I embarked on this analysis to test the hypothesis...is it possible BYU's lack of NCAA tournament wins in the 2000's is a random event or is it evidence of BYU choking or something about BYU's teams that have been unworthy of winning in the tournament.
Here are my assumptions:
1. Most accurate ranking system is the final ranking at Massey's college basketball comparison, which is an average of many computer models. IMHO, this is more accurate than any one computer model such as Sagarin, human polls, or any other ranking system.
2. There is no additional predictive quality of how teams will do in the tournament outside this ranking. i.e. great guard play, senior dominated team, etc.
3. I translated power rankings and probabilities using historical spreads of teams ranked 1 through 350 and applying to the final Massey comparison poll.
Analysis:
year, byu seed, byu ranking, opponent, opponent seed, opponent ranking, probability byu win, predicted MOV, actual MOV, byu should have been seeded, opponent should have been seeded, opponent should have been ranked, should be probability of byu win
(the should be seeding and probability is based on where byu should be seeded based on their ranking and the respective opponent they should have been matched up with)
2001, 12, 48, Cincy, 5, 22, .32, -4.3, -25, 12, 5, 18, .30
2003, 12, 34, UCONN, 5, 18, .36, -3.4, -5, 9, 8, 31, .46
2004, 12, 37, Syracuse, 5, 21, .38, -2.7, -5, 10, 7, 28, .43
2007, 8, 40, Xavier, 9, 41, .505, +0.1, -2, 10, 7, 25, .385
2008, 8, 33, T A&M, 9, 23, .41, -1.9, -5, 9, 8, 32, .485
2009, 8, 27, T A&M, 9, 34, .58, +1.7, -13, 7, 10, 38, .60
In total, we should have had 2.56 wins over those six games. (2.66 wins with the "should" probabilities). The probability that we lose all six with those assumptions: .033.
Conclusions and observations:
--We lost four games by five points or less. A few breaks and we easily win a couple of those games.
--Bottom line, the .033 number seems too low to blame on simple randomness. There definitely seems to be an additional factor(s). The rankings don't account for SOS enough and we get overranked? Something about our style of play? We're not battle tested and get nervous on a big stage? (Utes, here's your cue to act like jerks in this thread)
--2003, 2004, and 2008 we got screwed pretty good, either by our seed or our opponent seed (2008)
--Though our 0-6 record looks really bad, we were only favored in two games with both of those being pretty close to pick -em type games.
As a check to my data, this is what I compute the probabilities of first round games should be vs what Indy linked to from the Wikipedia site:
seed, my probability, actual probability
1 vs 16, .995, 1.0
2 vs 15, .95, .96
3 vs 14, .9, .85
4 vs 13, .81, .79
5 vs 12, .695, .66
6 vs 11, .67, .69
7 vs 10, .585, .61
8 vs 9, .525, .46
Note, the reason the probabilities go up so fast after the 5 vs 12 is due to conference champions being in those seeds that are below what a 13th seed would be just choosing the top 64 teams. Up until the 12th seed, the teams are seeded pretty close to their rank/4.
Here are my assumptions:
1. Most accurate ranking system is the final ranking at Massey's college basketball comparison, which is an average of many computer models. IMHO, this is more accurate than any one computer model such as Sagarin, human polls, or any other ranking system.
2. There is no additional predictive quality of how teams will do in the tournament outside this ranking. i.e. great guard play, senior dominated team, etc.
3. I translated power rankings and probabilities using historical spreads of teams ranked 1 through 350 and applying to the final Massey comparison poll.
Analysis:
year, byu seed, byu ranking, opponent, opponent seed, opponent ranking, probability byu win, predicted MOV, actual MOV, byu should have been seeded, opponent should have been seeded, opponent should have been ranked, should be probability of byu win
(the should be seeding and probability is based on where byu should be seeded based on their ranking and the respective opponent they should have been matched up with)
2001, 12, 48, Cincy, 5, 22, .32, -4.3, -25, 12, 5, 18, .30
2003, 12, 34, UCONN, 5, 18, .36, -3.4, -5, 9, 8, 31, .46
2004, 12, 37, Syracuse, 5, 21, .38, -2.7, -5, 10, 7, 28, .43
2007, 8, 40, Xavier, 9, 41, .505, +0.1, -2, 10, 7, 25, .385
2008, 8, 33, T A&M, 9, 23, .41, -1.9, -5, 9, 8, 32, .485
2009, 8, 27, T A&M, 9, 34, .58, +1.7, -13, 7, 10, 38, .60
In total, we should have had 2.56 wins over those six games. (2.66 wins with the "should" probabilities). The probability that we lose all six with those assumptions: .033.
Conclusions and observations:
--We lost four games by five points or less. A few breaks and we easily win a couple of those games.
--Bottom line, the .033 number seems too low to blame on simple randomness. There definitely seems to be an additional factor(s). The rankings don't account for SOS enough and we get overranked? Something about our style of play? We're not battle tested and get nervous on a big stage? (Utes, here's your cue to act like jerks in this thread)
--2003, 2004, and 2008 we got screwed pretty good, either by our seed or our opponent seed (2008)
--Though our 0-6 record looks really bad, we were only favored in two games with both of those being pretty close to pick -em type games.
As a check to my data, this is what I compute the probabilities of first round games should be vs what Indy linked to from the Wikipedia site:
seed, my probability, actual probability
1 vs 16, .995, 1.0
2 vs 15, .95, .96
3 vs 14, .9, .85
4 vs 13, .81, .79
5 vs 12, .695, .66
6 vs 11, .67, .69
7 vs 10, .585, .61
8 vs 9, .525, .46
Note, the reason the probabilities go up so fast after the 5 vs 12 is due to conference champions being in those seeds that are below what a 13th seed would be just choosing the top 64 teams. Up until the 12th seed, the teams are seeded pretty close to their rank/4.
Comment