Mount St. Mary's was the best defensive team in the league a season ago, and while both projection systems expect some regression in that department, they should have one of the better defenses yet again. It's the offense where there is a divergence; After being a below average offensive team last season, Kenpom projects the Mount to have the 2nd worst offense in the league, perhaps in large part due to the loss of Damian Chong Qui. Torvik isn't as concerned; it doesn't have them as world-beaters, but rather a slightly below average squad in terms of scoring efficiently. Again, the difference might be because of a talented transfer; Jalen Benjamin arrives from UAB with major credentials as a combo-guard. If Torvik is to be believed, Mount St. Mary's should be battling for a home quarterfinal game.


Favored by Kenpom:


1) St. Francis (PA) (60 spots)

What do you do with a team that underachieved a year ago, then brings back (almost) everyone and adds no one? Some would argue continuity is important, and that may be where Kenpom has arrived on St. Francis (PA). No team is projected to improve more on the offensive side of the ball by Ken Pomeroy's system than the Red Flash, going from the 3rd worst offense to the 3rd best, while the defense is also projected to improve. Torvik does see an improvement- both systems agree that SFU should be better in 21-22 than they were last season- but not to the same extent as Kenpom does. My guess? Kenpom is buying Rob Krimmel's history of putting teams on the floor that can score efficiently. Torvik, on the other hand, has concerns on the personnel, specifically the perceived lack of 3-point shooting.


2) Wagner (56)

Just about everyone who follows the NEC expects Wagner to be a title contender (and perhaps favorite), so what the hell, Bart? Torvik has the Seahawks finishing just 10-8 thanks to some serious regression on both sides of the ball, while Kenpom is buying what Bashir Mason was selling last season and has them as the best team in the Northeast Conference. Ultimately the disconnect is due to a defense that was only around average last season; Kenpom expects this defense to be more in-line with Mason's teams back in 2013-2018, when they were regularly one of the best in the league at preventing opponents from scoring. Torvik's system disagrees, and sees the defense regressing to below-average, possibly due to a lack of rim protection. That's a pretty fascinating difference that should be watched during the non-conference.


Who has improved according to the projections?


1) Merrimack

When you look at the projections as a whole, no team is expected to improve more than Merrimack; the Warriors have an average ranking of 191, which is 86 spots better than their final 20-21 average ranking of #277 (#267 at Torvik, #287 at Kenpom). Kenpom has Merrimack 3rd (#210), which seems bullish but reasonable. However, Torvik has Joe Gallo's group #1 in the NEC despite not adding a single experienced piece to last season's 9-9 team. Why? Because the system sees MC improving by 4.6 points/100 possessions on both offense and defense. That's 9.2 points/100 possessions! Umm...sure, maybe? However, I did hear directly from Bart regarding this seeming odd projection; "teams with less than 3 years' history could be particularly prone to anomalies". I have as much faith in Joe Gallo as anyone, but it's tough to reconcile that large of an improvement without a change in personnel, at least when the majority of NEC teams return most or all of their rotations. You can be high on Merrimack and still take the under on Torvik's projected 12-6 NEC record.


2) Long Island

While Merrimack is the clear leader in projected improvement, Long Island has an average pre-season ranking of #220 when compared to last season's #265. We've already touched upon the Sharks, but one question I'm asking; does the addition of Isaac Kante and Tyree Pickron (Quinnipiac) outweigh the losses of Jermaine Jackson, Jr. and Virshon Cotton? Maybe. But it's also worth noting that Torvik still has former CCSU player Karrington Wallace included in the projections, and as of this writing he is not included on the LIU roster. That wouldn't make a huge difference, but a difference nonetheless.


3) Sacred Heart

The anti-LIU, in that both projections systems are in agreement, is Sacred Heart; Torvik has the Pios #269 (6th in the NEC), while Kenpom has them #261 (also 6th), both of which are an improvement from last season's final average rankings of #297/#324. That probably shouldn't be surprising given that Anthony Latina returns his top 8 in minutes and adds Zach Pfaffenberger, who missed last season with an injury. Remember; SHU overachieved last season, finishing 9-7 in the league despite -2.9 Efficiency Margin (8th in the NEC). However, with Aaron Clarke healthy and a group of sophomores ready to make a jump, plus Latina's history of developing talent, these projections pass the "sniff test".


