(01-16-2023 12:46 PM)Tiger87 Wrote: (01-16-2023 12:30 PM)msu35 Wrote: (01-16-2023 11:55 AM)Stammers Wrote: (01-15-2023 01:32 PM)UofMstateU Wrote: (01-15-2023 01:07 PM)msu35 Wrote: As with all analytic systems for basketball, it's a push/pull dynamic. Our rating is influenced by not only our games but also our opponents. We could go down even if we won a game, which has already happened this year, therefore it's highly possible and even likely that we could go up after a loss.
I get that. About 14 years ago I wrote an RPI calculator program that would live game results and generate rpi's for the teams in real time. So I could see how the different metrics played out.
And NET has the RPI component in it. But its got some other special sauce in it as well. And some of that special sauce doesnt make sense.
To demonstrate how screwed the NET is as compared to the RPI, Ohio St's NET currently sits at 21. You look at number, and wonder how the hell that is even possible, when the following is also true:
1. They have no quad 3 wins, because they stuffed their schedule with Quad 4 teams
2. They lost to a quad 4 team AT HOME
3. Their non-conference schedule is #204 and their overall sos is 77. Quite ugly.
4. They have a single road win.
5. They are sitting at 10-6
6. Their RPI is 94
The RPI over-punishes for losing at home to extremely great teams. Which is one of its flaws. (The other is rewarding away wins against crappy teams.) But for its flaw, it appears they jacked with some calcs in the NET that overcompensates to the ridiculous degree. While Ohio St is not the #94 team as the RPI says, they are also nowhere near the #21 the NET seems to think they are.
Creighton's is even worse...
24 CREIGHTON 9-8
Q1: 1-6
Q2: 3-1
Q3: 2-1
Q4: 4-0
46 MEMPHIS 13-5
Q1: 1-2
Q2: 3-3
Q3: 5-0
Q4: 4-0
Q1 - Q2
4-7 CREIGHTON
4-5 MEMPHIS
Q1 - Q3
6-8 CREIGHTON
9-5 MEMPHIS
Q3 - Q4
6-1 CREIGHTON
9-0 MEMPHIS
The only reasonable conclusion is as I suggested. The "secret" component with the NET is that it gives disproportionate weight to the pure number of Q1 games. There is no other
I'm going to be a dissenting voice and throw a spanner in the works. Take a look at the margin of victory. It has to play a significant role in the calculations. This year, we've been unable to "blow" teams out to the degree the aforementioned teams have, and our rating has suffered.
But don't they claim the margin of victory is capped at 10 or 15 points, to prevent intentional blowouts?
Here is a very good article that answers all of our questions. Spoiler alert. Non G5 conferences get completely screwed over.
Quote:Scoring Margin
Simply the difference between a team’s score and its opponent’s score. However, the point differential is capped at 10 points, and all overtime games are capped at 1 point.
Quote:Team Value Index
The Team Value Index component of the NET is a results-oriented algorithm designed to reward teams for beating other good teams. The man remains mysterious behind the curtain, but the NCAA has said that this component includes factors such as who won, the opponent, and the location.
Scoring Margin is straightforward. A cap of 10 points, and 1 point for all overtime games. That means that even if we haven't been blowing out many teams, our 5 losses are by a combined 18 points, and we fall below the 10 point threshold in every loss. We have two wins by 9 points, one win by 8 points, and one win by 7 points. Presumably, we don't get maximum value in 4 of our wins because of a point here or there. It should also be noted that P5 teams play a much higher percentage of non P5 teams at home, which gives them greater opportunities to win by 10 or more points.
Team Value Index is where the real screwing takes places. Look into this component carefully. As "a results-oriented algorithm designed to reward teams for beating other good teams."
So just as I suspected, the P5 conferences get to double dip. They get rewarded/not punished for good losses as everyone else does, BUT they get disproportionately rewarded for beating good teams. The short story is that 6-12 in Q1 gets you rewarded more than 4-5, because there is a premium reward based on the pure number of wins. This 100% creates the inequities, end of story.
We only had 2 optimal chances at a Q1 win. Neutral against 41 Iowa State and home against 29 Murray State. All of our other games were a real grind. We only had two Q1 opportunities at home, and 3 out of our 7 Q1 games were against #2 Houston, with 1 at home, 1 on the road and 1 a de facto road game in Texas.
3 out of our 7 games were against the #2 team in the NET. 5 of our 7 games were against teams in the top 30.
Memphis 2022
Q1 3-4
W 22 Virginia Tech - N
L 41 Iowa State - N
L 29 Murray State -H
W 2 Houston - A
L 53 SMU - A
W 2 Houston - H
L 2 Houston - N
Compare this to UConn. They were 5-8 in Q1, and had a whopping 10 of their Q1 games against teams that were outside of the top 10 of the NET. In fact, 8 of their games were against teams outside of the top 30.
Key UConn Losses
38 Michigan State - N
73 West Virginia - A
48 Seton Hall - A
37 Xavier - A
51 Creighton - A
Easy Q1 Wins
50 Marquette - A
64 St. John's - A
48 Seton Hall - N
Their two best wins, #6 Villanova and #13 Auburn both at neutral sites, don't have the value of our best wins against Houston.
So not only did UConn have infinitely easier Q1 games than we did, they also got rewarded for 5 Q1 wins compared to our 4 despite having a worse record. UConn was rewarded with a #5 seed, while we got a #9 seed.
Sorry for the long post. The very very short summary, is that the NET puts a premium on the pure number of Q1 wins. IF you are a P5 team that goes 6-12 in Q1 with multiple games outside of the top 25, you get much more value than a non P5 team that goes 3-4, with multiple games against the top 10. It is complete bull****.
Partial Explanation Of NET Rankings