D3 Men’s VB Ratings Under the Microscope

Last week, I mentioned looking in to see how the other D3 Ratings Models “cook.” I take note of the leverage wins and losses from previous weeks exert on their respective systems. How much of a “hit” does St. John Fisher take when Santa Cruz reverse sweeps them (Inside Hitter #3 to #8, Massey #8 to #9, T100 #7 to #10) as compared to how the human element sees it? (AVCA from #5 to #7)  Or what about when Springfield wins a close one at home against Vassar? (Inside Hitter #7 to #5, Massey #7 to #2, T100 #5 to #3) And then comparing these to the AVCA Poll moving Coach Sullivan’s squad from #7 to #4. 

This exercise allows for some observations regarding how responsive each process is as it bakes in these outcomes.  For example, I find it interesting Massey moves Fisher down a single spot simultaneous to moving Springfield up five.  Does it value big wins more than it devalues a loss, or is it just because of the tight numerical ratings from earlier?  One thing is certain above all else, “Human’s are less likely to budge as much.” They often need more activation energy.  In fact, the total shift of 5 spots for Fisher (down 2) and Springfield (up 3) represented almost half of all from the top 15 teams in this week’s AVCA D3 Men’s Poll.  And compared to the 3 computer models, the AVCA movement was the least of all.

So the human component is slow to act, but what about the computer models? For one, they do not typically consider personnel on the court, or rather, off it.  Massey’s ranking of Wentworth last year demonstrates this fact.  This accounts for why earlier this season when everybody else had Wentworth in the top 5, it was still rating the Leopards at #17 – A residual from Laboo’s hand injury on Jake Baker’s “steel wall block” at Poly just a little over a year ago.  The T100 this week, because it is a point stealing algorithm, responded to 3 Stevens’ wins against some great opponents at Juniata last weekend to move past Wentworth to the #1 slot, while Coach Giglio’s team remained idle.  Even though Wentworth won their head-to-head match with Stevens about a week ago, the computer won’t typically respond to that match again after having already baked this outcome into its function.  No human being would do that, though, evidenced by the unanimity in first place votes this week in the AVCA Poll, as the coaches continue to love on Wentworth for that victory! Rightfully so, too.  You see, human’s have memory (Coaches are human, too).  … Well, so do computers, but not exactly the same kind. LOL  And then there’s Inside Hitter, whose pre-season and January rankings were kind of genius in how they encompassed the movement of talent across divisional and team lines.  If I understand correctly what they presently report, these ratings are now solely a function of the first 5 weeks of contests – Roughly 25% of the input expected to be there when it is all said and done.  Apparently it doesn’t yet recognize a middle of the pack team, starting 6-0 & soon to be 9-0, who will have dominated 9 of the weakest teams in the country (all among the ranks from #96 to #124) as not actually being among the top quarter of volleyball teams in the nation having done so. I ask myself, “What kind of picture might demonstrate these things of which I speak?”

The point is they all have their warts which in the stats world is often one source of “noise!”  However, the signal these 4 systems come together to compose, like beautiful music squelching out annoying missed notes and stray maniacal sounds emanating from the rear of the orchestra where the percussionists often dwell, is still pretty darn pleasing.  It’s called “MORE” – Median Order Rank of the Experts.  The intent for the remainder of this season is to have MORE no longer be a part of a weekly post, like T100 will continue to be, but instead be a standard offering on Wednesdays, found at the top of the site with the other headers where everyone can check in mid week, right after the coaches have done their thing.  

Median Order Rank by the Experts –  Give me MORE!

	As of 2/6/24	AVCA	MAS	IHIT	T100	MED	AVE
1	Stevens	2	1	4	1	1.5	2.0
2	Went	1	4	2	2	2.0	2.3
3	S. Va	3	5	1	4	3.5	3.3
4	Spring	4	2	5	3	3.5	3.5
5	Loras	8	3	3	6	4.5	5.0
6	Cal Lu	5	6	6	7	6.0	6.0
7	Vassar	6	7	9	5	6.5	6.8
8	SJFU	7	9	8	10	8.5	8.5
9	Nazareth	9	10	7	12	9.5	9.5
10	NC [IL]	10	12	11	8	10.5	10.3
11	Juniata	11	11	15	9	11.0	11.5
12	Mary	17	8	12	14	13.0	12.8
12	NYU	12	14	14	11	13.0	12.8
14	Carthage	13	13	13	13	13.0	13.0
15	Nichols	15	15	10	15	15.0	13.8
16	Dom (IL)	14	16	17	16	16.0	15.8
17	Ben (IL)	19	17	18	18	18.0	18.0
18	Miser	16	25	16	20	18.0	19.3
19	BW	18	19	26	17	18.5	20.0
20	Lasell	*21	20	24	19	20.5	21.0
21	UCSC	*22	22	20	21	21.5	21.3
22	SUNY NP	*24	18	26	22	23.0	22.5
23	Messiah	20	23	26	23	23.0	23.0
24	MIT	*23	26	22	25	24.0	24.0
25	Stevenson	*25	24	25	24	24.5	24.5

And because I like to be entertained, I also invented a game to play for the next 10 days -”DUAL DUELS!” It goes something like this:  I noted all teams from the 3 computer rankings that are double digits apart from one another.  For example, T100 might have a team > 10 spots better than Inside Hitter’s rank for that team, while maybe Massey might be < -10 spots lower than T100 for another, etc.  From the comprehensive list among all 3 computer models, I then went searching for matches in the next 10 days when any team rated a double digit rank better from any model would play another which is double digit ranks lesser from any other model. I found 8 matches to qualify. (Actually 7 because the same match came up two different times.) I then morphed their team ratings into win probability for each.  Now all that can be done is to wait and see how these contests go down to declare the winner of every DUAL DUEL.  It means nothing, of course, because of the small sample sizes, but like I mentioned earlier, mildly “Entertaining!”