What's new
By:
USCHO Fan Forum

This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

  • The USCHO Fan Forum has migrated to a new plaform, xenForo. Most of the function of the forum should work in familiar ways. Please note that you can switch between light and dark modes by clicking on the gear icon in the upper right of the main menu bar. We are hoping that this new platform will prove to be faster and more reliable. Please feel free to explore its features.

Pairwise and Bracketology 2013 Edition

  • Thread starter Thread starter Priceless
  • Start date Start date
Status
Not open for further replies.
Re: Pairwise and Bracketology 2013 Edition

I am off to Boston in a few hours so this will be the last time I am posting from a real computer. Based on the results I expect this is the bracketology I'd put forth.

Code: 
[B]Manchester (UNH)	Providence (Brown)	Toledo (BGSU)		Grand Rapids (Michigan)[/B]
Boston C		Quinnipiac		Miami			Minnesota
Yale			Lowell			Niagara			N Dakota
New Hamp		St Cloud		Mankato			Notre Dame
Union			W Michigan		Denver			Michigan

Now that I've written that, my predictions will start going wrong this afternoon...
 
Re: Pairwise and Bracketology 2013 Edition

Dream bracket for the committee, Priceless. No flights to Manchester, 2 to Providence, 2 to Toledo, 2 to Rapids. I was just thinking about Michigan this morning, too. Wow. What a reward for Minnesota for being #2 overall - playing what would be the hottest team in the country.
 
Re: Pairwise and Bracketology 2013 Edition

I agree this is what the brackets will probably end up (assuming Michigan wins the CCHA Tourney), but I still wish they would spread out the WCHA teams into the FOUR Regions
 
Re: Pairwise and Bracketology 2013 Edition

A really basic question that I've just realized I never knew the answer to and never inquired about:

How long have the selection committee used the PWR system, and have they ever veered from it for at-large selection or seeding? I've only been paying attention to this kind of detail for the last decade or so, and I can't think of an instance where they've veered from the formula.
 
Re: Pairwise and Bracketology 2013 Edition

ExileOnDaytonStreet said:
A really basic question that I've just realized I never knew the answer to and never inquired about:

How long have the selection committee used the PWR system, and have they ever veered from it for at-large selection or seeding? I've only been paying attention to this kind of detail for the last decade or so, and I can't think of an instance where they've veered from the formula.
Click to expand...

They've been using it in full since the 1994-95 season. In 1993-94 (and presumably some years before, but I don't know how many), they mostly used RPI and only used PWR to teams within a small RPI margin of one another on the bubble (i.e. teams within 0.01 or something like that).

Through 1998-ish, if I remember correctly, the committee got the PWR more as a grid of comparisons, and it could be more important for a bubble team to have bubble comparison wins than to have a good overall PWR rank. Somewhere around 1999 or 2000, they appear to have shifted to doing a straight-ranked PWR.
 
Re: Pairwise and Bracketology 2013 Edition

Does anyone know if the Thursday outcomes will have any effect on Yale's possible lock status, I understand that they are right on the edge of being a lock. Thanks for all of the analysis!
 
Re: Pairwise and Bracketology 2013 Edition

LTsatch said:
Does anyone know if the Thursday outcomes will have any effect on Yale's possible lock status, I understand that they are right on the edge of being a lock. Thanks for all of the analysis!
Click to expand...
RHamilton's earlier post listed what the updated possibilities would be after each of the possible Thursday outcomes. None of them list Yale as a lock.
RHamilton said:
Hey all,

I ran all 393,216 outcomes similar to JimDahl, but went a step further and assigned a percent likelihood using KRACH on each individual outcome. I took a brief look, and it looks like my unweighted outcomes match up with JimDahl's

Here are KRACH weighted likelihoods for all seeds:
http://pwr.reillyhamilton.com/

Note that "seed" refers to the PWR seed (after auto-bids have been accounted for), not the PWR rank. Ties broken using RPI.

Individual breakdowns for each of the outcomes tomorrow:

WCHA 'Quarterfinal' 1:
WCHA 'Quaterfinal' 2:
if Minnesota State won:
If Wisconsin won:



I'm working on some alternate ways of analyzing the data I've generated. I hope to release a dynamic "what-if" machine after tomorrow's games when the number of outcomes is a bit more manageable and less likely to crash my webhost's database server.

EDIT: Fixed typo per lugnut92 below :)
EDIT 2: Switched domain names.
Click to expand...
 
Re: Pairwise and Bracketology 2013 Edition

Up-to-date KRACH weighted breakdown:
http://pwr.reillyhamilton.com/

I just regenerated the KRACH figures, so the weightings are a bit different than the previously posted wis-cc.html file. The unweighted percents and number of outcomes are unchanged.

I'll post some scenarios for the early games tomorrow soon.
 
Re: Pairwise and Bracketology 2013 Edition

Updated, same model values as inputs.

