Hockeynerd88 Posted January 1, 2015 Share Posted January 1, 2015 What do you think, time to get rid of Pairwise? This year more than most it seems to not really be working. I have had issues with it in the past but this year the Pairwise is off. Thoughts? Quote Link to comment Share on other sites More sharing options...
siouxu31 Posted January 1, 2015 Share Posted January 1, 2015 It is meant to be a measuring tool at the end of the season, not 20 games in. It is too volatile with this few games played. Don't look at it again until early March and you will see that everything will shake itself out. Quote Link to comment Share on other sites More sharing options...
MafiaMan Posted January 1, 2015 Share Posted January 1, 2015 Well, there is the whole "evening out the whole season" issue. It was several years ago when Mankato and St Cloud had similar records near the end of the season. Mankato was on a roll and St Cloud had lost something like 5 or 6 in a row. Mankato even went into St Cloud and swept the Huskies in the 1st round of the WCHA playoffs. Thanks to the Pairwise, sputtering St Cloud made the NCAA tournament and the hot Mavericks were left at home. Not surprising, St Cloud was hammered in the first game of the tournament, I believe by Maine if memory serves me right. Quote Link to comment Share on other sites More sharing options...
The Sicatoka Posted January 1, 2015 Share Posted January 1, 2015 Well, there is the whole "evening out the whole season" issue. It was several years ago when Mankato and St Cloud had similar records near the end of the season. Mankato was on a roll and St Cloud had lost something like 5 or 6 in a row. Mankato even went into St Cloud and swept the Huskies in the 1st round of the WCHA playoffs. Thanks to the Pairwise, sputtering St Cloud made the NCAA tournament and the hot Mavericks were left at home. Not surprising, St Cloud was hammered in the first game of the tournament, I believe by Maine if memory serves me right. Pairwise used to contain a "last 16 games" component but folks complained that making the tournament should be based on a full season's resume. Quote Link to comment Share on other sites More sharing options...
gfhockey Posted January 1, 2015 Share Posted January 1, 2015 It's whoever did the schedule fault we scheduled so many cupcakes Quote Link to comment Share on other sites More sharing options...
siouxkid12 Posted January 1, 2015 Share Posted January 1, 2015 It's whoever did the schedule fault we scheduled so many cupcakes how is it the person who schedules the games fault? these games are scheduled a year or two in advances and no one can predict how the teams will do each year. Miami was suppose to be a darn good team last year and contend for the league title, they never did. 1 Quote Link to comment Share on other sites More sharing options...
siouxnami Posted January 2, 2015 Share Posted January 2, 2015 KRACH > Pairwise > Humans picking. Quote Link to comment Share on other sites More sharing options...
Hockeynerd88 Posted January 2, 2015 Author Share Posted January 2, 2015 I like the idea of humans being kept out of it for the most part, but I think they should look into it differently the way the RPI is weighted/used. Teams are rewarded for playing a "soft" schedule. Teams like Bowling Green, Michigan Tech, and Minnesota are virtual locks to make the NCAAs this year because of their schedule. (I am not debating the scheduling of these teams). Each of those teams did average to below average when playing the better teams. Then they do fine against the well below average teams. There needs to be much more of an emphasis for wins against top opponents, and I think they should lower the bar for teams that don't count against the RPI even if you get a win. IMHO Quote Link to comment Share on other sites More sharing options...
