IUHoosierJoe

I was under the impression that Pomeroy removes his pre-season rankings sometime in early January while Sagarin leaves at least a little percentage of his in for the entire season. But in looking at North Carolina for example, who was ranked high in pre-season, Sagarin has them #28 now, which is higher than anybody else. But Pomeroy has them #32, which is also pretty high compared to others. So maybe Pomeroy is leaving in some pre-season ratings.

Another thing about Sagarin's, I read an interview a few years ago where he said, in addition to making it entirely score-based, he leaves a percentage of his pre-season ratings in all the way to the end of the season because he found it predicted games even better that way. So he is somewhat determining his "computer" ratings with his pre-season ratings.

My system has Ohio State #39 in the score-based metric and #170 in the result-based. I've never seen that kind of differential, ever. I'm sure most know this, but that differential is what Pomeroy measures in his "Luck" rating. And of course, Ohio State is #363, lol.

Yeah, I agree. I know the computers don't like them (mine has them #45) because of the non-conference schedule and results. And I wonder sometimes if the voters follow the computers' lead for a school like Northwestern that rarely gets ranked and people don't see a lot. But their Big 10 resume speaks for itself.

My computer ratings using games through yesterday: 1 Alabama 2 Houston 3 UCLA 4 Kansas 5 Texas 6 Tennessee 7 Purdue 8 Baylor 9 Arizona 10 St Mary's CA 11 Gonzaga 12 Connecticut 13 San Diego St 14 Virginia 15 FL Atlantic 16 Kansas St 17 Marquette 18 Creighton 19 Xavier 20 Nevada 21 Indiana 22 Arkansas 23 Iowa St 24 Oral Roberts 25 Utah St

I understand what you guys are saying. He knows the game inside and out, and he knows these teams down to most finite detail. But he just talks too much for me. It sounds like this is a three-man crew, and the other two guys can barely get a word in. I can totally understand how a kid he’s coaching would tune him out. It’s just noise after a while.

My computer rankings through yesterday's games: 1. Alabama 2. Houston 3. Tennessee 4. UCLA 5. Purdue 6. Arizona 7. Texas 8. Kansas 9. St Mary's CA 10. Gonzaga 11. Baylor 12. Connecticut 13. FL Atlantic 14. Kansas St 15. San Diego St 16. Xavier 17. Marquette 18. Virginia 19. Iowa St 20. TCU 21. Utah St 22. Duke 23. Nevada 24. New Mexico 25. Indiana

My system likes the Mountain West as well. Their results-based metrics are especially good, because of their out of conference success. I have Clemson out (#69), Pittsburgh out (#59). I have West Virginia (#31) in as an 8-seed range. Pittsburgh, though, reminds me of Rutgers last year, as they have 5 wins over my top 50, so I think they'd get in based on that. My system does not like the ACC in general, and having to play Georgia Tech, Florida State, and Louisville won't help the rest of them any.

I’m with you. The play in games could be a second Big 10 tournament, lol.

Your system is quite interesting to me. Seems to be a unique way of looking at it, and hard to argue with the results. Pittsburgh is a team I’m keeping my eye on. Their resume is reminding me a little of Rutgers last year. I have them at 67 but they have 5 wins over my top 50.

My system actually ranks the Mountain West higher than the Pac 12 and ACC. In looking over their out of conference schedules, they have had some pretty good results. I’ll be interested to see how the tournament committee treats them.

Probably the same guy who tweeted that Tayven Jackson was going to Purdue, lol.

I wasn’t aware last year’s teams were playing. In any event, ESPN lost the Big 10 contract so they will be show the ACC every day and twice on Sunday, regardless of how good the teams are.