Code: 
   NAME   NCAA     S1     S2     S3     S4   CONF  ATLRG    TUC
1    AA 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
2    AF 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
3    AH 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
4    AI 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
5    Ak 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
6    Ar 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
7    BC 1.0000 0.4229 0.5764 0.0007 0.0000 0.2688 0.7312 1.0000
8    BG 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
9    Bn 0.1201 0.0000 0.0000 0.0000 0.1201 0.1201 0.0000 1.0000
10   BS 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
11   BU 0.1849 0.0000 0.0000 0.0007 0.1842 0.1650 0.0199 1.0000
12   By 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
13   Ca 0.2237 0.0000 0.0000 0.0000 0.2237 0.2237 0.0000 0.0000
14   CC 0.0790 0.0000 0.0000 0.0000 0.0790 0.0790 0.0000 1.0000
15   Cg 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
16   Ck 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
17   Cr 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
18   Ct 0.2615 0.0000 0.0000 0.0000 0.2615 0.2615 0.0000 0.6066
19   Da 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
20   DU 0.9956 0.0000 0.0038 0.8566 0.1352 0.0000 0.9956 1.0000
21   FS 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
22   Ha 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
23   HC 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
24   LS 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
25   MA 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
26   MD 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
27   Me 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
28   Mh 0.1878 0.0000 0.0000 0.0000 0.1878 0.1878 0.0000 0.0000
29   Mi 0.0951 0.0000 0.0000 0.0000 0.0951 0.0951 0.0000 0.3234
30   Mk 0.9993 0.0000 0.0940 0.8962 0.0091 0.0000 0.9993 1.0000
31   ML 1.0000 0.3129 0.6767 0.0104 0.0000 0.3110 0.6890 1.0000
32   Mm 1.0000 0.9033 0.0967 0.0000 0.0000 0.4271 0.5729 1.0000
33   Mn 1.0000 1.0000 0.0000 0.0000 0.0000 0.5355 0.4645 1.0000
34   Mr 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
35   MS 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
36   MT 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
37   ND 1.0000 0.0000 0.5257 0.4743 0.0000 0.0000 1.0000 1.0000
38   NE 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
39   NH 1.0000 0.0000 0.6338 0.3662 0.0000 0.0000 1.0000 1.0000
40   Ni 0.9887 0.0560 0.3854 0.3904 0.1569 0.3270 0.6617 1.0000
41   NM 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
42   NO 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
43   Nt 0.8427 0.1546 0.0470 0.3142 0.3269 0.3404 0.5023 1.0000
44   OS 0.1374 0.0000 0.0000 0.0000 0.1374 0.1374 0.0000 1.0000
45   Pn 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
46   PS 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
47   Pv 0.2562 0.0000 0.0000 0.0006 0.2556 0.2552 0.0010 1.0000
48   Qn 1.0000 1.0000 0.0000 0.0000 0.0000 0.4961 0.5039 1.0000
49   RM 0.0072 0.0000 0.0000 0.0000 0.0072 0.0000 0.0072 1.0000
50   RP 0.1242 0.0000 0.0000 0.0000 0.1242 0.0000 0.1242 1.0000
51   RT 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
52   SC 0.9797 0.0379 0.4296 0.3488 0.1634 0.2286 0.7511 1.0000
53   SH 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
54   SL 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
55   Un 0.5562 0.0000 0.0000 0.1382 0.4180 0.2518 0.3044 1.0000
56   Vt 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
57   Wi 0.2046 0.0000 0.0000 0.0482 0.1564 0.1569 0.0477 1.0000
58   WM 0.7756 0.0000 0.0000 0.0000 0.7756 0.0000 0.7756 1.0000
59   Ya 0.9805 0.1124 0.5309 0.1545 0.1827 0.1320 0.8485 1.0000

I'll update either VERY late Friday or Saturday morning.

I will also post pairwise code on Saturday with hopes that it will help with future module building.

edit: as in, entice help for the future with the goal of a full on model based monte carlo simulation for use earlier in the season.
 
Last edited:
Re: Pairwise and Bracketology 2013 Edition

Patman said:
I will also post pairwise code on Saturday with hopes that it will help with future module building.

edit: as in, entice help for the future with the goal of a full on model based monte carlo simulation for use earlier in the season.
Click to expand...

Cool, will be fun to see others models. Yours is in R? My admittedly sloppy PHP/MySQL database-driven model is here:
https://github.com/hamilr2/NCAA-Hockey-PWR-predictor

Obviously needs a lot of set-up and a DB schema to run, but if you like looking at methodology in an iterative language, it's worth a look. Maybe I'll prepare a readme later.

