Points leaders -- accounting for Strength of Schedule

Discussion of Minnesota Girls High School Hockey

Moderators: Mitch Hawker, east hockey, karl(east)

PACOTACO
Posts: 2
Joined: Fri Feb 01, 2013 11:26 am

Post by PACOTACO »

spending this much time crunching theoretical statistics for GIRLS hs hockey only makes you look like a


Image

just saying :P
MinnGirlsHockey
Posts: 204
Joined: Wed Sep 23, 2009 1:33 am

Post by MinnGirlsHockey »

PACOTACO wrote:spending this much time crunching theoretical statistics for GIRLS hs hockey only makes you look like a


Image

just saying :P
I can't keep up with all this math, but I think some of these models/formulas could be applied to even BOYS hockey and likely extrapolated to other sports as well. You're obviously a GIRLS sports hater based on your sexist comment, but maybe it could even be applied to your favorite BOYS sport. Wouldn't that be cool?

And if you think spending time discussing GIRLS hs hockey is for losers, then why are you spending your time reading this stuff? You must be a LOU-HOOO ZURR-HERR too. Please go away and don't come back.
PACOTACO
Posts: 2
Joined: Fri Feb 01, 2013 11:26 am

Post by PACOTACO »

I'm sorry. Maybe underlining GIRLS was not the correct way to state my point. Did not mean for it to come off as sexist whatsoever. This could without question be applied to the BOYS stats as well(which I would also consider a waste of time).

No one on this forum is Arthur Friedman(look it up) so what relevance does this really have?

Most people are aware that these stats are quite meaningless to college coaches and recruiters. Just seems like a silly thing to spend that much time figuring out??

Plus, who knows how accurate the stats actually are with how they seem to be adjusted on the mngirlsshockeyhub.

The underlining point is its incredible how people obsess with stats. If you watch girls hockey year round(elite, national camp, and season play) the same girls consistently score on the top of the charts. (but yes Im a huge sexist because I spend my free time watching and following GIRLS youth and HS hockey)

Just my two cents, now ill go away and wont come back :cry:
MinnGirlsHockey
Posts: 204
Joined: Wed Sep 23, 2009 1:33 am

Post by MinnGirlsHockey »

PACOTACO wrote:I'm sorry. Maybe underlining GIRLS was not the correct way to state my point. Did not mean for it to come off as sexist whatsoever. This could without question be applied to the BOYS stats as well(which I would also consider a waste of time).

No one on this forum is Arthur Friedman(look it up) so what relevance does this really have?

Most people are aware that these stats are quite meaningless to college coaches and recruiters. Just seems like a silly thing to spend that much time figuring out??

Plus, who knows how accurate the stats actually are with how they seem to be adjusted on the mngirlsshockeyhub.

The underlining point is its incredible how people obsess with stats. If you watch girls hockey year round(elite, national camp, and season play) the same girls consistently score on the top of the charts. (but yes Im a huge sexist because I spend my free time watching and following GIRLS youth and HS hockey)

Just my two cents, now ill go away and wont come back :cry:
Sorry too, I may have gotten a little too defensive but it really came across to me as a "girls sports aren't important, why are you spending time on this?" type thing. I apologize if I misinterpreted it, it was the ALL CAPS and underlining of 'GIRLS' that set me off....
sinbin
Posts: 898
Joined: Thu Oct 29, 2009 11:12 pm

Post by sinbin »

One man's hobby is another man's obsession is another man's Waterloo. And, of course, when I say "man", I mean it in the generic sense to include all men and all women, so as not to be sexist. :wink:
ghshockeyfan
Posts: 6132
Joined: Thu Jun 05, 2003 2:33 pm
Location: Inver Grove Heights, MN
Contact:

Post by ghshockeyfan »

sinbin wrote:One man's hobby is another man's obsession is another man's Waterloo. And, of course, when I say "man", I mean it in the generic sense to include all men and all women, so as not to be sexist. :wink:
Well said - very well said! :D What was amazing to me about the earlier post was the extra time to put the picture in to make the point... :lol:


