NPI Rankings 2024

[wihsuafs24]

There are a lot of teams in the top 50 that do not have a minimum of 10 wins.  What does that mean if they do not ultimately get the 10 wins?  Do these teams become ineligible for the tournament? Apologies if this is answered earlier in the thread.

No.  My understanding is that the 10 win minimum is to be eligible to drop "bad wins" for SoS purposes (you can't drop losses regardless), not to qualify for the tournament itself.  If you scheduled all of your games against the worst teams in DIII, but one game against a good team, you could theoretically drop all the games except that one "good win" and go in as an undefeated team with a good SoS.  This requires you to get to 10 wins before you start dropping anything else. 

This is what Coach Bianco of Denison tweeted about this in a thread earlier this year

https://x.com/BMBianco/status/1799429004522397850

This is very poorly written. It's not a "min win", but as I understand it, the min. # of games that must be counted in the algorithm.

Yes, all loses count as I understand it. So if you play a "bad" team and lose, you'll be punished. What this wants to do (right or wrong) is not punish you for beating a "bad" team. In their words, "a game you should win". This counts for league and non-league opponents

Thanks for clarifying. Who decides which games to drop? I would think everyone would just drop what they could to achieve the best SoS (or the algorithm would optimize somehow)...? 

[Kuiper]

There are a lot of teams in the top 50 that do not have a minimum of 10 wins.  What does that mean if they do not ultimately get the 10 wins?  Do these teams become ineligible for the tournament? Apologies if this is answered earlier in the thread.

No.  My understanding is that the 10 win minimum is to be eligible to drop "bad wins" for SoS purposes (you can't drop losses regardless), not to qualify for the tournament itself.  If you scheduled all of your games against the worst teams in DIII, but one game against a good team, you could theoretically drop all the games except that one "good win" and go in as an undefeated team with a good SoS.  This requires you to get to 10 wins before you start dropping anything else. 

This is what Coach Bianco of Denison tweeted about this in a thread earlier this year

https://x.com/BMBianco/status/1799429004522397850

This is very poorly written. It's not a "min win", but as I understand it, the min. # of games that must be counted in the algorithm.

Yes, all loses count as I understand it. So if you play a "bad" team and lose, you'll be punished. What this wants to do (right or wrong) is not punish you for beating a "bad" team. In their words, "a game you should win". This counts for league and non-league opponents

Thanks for clarifying. Who decides which games to drop? I would think everyone would just drop what they could to achieve the best SoS (or the algorithm would optimize somehow)...?

I'm sure the algorithm just does it automatically to maximize a team's ranking. It's not like the win is dropped for any purpose other than NPI, which is only used for selecting teams for Pool C bids to the NCAA tournament.

[Freddyfud]

Presumably the bad win exclusion is the reason for the new Adjusted Win Loss column which I don't think appeared before. In this column it appears some teams are already being capped at the 10 win mark (MWU for example.)

[rdanie03]

Here's a link to the NPI rankings through 10/20

https://stats.ncaa.org/selection_rankings/nitty_gritties/39652?utf8=%E2%9C%93&commit=Submit

Tufts jumping above Mary Washington is an example of how NPI differs from human rankings.  My guess is UMW does not drop in the USC poll in a week where it wins both games to remain undefeated, especially when one of the wins is against Christopher Newport and Tufts ties Middlebury at home (NPI doesn't distinguish between home and away games).

I disagree, mainly because I think that with both being undefeated but Tufts having probably(at least in the NCAA algorithim's eye) the best win in 2024 on the road at #4 and still being undefeated, they would probably go to #1

[Kuiper]

Here's a link to the NPI rankings through 10/20

https://stats.ncaa.org/selection_rankings/nitty_gritties/39652?utf8=%E2%9C%93&commit=Submit

Tufts jumping above Mary Washington is an example of how NPI differs from human rankings.  My guess is UMW does not drop in the USC poll in a week where it wins both games to remain undefeated, especially when one of the wins is against Christopher Newport and Tufts ties Middlebury at home (NPI doesn't distinguish between home and away games).