Siouxman Posted January 2, 2015 Share Posted January 2, 2015 I like the idea of humans being kept out of it for the most part, but I think they should look into it differently the way the RPI is weighted/used. Teams are rewarded for playing a "soft" schedule. Teams like Bowling Green, Michigan Tech, and Minnesota are virtual locks to make the NCAAs this year because of their schedule. (I am not debating the scheduling of these teams). Each of those teams did average to below average when playing the better teams. Then they do fine against the well below average teams. There needs to be much more of an emphasis for wins against top opponents, and I think they should lower the bar for teams that don't count against the RPI even if you get a win. IMHO I don't think anyone is a lock at this point. On the question of where they should be ranked, Mankato has played the third toughest schedule, Minnesota the 8th, and Michigan Tech the 13th. UND has played the 16th toughest. Each of them has a pretty good record at this point. I don't have a problem with them being ranked where they currently are for RPI. MSU and MTU both appear to be quality teams so far this year. Unless MN absolutely dominates the B1G, which they should, they could get hurt by the end of the year because those teams aren't doing well. There is no perfect system because of theoretical questions about what should be rewarded: complete season body of work; best record in last 8, 10, 12, pick a number of games. I wouldn't be happy seeing a team in the NCAA's that went undefeated their last 12 games of the regular season, but has a record of 12-22. MafiaMan pointed out a case where two teams got hot and cold at the wrong time of year, but those things will happen. If it is only occasionally, that is not too bad in the big picture except for the team that got left home. If a team has a great year but loses their last six games of the year because of season ending or serious injuries, how do you decide if they go to the NCAA's or stay home. Who gets to decide that? You can argue whether RPI or KRACH rankings should be used, but the current system works for both types of teams: good body of season work and getting hot late in the year. You can make the NCAA's based on the entire season body of work or by getting hot at the end and winning your conference tournament for the autobid. I am happy to see MTU and MSU up where they are, Penn St. too (once I get past my anti-B1G attitude). Quote Link to comment Share on other sites More sharing options...
MafiaMan Posted January 2, 2015 Share Posted January 2, 2015 I like the idea of humans being kept out of it for the most part, but I think they should look into it differently the way the RPI is weighted/used. Teams are rewarded for playing a "soft" schedule. Teams like Bowling Green, Michigan Tech, and Minnesota are virtual locks to make the NCAAs this year because of their schedule. (I am not debating the scheduling of these teams). Each of those teams did average to below average when playing the better teams. Then they do fine against the well below average teams. There needs to be much more of an emphasis for wins against top opponents, and I think they should lower the bar for teams that don't count against the RPI even if you get a win. IMHO Did you watch any of the GLI? Michigan Tech is solid. If you want to complain about teams making the tournament, I'd start with Atlantic Hockey and (most likely) Robert Morris. Quote Link to comment Share on other sites More sharing options...
Siouxperfan7 Posted January 2, 2015 Share Posted January 2, 2015 Sure the Pairwise isn't perfect, but it is a heck of a lot better than the alternative. We could have a committee select the 16 teams!!! 1 Quote Link to comment Share on other sites More sharing options...
The Sicatoka Posted January 2, 2015 Share Posted January 2, 2015 KRACH > Pairwise > Humans picking. + Quote Link to comment Share on other sites More sharing options...
siouxforcefans Posted January 2, 2015 Share Posted January 2, 2015 As much as I dislike the Pairwise, I don't have a better solution. People on here advocating for KRACH, but that'll never be perfect, either. Regardless of what system is used, someone is going to get left out who believed they should have been in, leaving the school, and their fans, to complain about that system. Quote Link to comment Share on other sites More sharing options...
siouxnami Posted January 2, 2015 Share Posted January 2, 2015 KRACH is pure mathematics and as fair as possible. Pairwise is math with some arbitrary human interventions that make it less perfect. I don't trust humans screwing with math or selections. 1 Quote Link to comment Share on other sites More sharing options...
InHeavenThereIsNoBeer Posted January 2, 2015 Share Posted January 2, 2015 KRACH is pure mathematics and as fair as possible. Pairwise is math with some arbitrary human interventions that make it less perfect. I don't trust humans screwing with math or selections. What human intervention are you referring to with the pairwise? As far as I know, the calculation is the same throughout the entire season with no human intervention except slight tweaks in between seasons. Quote Link to comment Share on other sites More sharing options...
aaronmanderson3 Posted January 2, 2015 Share Posted January 2, 2015 What human intervention are you referring to with the pairwise? As far as I know, the calculation is the same throughout the entire season with no human intervention except slight tweaks in between seasons. The only human intervention is removing bad wins when calculating the RPI, unless you want to count moving teams for attendance/inter-conference matchups/etc. when deciding regionals. Quote Link to comment Share on other sites More sharing options...
siouxnami Posted January 3, 2015 Share Posted January 3, 2015 The pairwise itself uses arbitrary weightings of common opponents, RPI, and head to head. This determination in and of itself is arbitrary. KRACH is a pure mathematical formula with not arbitrary weightings. Biggest real world difference is the KRACH will not punish a conference that is loaded where as PWR can. What human intervention are you referring to with the pairwise? As far as I know, the calculation is the same throughout the entire season with no human intervention except slight tweaks in between seasons. Quote Link to comment Share on other sites More sharing options...
The Sicatoka Posted January 3, 2015 Share Posted January 3, 2015 Pairwise is the worst possible system out there ... except for all the rest. I'd use KRACH because it's the most objective. What I don't like is that most think some guy named Ken came up with it out of the clear blue sky. KRACH is actually based on a very respected mathematical model called the Bradley-Terry Method. Bradley-Terry was first used to model chess master rankings from tournaments where there wasn't time for every master to play every other master. http://www.encyclopediaofmath.org/index.php/Bradley-Terry_model 1 Quote Link to comment Share on other sites More sharing options...
jimdahl Posted January 7, 2015 Share Posted January 7, 2015 This isn't even a PWR vs KRACH question, because the rankings mentioned are all actually pretty much in agreement right now: Rankings comparison Harvard is #1, UND is #6-7. Where it gets interesting is teams like Minnesota-Duluth which only has the 15th best win percentage, but KRACH/PWR/RPI all put at #3 because of strength of schedule (or Mass.-Lowell getting the opposite effect). I think this is actually a rankings vs polls question. While polls have some relative strengths (they can take changes like injuries or hot streaks into account), they also have relative weaknesses (historical performance bias, lack of voters' consistent exposure to all the teams). I personally find the rankings a bit more credible. Quote Link to comment Share on other sites More sharing options...
Shawn-O Posted January 7, 2015 Share Posted January 7, 2015 The rankings mentioned are all actually pretty much in agreement right now: Rankings comparison Harvard is #1, UND is #6-7. Where it gets interesting is teams like Minnesota-Duluth which only has the 15th best win percentage, but KRACH/PWR/RPI all put at #3 because of strength of schedule (or Mass.-Lowell getting the opposite effect). I think the only reason people think those rankings seem off is that they differ from the polls. While polls have some strengths (they can take changes like injuries or hot streaks into account), they also have pretty severe weakness (historical performance bias, lack of voters' consistent exposure to all the teams). I'd personally believe the rankings. NCAA media polls after the halfway mark are useless. If I'm the SID I'd just create a link to Jim's KRACH page, copy paste, and done. Quote Link to comment Share on other sites More sharing options...
jimdahl Posted January 7, 2015 Share Posted January 7, 2015 NCAA media polls after the halfway mark are useless. If I'm the SID I'd just create a link to Jim's KRACH page, copy paste, and done. That's a good point -- polls are more useful than rankings for the first couple months because historical performance is probably a better indicator of likely strength than the small sample of games. Quote Link to comment Share on other sites More sharing options...
Shawn-O Posted January 7, 2015 Share Posted January 7, 2015 That's a good point -- polls are more useful than rankings for the first couple months because historical performance is probably a better indicator of likely strength than the small sample of games. Jim, at what point does the number of games become statistically relevant enough, in your opinion? Quote Link to comment Share on other sites More sharing options...
jimdahl Posted January 7, 2015 Share Posted January 7, 2015 Jim, at what point does the number of games become statistically relevant enough, in your opinion? I'm not a good enough statistician to answer that correctly. If you think of all of the games as the population and the completed results as the sample, you could probably create a confidence interval for some hypothesis (the ranked quality of a team?). But, the games aren't independent so I'm already lost. What I have done is some analysis as to how well the as-of-today PWR predicts the final PWR, which is what we probably really want to know. That's in When to start looking at PWR (revisited). The bottom line is that it has some predictive power right away in November, but gets continually better. I personally start paying attention after the Christmas break tournaments. Quote Link to comment Share on other sites More sharing options...
Shawn-O Posted January 7, 2015 Share Posted January 7, 2015 I'm not a good enough statistician to answer that correctly. If you think of all of the games as the population and the completed results as the sample, you could probably create a confidence interval for some hypothesis (the ranked quality of a team?). But, the games aren't independent so I'm already lost. What I have done is some analysis as to how well the as-of-today PWR predicts the final PWR, which is what we probably really want to know. That's in When to start looking at PWR (revisited). The bottom line is that it has some predictive power right away in November, but gets continually better. I personally start paying attention after the Christmas break tournaments. You should get back into the consulting game. Quote Link to comment Share on other sites More sharing options...
AZSIOUX Posted January 7, 2015 Share Posted January 7, 2015 That's sweet info jim, thanks again! Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.