Through yesterday's games, my computer-generated rankings, 50/50 weighting of result-based and score-based: 1. Tennessee 2. Houston 3. Alabama 4. Purdue 5. UCLA 6. Kansas 7. Arizona 8. Texas 9. St Mary's CA 10. Gonzaga 11. Kansas St 12. Baylor 13. FL Atlantic 14. Connecticut 15. San Diego St 16. Marquette 17. Virginia 18. Iowa St 19. Xavier 20. TCU 21. New Mexico 22. Boise St 23. Duke 24. Utah St 25. Indiana As I mentioned a couple weeks ago, these have been a decent predictor of tournament selections and seeds with a couple exceptions: I tend to rank teams from smaller conferences higher than their seeding. For example, I have Florida Atlantic in the 4-seed range, and my sense, just based on past experience, is their ceiling is probably an 8-seed even if they win out, and even if they're ranked in the top 15 of the AP poll. But we'll see. The other exception, as I mentioned before, is a team here and there that gets in the tournament because of an abnormal number of Quad 1 wins versus their entire body of work. For example, I had Rutgers ranked nowhere near the cut off line last year. With that being said, I have Indiana a solid 7 seed now, playing 10-seed NC State in the first round.

True, but that year all the power conference teams had at least 5 regular-season losses. Gonzaga only had 2.

Interesting. Just to be clear, what I posted for 2013 isn't what I used for my top 25 I posted Monday, but what I have come up with to closely replicate the NET. For 2013, my NET-duplication formula resulted in Kansas ranked #6, behind Louisville, Indiana, Florida, Gonzaga, and Duke. Florida, with a 3 seed, certainly ended up below their ranking, lol.

I found it, and I was actually mistaken. I had Louisville the slight #1 over us in my attempted duplication. I had IU #2. Here are what I had as Quad 1 games: W N Georgetown (#14) W North Carolina (#29) L N Butler (#39) W @ Iowa (#51) W Minnesota (#26) L Wisconsin (#16) W Michigan State (#12) W Michigan (#8) L @ Illinois (#43) W @ Ohio State (#7) W @ Michigan State (#12) L @ Minnesota (#26) L Ohio State (#7) W @ Michigan (#8) W N Illinois (#43) L N Wisconsin (#16) So it looks like I had us 10-6 in Quad 1 in the regular season. And I'll note North Carolina is on the borderline and I can't duplicate it exactly, but pretty close.

I actually did that a while back, tried to replicate the NET formula (I can get very close to the current rankings but not exact) for that season. I think IU finished the regular season #1 with Louisville a close second. I can try to find it if you’re interested.

NET is reflective and predictive but there is more weight on predictive.

Alright, you guys have inspired me to post my computer-generated rankings 1. Alabama 2. Houston 3. Tennessee 4. UCLA 5. Purdue 6. Kansas 7. Texas 8. St Mary's CA 9. Arizona 10. Gonzaga 11. Kansas St 12. Connecticut 13. FL Atlantic 14. Baylor 15. Iowa St 16. Virginia 17. Auburn 18. New Mexico 19. Marquette 20. San Diego St 21. Xavier 22. TCU 23. Boise St 24. Utah St 25. Kent (Indiana #30)

Well, you forgot Billy Donavan, lol.

I realize that. There are no “set criteria.” They have criteria until they don’t. For instance, they look at “total body of work” until they want to take a team like Rutgers last year and ignore their bad losses. They have the NET, but it does not line up with the seeding because it weighs the predictive side too much. You don’t even need to dig into the sheet much, just weigh the result-based metric more than the NET does. With the RPI, there was no predictive side at all, so it wasn’t a set criteria either. They’re never going to put anything out there that gives away the whole field, because they want to leave themselves the flexibility to pick who they want. But Sagarin’s old ratings, before he changed his formula, was a really good predictor of the field and fairly good on seedlings, outside of, again, your teams like Rutgers last year getting in.

You may notice the NET rankings have Indiana 32 and Maryland 52. Mine are also computer generated but I put more weight on the result-based side than the NET does. I actually put equal weight on result-based and predictive, so mine are nearly identical to Sagarin’s old version in which he used pure points and elo chess. The committee used to rely on that old version of Sagarin’s rankings very heavily.

I’ve got Indiana in as a 12 seed (#47) and Maryland well out (#62).

99.9% of “disruptors” are a very small fraction as intelligent as they think they are.