Team setup and iteration is done in pwr.php, relevant PWR calculations happen in the doPWR function here:
https://github.com/hamilr2/NCAA-Hockey-PWR-predictor/blob/master/include.php#L295
 
Re: Pairwise and Bracketology 2013 Edition

I looked at RHamilton's most recent chart. The following schools have a chance yet at the last 2 #1 seeds:
#3 overall: Miami (~55%), BC (~19%), Lowell (~18%), Yale (~1.5%), NoDame (~5%) St Cloud (~1%)
#4 overall: Miami (~33%), BC (~23%), Lowell (~10%), Yale (~15%), Niagara (~6%), St Cloud (~3%), NoDame (~10%).

That is 7 schools still with a chance at a #1 seed, besides the 2 guarantees. Wow. Admittedly, Niagara and St Cloud the smallest, but still.
 
Re: Pairwise and Bracketology 2013 Edition

I found an odd thing in JimDahl's blog tonight. The assumption here has been that St Cloud can now clinch with a win.

However, on his blog, updated with today's results, we find this for St Cloud, under the 1 win column:

#13 0.0%

Which in interesting to me, because the #14 row is totally empty. And, if that really means 0.01%, where there might be a strange case yet, then a win tomorrow afternoon does NOT guarantee them in, because we all know that CC, Mich, Prov and Brown could yet win tourneys, in which case #13 in the PWR misses.

Can, RHamilton or PatMan add more information?
 
Re: Pairwise and Bracketology 2013 Edition

RHamilton said:
Cool, will be fun to see others models. Yours is in R? My admittedly sloppy PHP/MySQL database-driven model is here:
https://github.com/hamilr2/NCAA-Hockey-PWR-predictor

Obviously needs a lot of set-up and a DB schema to run, but if you like looking at methodology in an iterative language, it's worth a look. Maybe I'll prepare a readme later.

Team setup and iteration is done in pwr.php, relevant PWR calculations happen in the doPWR function here:
https://github.com/hamilr2/NCAA-Hockey-PWR-predictor/blob/master/include.php#L295
Click to expand...

right now I'm not running a model I'm using Robin Lock's CHODR model... if I were to run things... and admittedly i operate as a Bayesian, I'd probably do some variant of Poisson regression using the hessian matrix as a measure of parametric uncertainty and then simulate from that assuming normality... I will then get away with any difficult posterior distribution construction and related Markov Chain Monte Carlo inference issues... anyhow...

no, what I will post is straight through, my code to calculate pairwise. I want to build a bigger simulator. I argue that parts would likely have to be written in another language (C or fortran) to achieve desired efficiency... but i seem to say that about most of the stuff i do right now :)
 
Re: Pairwise and Bracketology 2013 Edition

I've got PWR code written in Stata if anyone is interested. It's not yet ready to do extensive Monte Carlo simulations yet, because it's w-a-a-a-y too slow on the CommOpp criterion. Everything else is working pretty well, much faster than I'd have expected.
 
Re: Pairwise and Bracketology 2013 Edition

Numbers said:
I found an odd thing in JimDahl's blog tonight. The assumption here has been that St Cloud can now clinch with a win.

However, on his blog, updated with today's results, we find this for St Cloud, under the 1 win column:

#13 0.0%

Which in interesting to me, because the #14 row is totally empty. And, if that really means 0.01%, where there might be a strange case yet, then a win tomorrow afternoon does NOT guarantee them in, because we all know that CC, Mich, Prov and Brown could yet win tourneys, in which case #13 in the PWR misses.

Can, RHamilton or PatMan add more information?
Click to expand...

I see 10 outcomes (out of 49,152) where St. Cloud ends up #17 when winning their semi.

The scenarios all involve:
- Niagara losing to Mercyhurst or Connecticut
- Miami losing to Ohio State or Notre Dame
- Union DEFEATING Brown in the ECAC Final
- Quinnipiac and Yale TYING in the ECAC Consolation
- 8 of 10 feature BU winning hockey east, the other two have Providence beating BU -- the Providence wins require Connecticut to beat Niagara
- CC winning the WCHA

EDIT: Putting this in the Pairwise Predictor, I see that St. Cloud ends up in a 4-way tie for #9, becomes #12 on tiebreakers. Union finishes #13, and BU finishes #14, but they both move up one seed to make room for Ohio State, Colorado College, and Mercyhurst. This bumps SCSU down to #17.

Ugly is all I can say!
 
Last edited:
Re: Pairwise and Bracketology 2013 Edition

To help all the pairwise prognosticators out there, I put together this:

http://pwr.reillyhamilton.com/query.php
Simply input your choice, or fill in games as the day goes on, and receive the weighted (and unweighted) outcomes.

It may take up to a minute or so for some queries. The more teams you switch from "Don't Care" to something else, the quick it will run. Impossible conditions (a team winning a final than won neither semi) will not return any results. Results are cached, though, so repeated visits to the same query should be instantaneous. Use the links to share.

Let me know if you have any issues running it. We'll see how long it stays up, and if my webhost complains!
 
Status
Not open for further replies.
Back
Top