How good/bad is Pomeroy?

[ATWizJr]

Through last night's games, 14 of the Pomeroy top 20 are out.  Seems high, no?  The following have lost but it seems to have no impact on their ranking with Pomeroy:

2   Duke
4 TX
5 Pitt
7 WI
8 SDSU
9 Purdue
10 SYR
12 BYU
13 "Ville
14 ND
17 WA
18 IL
19 Belmont
20 UT St.

Yet, these teams remain in the top 20?  And these rankings are after yesterday's games! 

How can Duke still be # 2 after yesterday?  How can WI stay # 7 after the Butler (ranked # 46!) loss?  How can MU beat the # 10 Orange and move up one stinking place to # 26 while Syr remains # 10?

Does not make sense.

[Stuckin1977]

While I love Pomeroy's website and check it almost daily throughout the season, you have to take it with a grain of salt.

His rankings are about 99.9% quantitative and 0.1% qualitative.  His formula is good because it takes into account close losses, where the RPI doesn't care how much you won or lost by, just whether you were on the road or at home.  What Pomeroy doesn't take into account are the qualitative factors: eye test, coach's resume in the tourney, tournament experience of the team, year of collegiate experience on the team, one-on-one matchups, etc.

[Tugg Speedman]

Also Pomeroy is the compilation of the entire season.  With 30+ games already in the books, one more win or loss is not going to change much.

Are you suggesting that Butler move from #46 to #10 because they beat Bucky last night?

If so, why didn't St. Johns move to #1 when they smoked Duke in February?

[NersEllenson]

Also Pomeroy is the compilation of the entire season.  With 30+ games already in the books, one more win or loss is not going to change much.

Are you suggesting that Butler move from #46 to #10 because they beat Bucky last night?

If so, why didn't St. Johns move to #1 when they smoked Duke in February?

You hit the nail on the head....1 game is not going to move a team significantly in Pomroy's rankings as they are based on body of work for the season.  I find Pomroy's ranking/predictions to be far and away the best...better than RPI, Saragin, and definitely the AP and Coaches polls.

[Dr. Blackheart]

Also Pomeroy is the compilation of the entire season.  With 30+ games already in the books, one more win or loss is not going to change much.

Are you suggesting that Butler move from #46 to #10 because they beat Bucky last night?

If so, why didn't St. Johns move to #1 when they smoked Duke in February?

Bingo...the reverse side of this is that the Big Dance is sudden death and any team can beat a top team on any given day. As an 11 seed, MU has no pressure on them...this is our second lowest seed ever in school history. Just have fun and play determined. Pressure does weird things to talented teams. MU has faced adversity all year and is in a familiar locale. Kentucky has already switched team hotels but MU is set up at their Seton Hall hotel out here.

[The Man in Gold]

Through last night's games, 14 of the Pomeroy top 20 are out.  Seems high, no?  The following have lost but it seems to have no impact on their ranking with Pomeroy:

So he has 6/12 remaining teams as of right now... That would be a decent bracket the way things have shaken out this year.

[THRILLHO]

First, as someone else pointed out, 6/12 isn't bad.  But you also need to take into account that these games are single elimination, top teams have played each other a lot at this point, and much of the difference between close teams is noise or 'luck'.

With regards to the 2nd point, many the top 20 teams lost to other top 20 teams.
Belmont (19) lost to Wisconsin (9)
Illinois (18) lost to Kansas (3)
Washington (17) lost to UNC (15)
BYU (12) lost to Florida (16)
SDSU (8) lost to UConn (11)

2 other teams were knocked out by Arizona (23).

That's the nature of a bracketed tournament.

One thing I think that Pomeroy doesn't account for is "hotness" -- I believe that all games in the season are weighted equally.  Obviously new games will move the needle a bit, but the model treats it as an extra data point from a fixed distribution, not evidence of a changing underlying distribution.  I humbly submit that this models experienced teams like Duke well but probably does not account for young teams with good coaches like Arizona that can improve throughout the season.  I'm not sure that there is a good way to account for teams that are as variable as Marquette or teams like Butler that just seem to come to play.

[Jay Bee]

The RPI in late November and early December is much better. 

[ATWizJr]

So, then, why would we drop 5 spots after last night's loss?

[Stuckin1977]

So, then, why would we drop 5 spots after last night's loss?

Pomeroy's formula takes into account how much you win or lose by.  If you looked at MU's page before the game, he projected to lose by (I think) 6 points.  Since we got "blown out" in relation to what his model was predicting, we got bumped back more than just a spot or two.

[brewcity77]

Pomeroy's formula takes into account how much you win or lose by.  If you looked at MU's page before the game, he projected to lose by (I think) 6 points.  Since we got "blown out" in relation to what his model was predicting, we got bumped back more than just a spot or two.

Maybe it's just me, but I think that's crap. When you look at some of these other teams, it's inexplicable that a loss to North Carolina would drop us 5 spots, even if it was a blowout loss. We lose a game we're expected to lose by 12 points more than expected and we drop 5 spots, whereas Wisconsin loses by 7 to a team they were supposed to beat by 6 and they drop 1 spot?

I don't know how his algorithm works, but I can't see any logical way that makes sense.

[Brewtown Andy]

Maybe it's just me, but I think that's crap. When you look at some of these other teams, it's inexplicable that a loss to North Carolina would drop us 5 spots, even if it was a blowout loss. We lose a game we're expected to lose by 12 points more than expected and we drop 5 spots, whereas Wisconsin loses by 7 to a team they were supposed to beat by 6 and they drop 1 spot?

I don't know how his algorithm works, but I can't see any logical way that makes sense.

A lot of it has to do with the efficiency ratings.  If you play a game where you're vastly different than your average efficiency for the season, you move up or down depending on whether you were better or worse than average.

A team could get blown out, but if they played an average game for them and the winner was miles above their average, the losing team probably wouldn't move all that much.