What are these quadrants you speak of?
Quadrant 1: Home 1-30; Neutral 1-50; Away 1-75
Quadrant 2: Home 31-75; Neutral 51-100; Away 76-135
Quadrant 3: Home 76-160; Neutral 101-200; Away 136-240
Quadrant 4: Home 161-plus; Neutral 201-plus; Away 241-plus
3 Likes
Awesome that theyâre giving road wins more credence. Crappy that they are somewhat arbitrarily making a win against 31 being categorically different than a win against 30 (for example). Really crappy that theyâre using RPI to define the quadrants.
5 Likes
âReally crappy that theyâre using RPI to define the quadrants.â
^^^^THIS
Iâm going to keep re-hashing this but RPI is still the end-all-be-all by the lazy committee.
Is there a reson they couldnât create an additional set of quadrants based on Kenpom or other metrics? If you want to add those to the sheet, why not create quadrants out of those as well?
2 Likes
Where does Michiganâs updated page w/ results against the 4 quadrants live? Iâll bookmark it.
Believe that just has âprojectedâ quadrant data right?
I think you have to download the page with all the team sheets and scroll down to find UM. Thereâs a link to the file here:
https://extra.ncaa.org/solutions/rpi/SitePages/Home.aspx
You can also use this warrennolan.com page which includes a handy key to what the quadrants are and the teamâs record right below it:
1 Like
There will always be arbitrariness when you have cutoffs, the question is whether they put them in reasonable spots. I wonder if they still havenât properly taken into account road/home differences in spots â i.e., is a road win against Team 76 the same as a home win against Team 75? Is there too much overlap between home and neutral venues? But considering they wanted to keep it simple with four categories and have to start with some overlap (i.e., wins against top 30 teams are quadrant 1 no matter where they take place), it seems a definite improvement.
1 Like
Not sure Iâd call it improvement when a win at home against Buffalo is the same as a win at home against Michigan St.
Extremely flawed system of measurement.
Thanks for the team sheet info, dudes.
Yeah I understand. Itâs necessary, I just hope the committee dives deeper than just that sheet.
For example, if someone theoretically beat 31 and 32 at home,hopefully someone points that out that, â2 of this teamâs quadrant 2 wins are practically quadrant 1 wins, lets give them some credit for that.â It would at least be helpful in head to head comparisons maybe.
And agreed that if theyâre going to put some advanced metrics on there, that the quadrant wins for the advanced metrics are on the table somewhere as well.
The old system used RPI too and simply separated them into top 50, top 100, top 200, 201+. This seems like an improvement on that.
As to RPI generally, Iâm glad theyâre using advanced metrics more, but Ws and Ls are still significant data points. Certainly replicating the system with advanced metrics would be valuable, although youâd have to pick one or maybe average the three they use. Itâs not like kenpom is a platonic ideal of efficiency analysis.
1 Like
Now that I would love. Why not use the average of the averages with the advanced metrics and build a second set of quadrants based off that?
I would still weigh Kenpom heaviest. If you think about it, the Kenpom spreads are damn near identical to the Vegas spreads. Whatâs the point of me saying that? It tells you just how accurate his system of measurement is when Team A plays Team B at home/neutral/road. Therefore, his system can accurately tell us the difference between Team A and Team B, which is the whole point of the committee. We have a true understanding of just how good his rankings and system of measurement is.
Correct. But pretty easy to look at the âSchedule & Resultsâ on the right side of that page and determine the current quadrant records from the specified Q next to each opponent.
My thing with this has always been that the actual results on the floor should matter more than they do in a predictive model like KenPomâs. We should be trying to measure what teams accomplished rather than who would win a game. Iâve seen people use the language from the NCAA that maybe that isnât the exact written goal, but doesnât it just make sense?
I think Michigan lost to Purdue, I donât really think it is fair that that loss should count less than if they had lost by 10 or 20 points.
Just my 2 cents. I donât know what the proper solution is, but I donât think KenPom is just an easy no brainer answer.
I donât know, if Iâm on the fence of seeding identical teams or deciding between two teams to make the tournament, I wouldnât mind Vegas (or Kenpom in this instance) telling me which team would be favored on a neutral court. To me, this tells you who the better overall team is. Keep in mind, that spread would be from their overall season record and statistics.
Thatâs more accurate than RPI in my opinion. Iâd rather have that be my answer than some committee members who probably donât know a whole lot about these teams to begin with.
If people donât want to use one metric such as Kenpom, then I would be all on board of ranking teams based on the average of the averages. Use those rankings to fill out the quadrants instead of RPI, or at least as a second set with the RPI.
I feel like they could do a similar quadrant system but use kenpom (or whatever metric of their choice) to separate the quadrants. Then you still look at wins and losses and use them as a primary factor but the way of actually dividing the quadrants and defining what a good win or bad loss is would be better. The wins and losses of other teams shouldnât matter for your resume IMO (this is where I could see some contention), how good they are should. If Iâm being unclear I can clarify.
Agree with your take on KenPom. But I wonder if youâd end up with a lot more uncompetitive games if you went solely by accomplishment over the breadth of a season. A lot of teams are pretty different animals by seasonâs end. Michigan could get yoinked. I think they should use some combination of the above and then try to put together a great tourney, and without apologies. Everyone will be fighting about seedings long after weâre dead.
This distinguishes between how good the teams you play are and how many wins they have. If the goal is to win basketball games, then maybe the teams you play are good if theyâve won a lot of games.
A few additional things to consider:
-
RPI and kenpom are already highly correlated, but if kenpom truly does what it intends to (predict future W/L), if the teams played endless games and every other team in the NCAA, eventually kenpom and rpi (which is determined by actual W/L) would converge. But the teams donât play endless games and they donât play every team. And the largest differences will be result from which teams won or lost close games and who the teams played. Do we not want to reward those things?
-
Related point â we sometimes complain about teams âgamingâ the RPI, but generally that just means playing teams with better W/L records â which I believe is one of the committeeâs goals, to have teams play tougher non conference schedules. I have heard the complaint that teams can raise RPI by playing mediocre instead of bad teams without every playing really good teams. But itâs still playing higher competition.
-
I miss that time before kenpom was mainstream when you could do well in NCAA pools simply by exploiting the gap between RPI and kenpom.
-
As weâre talking about kenpom so much, here are some relevant, interesting thoughts from KP himself: