Announcement

Collapse
No announcement yet.

pairwise -- post jan 8th

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Shirtless Guy
    replied
    Re: pairwise -- post jan 8th

    Originally posted by ticapnews View Post
    I haven't looked that far ahead yet (I so miss the PairWise calculator at slack.net) but there appears to be an RPI cushion at 12 that it will take some unusual results for any of them to miss. The bubble really is down to 13-18.
    There is a huge RPI cliff after 10 that is hard for anyone to overcome. UML and Harvard also have a great shot if they don't get swept and there are a ton of AQs below 12 in the final RPI.

    I missed the slack.net page so much that I built my own excel calculator. Tech Hockey Guide is hoping to have a working web based pairwise calculator with ability to enter results for all remaining games by tomorrow night (if all goes well). I'll keep this thread posted. Big issue might be handling a load if enough people like us want to fiddle with things. As nice as CHN's calculator has been, it has never included conference tournaments (until games are scheduled) and was very difficult to tweak results (having to start from scratch every time).

    Leave a comment:


  • ticapnews
    replied
    Originally posted by Shirtless Guy View Post
    I could be wrong but it does appear that it is impossible for to get the top 8 & Notre Dame out of the tournament because the only way they finish below 10, is with enough AQs above them to guarantee the cut line is below them.
    I haven't looked that far ahead yet (I so miss the PairWise calculator at slack.net) but there appears to be an RPI cushion at 12 that it will take some unusual results for any of them to miss. The bubble really is down to 13-18.

    Leave a comment:


  • Patman
    replied
    Re: pairwise -- post jan 8th

    Originally posted by Shirtless Guy View Post
    You didn't read the rules of what that link is did you? Not to mention anyone that can finish 11th could technically be bumped by 6 lower ranked teams winning their conference tournament.

    While it's likely that those 20,000 simulations are the most of the highest likelihood outcomes, I wouldn't say anyone below #6 is actually a lock.
    No, but really, Monte Carlo simulation implies the chances are god awfully small.

    edit: technically, they aren't the "most likely"... they're just draws from a generated distribution. For instance, I have a hard time believing the situation that puts Mankato at 3rd overall is more likely than the others in between with no results. Its just a generation in the tails... therefore (probably) not likely yet be drawn.
    Last edited by Patman; 03-09-2016, 02:52 PM.

    Leave a comment:


  • Shirtless Guy
    replied
    Re: pairwise -- post jan 8th

    Originally posted by ticapnews View Post
    My story will be posted shortly but teams ranked 1-12 at the start of conference quarters have made the tournament 94% of the time. Of teams in the top 10 only the 10th ranked 2005 Dartmouth team missed the tournament.
    I could be wrong but it does appear that it is impossible for to get the top 8 & Notre Dame out of the tournament because the only way they finish below 10, is with enough AQs above them to guarantee the cut line is below them.

    Leave a comment:


  • ticapnews
    replied
    Re: pairwise -- post jan 8th

    My story will be posted shortly but teams ranked 1-12 at the start of conference quarters have made the tournament 94% of the time. Of teams in the top 10 only the 10th ranked 2005 Dartmouth team missed the tournament.

    Leave a comment:


  • Shirtless Guy
    replied
    Re: pairwise -- post jan 8th

    Originally posted by Tater View Post
    #1-8 & #10 are 100% locks. #9 & #11 are 99%. So 1-11 are essentially locks and #12 Harvard is at 95%. CHN has done the math.

    http://www.collegehockeynews.com/rat...lityMatrix.php
    You didn't read the rules of what that link is did you? Not to mention anyone that can finish 11th could technically be bumped by 6 lower ranked teams winning their conference tournament.

    While it's likely that those 20,000 simulations are the most of the highest likelihood outcomes, I wouldn't say anyone below #6 is actually a lock.

    Leave a comment:


  • Tater
    replied
    Re: pairwise -- post jan 8th

    Originally posted by JimDahl View Post
    I would call the top 5 locks, and just added Denver to the list of teams I expect to see in the tournament:
    http://collegehockeyranked.com/2016/...aa-tournament/

    It's tough to decide when to call these, though. Think of a presidential election. With 60% of the votes in, and 80% of them going to one candidate (assuming reasonable demographics distribution), do you say the candidate has won in a landslide or do you say it's not over because the loser can still win if he gets every single remaining vote? Obviously the former. Denver is that winning candidate--it's mathematically possible for them to miss, but I wouldn't worry about it.