I disagree, mainly because I think that with both being undefeated but Tufts having probably(at least in the NCAA algorithim's eye) the best win in 2024 on the road at #4 and still being undefeated, they would probably go to #1

You may be right, but I was comparing the NPI decision to make that change with what might happen under the human rankings.  We know that the NPI doesn't take Tufts away win into account because the men's soccer committee chose to leave that Home v. Away "dial" set at 0/0. but we have no knowledge of what the coaches think about home versus away.  My speculation was based on the notion that I haven't seen the coaches bump down a #1 team that continued to win for a team that had a win and a tie.  I tend to think there is an anchoring bias in favor of keeping a team in that spot if it didn't do anything to lose it.  Could be wrong, though, but that was the reason to go to NPI - it theoretically reduces the guesswork/predictions or at least it makes it clear the factors it considers.

One factor that a human committee might consider in elevating Tufts over UMW that the NPI would not consider is goal difference in a game.  A ranking committee might think UMW's 1-0 win over Stevenson on a goal in the 89th minute was "too close" for a #1 team against a 3-4-4 Stevenson team even though UMW outshot Stevenson 33-4.  For the NPI, however, UMW's SoS drops playing Stevenson (NPI #214 last week) whether it beat them 1-0 or 100-0.

That was really my only point.  NPI makes that move because of SoS - since that matters even more than wins versus ties in this instance.  A human committee (or group of voters) could consider a bunch of factors that we might agree are relevant, but the NPI can't deviate from its algorithm.

Incidentally, it's useful to remember that the NPI shifts are typically momentary when you look at the rankings week-by-week.  For example, Tufts will be hurt next week by playing Husson (NPI #182) regardless of whether it crushes Husson or squeaks by them on an own goal in the 90th minute simply because it will drag down its SoS (although they will most likely be dropped at season's end and the same would happen for Stevenson for UMW).  UMW also plays a lower-ranked team next week, but N.C. Wesleyan is ranked #58 in the current NPI, so that isn't quite as big a hit.  And, of course, they can both be helped or hurt by how their prior opponents do because SoS and QWB are both ultimately determined at the end of the season, not by these interim numbers.

[ziggy]

Because Tufts has reached the minimum wins dial they will not see their NPI lowered because of a win over Husson. If adding that game in would lower their NPI it would be dropped immediately.

Perhaps a helpful way to think of it is up until the minimum wins dial, NPI treats every game as "need more information". Once that is reached (which it is in the case of Tufts), NPI says "you are this score and if beating a team you should beat would lower your score it will not count against you".

On the NCAA NPI report you'll notice two important columns: Win Value and Loss Value. These show the game npi score a team would get for beating or losing to that team in the current iteration of NPI. So, in this NPI report, every team who has beaten Husson has a game NPI of 57.538 factoring into their overall NPI. This would obviously lower Tufts' NPI, which currently sits at 61.778, and so it immediately falls below the cutline for games that get factored in because they have met the minimum wins dial.

If Tufts does drop after beating Husson it will be because of the overall shift of the entire interconnected national picture - not directly because they played Husson.

[UPDATE: Actually, this may not be totally correct. It's possible it would count, but in a way that bumps out another of their games that is currently counted. Just quick glancing at who they have played there are opponents who rank below Husson. If any of them are currently counting, beating Husson may lead to a direct increase in Tufts' NPI because beating Husson is better than, say, beating Trinity (CT)]

[Kuiper]

Because Tufts has reached the minimum wins dial they will not see their NPI lowered because of a win over Husson. If adding that game in would lower their NPI it would be dropped immediately.

Perhaps a helpful way to think of it is up until the minimum wins dial, NPI treats every game as "need more information". Once that is reached (which it is in the case of Tufts), NPI says "you are this score and if beating a team you should beat would lower your score it will not count against you".

On the NCAA NPI report you'll notice two important columns: Win Value and Loss Value. These show the game npi score a team would get for beating or losing to that team in the current iteration of NPI. So, in this NPI report, every team who has beaten Husson has a game NPI of 57.538 factoring into their overall NPI. This would obviously lower Tufts' NPI, which currently sits at 61.778, and so it immediately falls below the cutline for games that get factored in because they have met the minimum wins dial.

If Tufts does drop after beating Husson it will be because of the overall shift of the entire interconnected national picture - not directly because they played Husson.

