? About ESPN bracketology | The Boneyard

? About ESPN bracketology

HuskyNan

HuskyNan

You Know Who
Joined
Aug 15, 2011
Messages
29,934
Reaction Score
276,552
How often is Creme’s final bracket the same as the actual one? Anyone ever do an analysis?
 
HuskyNan said:
How often is Creme’s final bracket the same as the actual one? Anyone ever do an analysis?
Expand ...
I’ve never heard of anyone doing an analysis, but it’s safe to say the bracket per se is never exactly the same.

What he has been right about more often than not is (a) the #1 and #2 seeds and their placement, and (b) the selection of the at-large field, with maybe 1 or 2 misses.
 
Plebe said:
I’ve never heard of anyone doing an analysis, but it’s safe to say the bracket per se is never exactly the same.

What he has been right about more often than not is (a) the #1 and #2 seeds and their placement, and (b) the selection of the at-large field, with maybe 1 or 2 misses.
Expand ...
\

Predictions are typically educated guesses. If your data/statement is accurate (no reason to think otherwise) he is doing pretty good. I wouldn't bet the house on his data but it could be encouraging news for some.
 
Also "same" has multiple meanings.

Same in terms
  • of the 64 teams? (I think he usually gets 61-64)
  • accurate seeding (Ive seen him post stats about this but I forget what they were)
  • accurate top 4 by region
?
 
Earlier this week there was a "Creme" thread and no one was able to give any statistics on his efforts. I called his work "cr*p" in an effort to get some real statistical information but that didn't go over too well. My thinking was that he just is doing it for his ESPN paycheck because no matter how close he is or how far off he is, it doesn't matter because it's purpose is just to keep people talking/posting.
 
HuskyNan said:
How often is Creme’s final bracket the same as the actual one? Anyone ever do an analysis?
Expand ...
I've been curious about this with Lunardi as well.
 
.-.
Those articles don't really say much - they don't compare him with anyone or anything else. If you just took the Massey Ratings, say, and made the obvious adjustments to meet the NCAA rules for preventing rematches, would it be any less accurate? I bet it would be close.
 
Centerstream said:
Earlier this week there was a "Creme" thread and no one was able to give any statistics on his efforts. I called his work "cr*p" in an effort to get some real statistical information but that didn't go over too well. My thinking was that he just is doing it for his ESPN paycheck because no matter how close he is or how far off he is, it doesn't matter because it's purpose is just to keep people talking/posting.
Expand ...
He is doing his best. Making his best guess, really. One of the things I saw admitted (at some point) was that he doesn't claim his placements for sub-regionals are accurate, there are just too many variables. He is pushing his top 4 lines (completely), his "who's in / who's out?" and his seed lines.

While I don't have numbers, I would suggest that he is as accurate as anyone would be. Anyone that has the detailed knowledge of how the committee works and access to whatever data he has could probably come equally close. Most of us don't have the time and lack some of the details of the committee process.
 
It would be interesting if Creme gave us, on air, his updated bracketology for the top 16 teams minutes before the next tournament committee reveal.
 
I remember thinking last year that Creme missed in quite a few of his calls. I wish I could remember which ones.
 
.-.
Centerstream said:
Probably the ones where the Committee ignored their own "rules".:rolleyes:
Expand ...
I hate to break it to you. The committee has no rules, merely guidelines that they can uphold or ignore at their discretion.

If the NCAA really wanted to take the “subjective element” out of it’s selection process, they could have some computer gurus develop an algorithm to select all 64 teams, seed and locate them, without any further human intervention.

It really wouldn’t be very hard to develop such a process. Clearly, that’s not what the NCAA wants.
 
When I watch the WCBB selection show I have Charlie's last bracketology and circle where he's right and x out where he's wrong! He's usually in the 35% to 50% range. He might have right team in right seed but in another region! Or 1 seed up or 1 seed down, example, 9th or 10th seed! He's usually accurate in the top 8 seeds.
I think this year he may be more wrong than right as the parity between teams is a hairs breath apart. Too many upsets by underdogs!
 
oldude said:
I hate to break it to you. The committee has no rules, merely guidelines that they can uphold or ignore at their discretion.