    After this weekend, things will be tight enough that I'll switch over to pure mathematical certainty.
    #1-8 & #10 are 100% locks. #9 & #11 are 99%. So 1-11 are essentially locks and #12 Harvard is at 95%. CHN has done the math.

    http://www.collegehockeynews.com/rat...lityMatrix.php

    Leave a comment:


  • JimDahl
    replied
    Re: pairwise -- post jan 8th

    I would call the top 5 locks, and just added Denver to the list of teams I expect to see in the tournament:
    http://collegehockeyranked.com/2016/...aa-tournament/

    It's tough to decide when to call these, though. Think of a presidential election. With 60% of the votes in, and 80% of them going to one candidate (assuming reasonable demographics distribution), do you say the candidate has won in a landslide or do you say it's not over because the loser can still win if he gets every single remaining vote? Obviously the former. Denver is that winning candidate--it's mathematically possible for them to miss, but I wouldn't worry about it.

    After this weekend, things will be tight enough that I'll switch over to pure mathematical certainty.

    Leave a comment:


  • LTsatch
    replied
    Re: pairwise -- post jan 8th

    Originally posted by Shirtless Guy View Post
    unless they changed their method playoffstatus.com doesn't use pwr to rank teams and can't be trusted for NCAA tournament predicting.
    I was wondering that when they had Yale at 7 instead of 8, i guess they use their own power rankings. That said, Yale just jumped over Michigan to 7 with the second Ohio State loss.

    Leave a comment:


  • Shirtless Guy
    replied
    Re: pairwise -- post jan 8th

    Originally posted by LTsatch View Post
    Playoffstatus.com has the top five teams as 100% locks even if they lose out the rest of the way, they also have 6-8 with a less than 1% chance of falling out even if they lose out the rest of the season.
    unless they changed their method playoffstatus.com doesn't use pwr to rank teams and can't be trusted for NCAA tournament predicting.

    Leave a comment:


  • LTsatch
    replied
    Re: pairwise -- post jan 8th

    Playoffstatus.com has the top five teams as 100% locks even if they lose out the rest of the way, they also have 6-8 with a less than 1% chance of falling out even if they lose out the rest of the season.

    Leave a comment:


  • Biddco
    replied
    Re: pairwise -- post jan 8th

    Now I can't wait for the next bracketology to be released.

    Leave a comment:


  • Patman
    replied
    Re: pairwise -- post jan 8th

    Originally posted by Shirtless Guy View Post
    to be fair, a WCHA/NCHC team could go 4-0 over the next 2 weeks and be done for the season. Big Ten could go 0-7 starting friday...ECAC/HEA/AHA are the ones that are done...the west still has some regular season games.
    right... i'm just explaining general principle

    Leave a comment:


  • Shirtless Guy
    replied
    Re: pairwise -- post jan 8th

    Originally posted by Patman View Post
    Its a lot of nibbling at the margins anyhow. First we don't really have "enough" data... and I mean statistically enough... second... its harder to drop out at tournament time than it is to jump in.

    Now, college with its best of 3 format does make it doable to drop two results without a win. However, a borderline team getting hot and going 4-1 (read as first round+ QF sweep+loss in SF) is possible. For any team the most games they can drop NET is two games. (Yes, a team can go 2-3 by playing two playoff series in sequence... but you get my point).

    Nobody is "fire proof" per se... I wouldn't bet heavily on all teams 12th and above... but it isn't easy.
    to be fair, a WCHA/NCHC team could go 4-0 over the next 2 weeks and be done for the season. Big Ten could go 0-7 starting friday...ECAC/HEA/AHA are the ones that are done...the west still has some regular season games.

    Leave a comment:


  • Patman
    replied
    Re: pairwise -- post jan 8th

    Originally posted by ticapnews View Post
    None. For the first eight seasons of this current format we had the CHA taking up a spot. For those eight seasons the top 12 teams qualified for 89 of 96 spots (92.7%). For the three seasons without the conference it was 34 of 36 spots (94.4%).
    Its a lot of nibbling at the margins anyhow. First we don't really have "enough" data... and I mean statistically enough... second... its harder to drop out at tournament time than it is to jump in.

    Now, college with its best of 3 format does make it doable to drop two results without a win. However, a borderline team getting hot and going 4-1 (read as first round+ QF sweep+loss in SF) is possible. For any team the most games they can drop NET is two games. (Yes, a team can go 2-3 by playing two playoff series in sequence... but you get my point).

    Nobody is "fire proof" per se... I wouldn't bet heavily on all teams 12th and above... but it isn't easy.

    Leave a comment:

Working...
X