[UPDATE: Actually, this may not be totally correct. It's possible it would count, but in a way that bumps out another of their games that is currently counted. Just quick glancing at who they have played there are opponents who rank below Husson. If any of them are currently counting, beating Husson may lead to a direct increase in Tufts' NPI because beating Husson is better than, say, beating Trinity (CT)]

Yes.  That's correct that it wouldn't affect this week's rankings.  You'll notice I mentioned that (although not in the detail you went into) in the parenthetical I added after the statement you focus on

(although they will most likely be dropped at season's end and the same would happen for Stevenson for UMW)

It's really a season's end determination of whether a particular team's game will be dropped or kept in the 10 that are used for tournament selection beyond auto bids, so we can't say a week ago, when I posted that comment, which 10 games will "count" in the final ranking that matters other than to say the highest NPI wins will count.  The more "bad wins" you have, though, the more likely you will have to include a bad win in your 10, which would lower your overall NPI.  That's really the effect of playing Husson, rather than a better non-conference opponent.  The Men's Soccer Committee presumably selected a 10 win standard to incentivize teams to schedule stronger non-conference opponents.

[Kuiper]

Mary Washington jumps back above Tufts in this week's rankings (which, as mentioned above, is just for fun because of weekly fluctuations)

https://stats.ncaa.org/selection_rankings/nitty_gritties/39798?utf8=%E2%9C%93&commit=Submit

Among the top 15, Williams drops a couple of spots from 6 to 8, Kenyon rises from 12 to 9, Babson drops from 9 to 12, Trinity TX rises from 15 to 13 (after tying Colorado College), North Park drops from 13 to 14, and Johns Hopkins rises from 17 to 15.

There's naturally more volatility the lower you go in the rankings.  A few big risers are Lynchburg, from 43 to 25, and Redlands from 37 to 24.

[eaglesoccerdad]

Other big moves
W&L up from 62 to 42
Montclair drops 21 to 37
Messiah 70 to 50
CMS 23 to 33
John Carroll 40 to 28
Macalester 41 to 29
OWU 30 to 19

[stlawus]

So ties are being included in the 10 win threshold then, in terms of being 0.5 wins for each contest?  I'm assuming that's what the adjusted win/loss column indicates. 

As everyone else said, these rankings are predictably volatile given such small sample sizes.  Vassar drops from 24 to 43 and SLU goes from 104 to 74.

[SKUD]

It is funny to see teams win two games in a week and drop because the quality of opponent.

[Gregory Sager]

It is funny to see teams win two games in a week and drop because the quality of opponent.

Heck, North Park went on the road and beat a team with a winning record (North Central), and yet the Vikings still dropped a spot in the national NPI ranking. Of course, NPU remained atop the regional NPI ranking, which is the ranking of primary importance.

[Kuiper]

So ties are being included in the 10 win threshold then, in terms of being 0.5 wins for each contest?  I'm assuming that's what the adjusted win/loss column indicates. 

As everyone else said, these rankings are predictably volatile given such small sample sizes.  Vassar drops from 24 to 43 and SLU goes from 104 to 74.

Yes.  Ties are considered half a win and half a loss for purposes of the Adj/W-L column.

[Kuiper]

It is funny to see teams win two games in a week and drop because the quality of opponent.

Heck, North Park went on the road and beat a team with a winning record (North Central), and yet the Vikings still dropped a spot in the national NPI ranking. Of course, NPU remained atop the regional NPI ranking, which is the ranking of primary importance.

Is it?  I thought the regional rankings are irrelevant now, with the selection of Pool C bids strictly based upon numerical order of NPI scores after the AQs are removed.  Perhaps I missed something though.

[Gregory Sager]

It is funny to see teams win two games in a week and drop because the quality of opponent.

Heck, North Park went on the road and beat a team with a winning record (North Central), and yet the Vikings still dropped a spot in the national NPI ranking. Of course, NPU remained atop the regional NPI ranking, which is the ranking of primary importance.

Is it?  I thought the regional rankings are irrelevant now, with the selection of Pool C bids strictly based upon numerical order of NPI scores after the AQs are removed.  Perhaps I missed something though.

That's imprecise wording on my part. Sorry about that. I was speaking strictly of NPU, which is in good shape for a Pool C if needed at #14, just as it was at #13. The "primary importance" to which I referred concerns hosting privileges for the first weekend, which is a pretty big topic for discussion in the North Park athletic department right now (in part because it would greatly affect departmental scheduling and resources in light of the start of men's and women's basketball seasons).

I've loved every chance I've had over the past decade to call D3 tournament soccer games, and I'm eager to get another opportunity to do it.