If the NCAA really wanted to take the “subjective element” out of it’s selection process, they could have some computer gurus develop an algorithm to select all 64 teams, seed and locate them, without any further human intervention.

It really wouldn’t be very hard to develop such a process. Clearly, that’s not what the NCAA wants.
Expand ...
Gee, I wondered why I used "rules".:rolleyes:.
:)
 
Charlie Creme’s latest top 16, in order of seed in each region:
Greensboro: Baylor, Mississippi State, Oregon State, Miami
Chicago: Louisville, Stanford, Iowa, Marquette
Albany: UConn, NC State, Maryland, South Carolina
Portland: Notre Dame, Oregon, Iowa State, Gonzaga
 
.-.
jaydee602 said:
Charlie Creme’s latest top 16, in order of seed in each region:
Greensboro: Baylor, Mississippi State, Oregon State, Miami
Chicago: Louisville, Stanford, Iowa, Marquette
Albany: UConn, NC State, Maryland, South Carolina
Portland: Notre Dame, Oregon, Iowa State, Gonzaga
Expand ...
looks like he gave UConn the easiest path.
 
jaydee602 said:
Charlie Creme’s latest top 16, in order of seed in each region:
Greensboro: Baylor, Mississippi State, Oregon State, Miami
Chicago: Louisville, Stanford, Iowa, Marquette
Albany: UConn, NC State, Maryland, South Carolina
Portland: Notre Dame, Oregon, Iowa State, Gonzaga
Expand ...

I wouldn't say Louisville has that tough a draw. While Stanford is the only team who beat Baylor, outside of that game when they shot way over their heads from the perimeter, they have looked underwhelming this year (IMO) And yes, I know we're just commenting on Creme's fantasy brackets. Reality could be much different.
 
oldude said:
It really wouldn’t be very hard to develop such a process. Clearly, that’s not what the NCAA wants.
Expand ...

I have to disgree with this statement. IMO this would be extremely hard to develop because their are too many variables. The first question I feel that would need to be dealt with is whether you want the 64 teams that are best at the end of the season or the 64 teams that have the best overall body of work (I personally have no idea which the committee shoots for). All of the computer models currently used to rank teams have flaws. Some less so then others. If you are comparing teams that have not played each other and somehow determine an accurate measure of SOS you would still -need to factor in such things as teams improving during the year, the effect of injuries - short and long term - and a plethora of other factors. I say if you think it is not hard to do give it a shot.

I do think that when discussing what 64 teams are going to get in that at least 56 are pretty much locks so at most there are only 8 tough choices (and probably fewer).

I am a strong advocate of limiting at-large selections to teams with at least .500 records in the league. I have not yet decided if this should or should not include the league tournament.
 
Yorkeez4 said:
I have to disgree with this statement. IMO this would be extremely hard to develop because their are too many variables. The first question I feel that would need to be dealt with is whether you want the 64 teams that are best at the end of the season or the 64 teams that have the best overall body of work (I personally have no idea which the committee shoots for). All of the computer models currently used to rank teams have flaws. Some less so then others. If you are comparing teams that have not played each other and somehow determine an accurate measure of SOS you would still -need to factor in such things as teams improving during the year, the effect of injuries - short and long term - and a plethora of other factors. I say if you think it is not hard to do give it a shot.

I do think that when discussing what 64 teams are going to get in that at least 56 are pretty much locks so at most there are only 8 tough choices (and probably fewer).

I am a strong advocate of limiting at-large selections to teams with at least .500 records in the league. I have not yet decided if this should or should not include the league tournament.
Expand ...
And everything you just outlined can be written into a program and weighted by whatever priority you choose.
 
RSHERMVIKES said:
When I watch the WCBB selection show I have Charlie's last bracketology and circle where he's right and x out where he's wrong! He's usually in the 35% to 50% range. He might have right team in right seed but in another region! Or 1 seed up or 1 seed down, example, 9th or 10th seed! He's usually accurate in the top 8 seeds.
I think this year he may be more wrong than right as the parity between teams is a hairs breath apart. Too many upsets by underdogs!
Expand ...
I am really pretty sure that - after the top 4 - his regions are not intended to be completely accurate. As I said above, he is pretty confident in his seed lines, remember, teams can be moved one seed line to accommodate geographic concerns, IIRC. There are too many variables - especially since they no longer simply fill in the S curve - to get the sub-regionals correct. I would suggest that he is usually north of 90% on who's in, and probably north of 80% on the seed lines. And often, his discrepancies are the ones that puzzle lots of folks, when the committee does something weird.
 