Who has gotten worse, according to the projections?


1) Fairleigh Dickinson

Objectively, no team was hit by the transfer bug worse than Fairleigh Dickinson; Jahlil Jenkins is at Stony Brook and Elyjah Williams went home to Northwestern, while rotation players Brandon Powell and Callum Baker are also gone. So it's not surprising that the projection systems see the Knights as the largest "droppers", going from an average ranking of 284 at the end of last season to #318 in the pre-season, with both systems picking FDU 9th in the NEC. But let's remember; as long as Greg Herenda is still roaming the sidelines, the Knights will be a tough out come February.


2) Mount St. Mary's

The Mount lost Damian Chong Qui and added Jalen Benjamin. Torvik sees them as interchangeable, at least from a value standpoint (Chong Qui was worth 2.8 PORPAGATU!, while Benjamin is projected to be worth 2.7). However, Kenpom sees the offense taking a step-back without DCQ, and the defense experiencing quite a bit of regression; in Engelstad's first two seasons in Emmitsburg, they finished 282nd and 283rd at Kenpom in Adjusted Defensive Efficiency, then last season they were at 150th. The pre-season projection has them at 233rd; effectively splitting the difference. My brain tells me that's wrong, but perhaps that's 'recency bias' at play.


3) Bryant

The third team which has a pre-season ranking significantly lower than it's final 20-21 ranking is Bryant. Wait...what? Both projection systems have the Bulldogs between 19-24 spots below where they finished last season, and that makes sense for a few reasons: 1) Historically, at Kenpom, Bryant has finished better than last season's #185 exactly once, back in 2013 (#176). It's called regression to the mean; 2) Michael Green III is gone. Last season Torvik saw MG3 as the 21st most valuable player in the NEC, while it projects none of the incoming transfers to be in the top 50 in the league. Jared Grasso may believe the Bulldogs got better when Green left, but the projection systems disagree. Of course, Grasso knows his team better than the computers, right?; and 3) Generally the best NEC teams finish somewhere between #180 and #210 at Kenpom. So Wagner and Bryant, at #192 and #204, respectively, are right in line with your average NEC title contender. Nothing to see here, folks.


They are who they are


St. Francis (NY)

No team lost more than St. Francis-Brooklyn; the Terriers lost Chauncey Hawkins to the Transfer Portal (and, eventually, Bryant) in September after graduating 4 rotation players off last season's team. Torvik (#282) is more bullish than Kenpom (#323), and that's easily explainable; Torvik likes their 4 D1 transfers more than Kenpom does, especially Michael Cubbage (Marist). In my opinion, the difference in projections accurately shows the range of outcomes for Glenn Braica's team; if the newcomers fit in well, a .500 season is within reach. If not? They could be battling for an NEC playoff berth.


Central Connecticut St.

Essentially, the systems see the Blue Devils as they've been; a basement dwelling NEC program, with an average pre-season ranking of 343 (out of 358 D1 teams). In fact, Central hasn't finished better than #305 at Kenpom since back in 2013 (shouts to Joe Efese!). The fact that new coach Pat Sellers only has five experienced D1 players helps further explain why the Blue Devils' projected NEC record of 4-14 is two games behind the next worst team. Better days ahead, hopefully.


The Players

We don't get to peak under the hood to see the Kenpom engine, but over at Torvik we can see his player projections. So without further ado, here is Torvik's pre-season All-NEC 1st, 2nd, and 3rd teams, along with an "All-Newcomer team", based on projected PORPAGATU!. Note: Torvik assumes all freshmen at this level are essentially the same.


1st Team: Eral Penn (LIU), Peter Kiss (Bryant), Elijah Ford (Wagner), Alex Morales (Wagner), Isaac Kante (LIU)

2nd Team: Brandon Rush (FDU), Ty Flowers (LIU), Tyler Thomas (SHU), Mike Sixsmith (SHU), Charles Pride (Bryant)

3rd Team: Devon Dunn (FDU), Jalen Benjamin (Mount), Ramiir Dixon-Conover (SFU), Aaron Clarke (SHU), Hall Elisias (Bryant)

Newcomer: Jalen Benjamin (Mount via UAB), Michael Cubbage (SFC via Marist), Bahaide Haidara (SFC via George Mason), Jahbril Price-Noel (Wagner via Pacific), Patrick Emilien (SFC via Western Michigan)