2025-2026 MBB computer projections thread
- Thread starter Sigmapolis
- Start date
At some level I appreciate BT throwing out MoV at a certain point.
Who cares if you beat some cupcake by 45 instead of 35 on your own court.
You kicked their ass either way. The higher MoV is probably more indicative of just how motivated the power team was that night and/or how long the coach decided to play the back of the bench.
Then again... beating Purdue on their court by 25 instead of 15 might be useful data.
It's not often somebody does that. Much less they can pour it on at the end.
However, in the end these decisions are small modeling choices that really only make a difference when you are trying to do "angels on a head of a pin" comparisons between teams up in rarified air.
Generally a good problem to have if we're noodling on these choices as to why we're #2 or ~#4.
The slight divergence between the models is why the committee is given multiple rankings.
I've been wondering how some of the different systems phase out last year's data. If they keep all of last season's data together and phase it out as a group as we get farther into the new season that would yield different results for us than if they start phasing out the oldest games from the previous year until it is all gone. Because our metrics were better for the first half of last year than the second half.
I ran it by my wife, she says I am f'ing crazy but she doesn't even know who Grant Hill.
I thought I was crazy until I started looking at numbers.
I think kenpom gradually (game by game) reduces the number of games from the previous season that are included in determining this season's rating, with eventually only data from the current season being used, probably around this time. I can't find the post about this either however. There definitely was one in recent years.
That's a good question, but I'm pretty sure it's more similar to the first one. Because the preseason ratings aren't just taking data from the last season.
The preseason ratings come from formulas that include transfers, recruits, and outgoing players, and those players' prior production (or recruiting rankings, for freshmen) to calculate the projected offensive rating. Then the defensive ratings do more heavily lean on the program's previous years' defensive ratings, because individual defensive metrics don't translate as much as a coach's defensive success from year to year.
If all of that is accurate, I don't think you could simply remove games from the prior year in the formula as you get further into the current season. My assumption has always been that it's more of a weighted average. For instance, after game 1, Bartthag might be 95% preseason and 5% current season, while after game 10, it might be 85% current season and 15% preseason.
That has always been my assumption, but I'm starting to wonder. And I wouldn't think they remove them one by one as the season goes one, but some type of phase out that emphasizes most recent data points. It sure has felt like on Torvik and a couple other metrics that we have faced headwinds over the last month or so that keep our overall rankings below what they would be if it was just this years data, when it feels like it should be the opposite and we keep slowing moving up as last years data falls away.
10 quad 1 games in the remaining 15 games. its nuts that this team is projected to win 27 or 28 games, but they are that good.
The model doesn't project discrete Ws and Ls but rather probabilities.
A 50% game is projected as 0.5 of a W and 0.5 of a L in the results.
Hence, if you add up the possibility of losing the Q2 games eventually their sum (even if each individual one is in the 15% to 20% range) gets up to the point where you project one Q2 loss.
In the real world, that might be like the Oklahoma St. game last year.
Road game. Sleepy gym. Tired team after a long grind. Opponent who isn't a world beater (not Arizona or BYU or somebody genuinely dangerous) but just good enough to beat you on their best night.
I hope it doesn't happen but even good teams can have a night like that one.
It's probably just more of an assumption that statistically based on the number of quad 2 games we have remaining we will lose one of them rather than it being a specific opponent.
As expected. Cyclones up to #2 in the AP poll, and received one first place vote.
apnews.com
NCAA College Basketball Rankings: AP Top 25 Basketball Poll
The official page for the AP Top 25 men's college basketball poll. Keep up to date and view the latest NCAA men's basketball ranking updates each week for the 2025-26 season.
apnews.com
Michigan drops 3 spots with a home loss against unranked Wisconsin. If we lose at KU, I bet any amount of money that we drop much further.
I agree. Luckily we're not going to find out.
If we lose a close game at Kansas and win at Cincinnati, I would take that bet we don't fall below Purdue, Duke, and Houston. None of those three play a ranked team this week and its still in people's consciousness that we killed Purdue at Mackey. We would almost certainly fall below UConn and Michigan.
David Jablonski (Cox Media Group) has us 5th, after UA, Nebby, Mich & Vandy. I think that's the lowest. Most have us #2. A good number have UConn 2 and us 3.