.-.
Yorkeez4 said:
I have to disgree with this statement. IMO this would be extremely hard to develop because their are too many variables. The first question I feel that would need to be dealt with is whether you want the 64 teams that are best at the end of the season or the 64 teams that have the best overall body of work (I personally have no idea which the committee shoots for). All of the computer models currently used to rank teams have flaws. Some less so then others. If you are comparing teams that have not played each other and somehow determine an accurate measure of SOS you would still -need to factor in such things as teams improving during the year, the effect of injuries - short and long term - and a plethora of other factors. I say if you think it is not hard to do give it a shot.

I do think that when discussing what 64 teams are going to get in that at least 56 are pretty much locks so at most there are only 8 tough choices (and probably fewer).

I am a strong advocate of limiting at-large selections to teams with at least .500 records in the league. I have not yet decided if this should or should not include the league tournament.
Expand ...

It's never a selection of the best 64 teams, due to conference tournament winners. Upsets there can put teams into it that wouldn't have qualified for an at large bid.
 
Here are his results from last year. Less accurate than I expected.
  • He only missed 1 team. (Rutgers instead of Oklahoma selected). I would expect 0-1.
  • He got 56% team's seed correct and 88% within 1 seed. So only 8 teams missed by >1 seed.
  • Though only 17 of 64 perfect (27%) i.e. Seed + Regional.
I find them an interesting read but not worthy of FOMO.

2017-18 NCAA WBB Bracket vs Creme.png
 
maccca3232 said:
Here are his results from last year. Less accurate than I expected.
  • He only missed 1 team. (Rutgers instead of Oklahoma selected). I would expect 0-1.
  • He got 56% team's seed correct and 88% within 1 seed. So only 8 teams missed by >1 seed.
  • Though only 17 of 64 perfect (27%) i.e. Seed + Regional.
I find them an interesting read but not worthy of FOMO.

View attachment 40119
Expand ...
This is actually pretty darn good, if you ask me. To predict 97% of the at-large field correctly and to get 88% of the field within one seed line is quite good. Let's recall that the committee can move teams up or down a seed line in order to achieve a "better" bracket (in its estimation). Regional placement is inherently unpredictable because it can be impacted in a domino effect by widely varying permutations of decisions.
 
oldude said:
And everything you just outlined can be written into a program and weighted by whatever priority you choose.
Expand ...
And every program ever written (possibly with the exception of NASA) will have bugs in it, guaranteed. The more conditions/variables in it, the bigger the chances.
 
Centerstream said:
And every program ever written (possibly with the exception of NASA) will have bugs in it, guaranteed. The more conditions/variables in it, the bigger the chances.
Expand ...
Just to be clear. I’m playing devils advocate by suggesting a completely computerized selection process in answer to those who want to trash the work of the committee or Charlie Creme’s projected brackets.

There is, of course, no perfect system. As Geno suggests, teams need to take care of what they can control and not worry about where they’re sent and who they play.
 
oldude said:
Just to be clear. I’m playing devils advocate by suggesting a completely computerized selection process in answer to those who want to trash the work of the committee or Charlie Creme’s projected brackets.

There is, of course, no perfect system. As Geno suggests, teams need to take care of what they can control and not worry about where they’re sent and who they play.
Expand ...
No problem and I understand you being devil's advocate.
I was basing my post on the fact that I was a software tester for the final 25 years of my working life. No matter how many times some program has been tested, once placed into production, an issue will come up.
I am a Charlie basher but only because the Committee puts out a couple Top 16s during the season and the only bracket that matters is the Committee's final decision.
 
.-.
You must log in or register to reply here.

Forum statistics

Threads
168,535
Messages
4,581,050
Members
10,491
Latest member
7774Forever


Top Bottom