That aside - long ago I tried to make the point about KRACH SOS being the average RANK instead of the average RATING... At that point in time SSP was crazy good and so much so that their rating was 10x everyone else. What that does to an average is push every team that played SSP to the very top of the average rating list. So, instead, we went to ranking average. Because we don't have a runaway team like those SSP teams from a number of years ago, then I guess the rating average for SOS may again be useful. It just wasn't meaningful back then as such high SOS averages were being achieved just by virtue of seeing SSP 1x on a schedule.
sinbin
Posts: 898
Joined: Thu Oct 29, 2009 11:12 pm

Post by sinbin »

Yes, it's prudent to adjust for outliers whenever possible. One simple means (no pun intended) in the real world is to use median rather average, so you don't have the Bill Gates effect. For those who like both numbers and hockey (and there appear to be more than a few of us), we are living in the golden age of KRACH, SOS-adjusted points, Section realignments, and I'm sure the next ground-breaking statistic percolating in someone's garage or basement or the privacy of their own mind. Bandy and EPDad have had a stimulating discourse on parametric and non-parametric statistics and related topics, much to the delight of some and probably to the chagrin of many. At the day's penultimate moment, we've developed models that approximate reality as closely as possible, experimenting and tweaking until we come up with some that is reasonably close, but also has some intuitiveness and allows for fairly easy calculation. At the very end of the day, though, a hot goalie or a lucky bounce or a team on a wicked streak can often throw all of our math out the window and leave us with some very unpredictable, non-mathematical results.
Bandy
Posts: 153
Joined: Wed Dec 01, 2010 3:35 pm

Post by Bandy »

End of season KRAPPIs. I calculated KRAPPI & a new stat based on goals (KRAPGI) for the top 210 points leaders off of Hub. I’ll post an update in a couple minutes. Here’s a refresher on calculations:

Adjusted Points Per Game = (Points – (PIM*penalty))/(Number of games). The penalty for PIM for KRAPPI = 0.15 (slight increase because the old value didn’t really penalize players much for having lots of penalty minutes). The penalty for KRAPGI (introduced below) is half that value, or 0.075 per PIM.

KRAPPI is the KRAch and Pim-adusted Points Index. KRAPPI = [(Adjusted pts per game) / (AVG SOS)^0.5].

KRAPGI is the KRAch and Pim-adjusted Goals Index. KRAPGI = [(Adjusted GOALS per game) / (AVG SOS)^0.5].

The following two posts will list top 60 KRAPPI scores, and the top 60 KRAPGI scores.
Last edited by Bandy on Mon Feb 18, 2013 10:32 am, edited 1 time in total.
Bandy
Posts: 153
Joined: Wed Dec 01, 2010 3:35 pm

Post by Bandy »

Top 60 KRAPPI ranking players, based on MN Girls Hockey Hub stats as of 2/18/2013 and KRACH SOS as of 2/16/2013.

Player | Team | PtsPerGame_PIMadjusted | KRAPPI | Rank_PPG | Rank_KRAPPI
Kelly Pannek BSM 3.42 5.7 3 1
Nicole Schammel REDWNG 4.05 5.3 1 2
Dani Cameranesi BLAKE 3.56 5.0 2 3
Lisa Marvin WRRD 2.96 4.5 7 4
Amy Petersen MNTKA 1.91 4.5 38 5
Laura Bowman MNTKA 1.79 4.2 46 6
Kayla Gardner WRRD 2.69 4.1 10 7
Kiersten Falck BLAINE 2.49 4.0 13 8
Katie Rooney BUFF 2.45 4.0 14 9
Amy Menke SHAK 2.43 3.8 15 10
Megan Wolfe EAGAN 2.30 3.8 16 11
Brittany Wheeler BSM 2.27 3.8 19 12
Corbin Boyd HPKNS 2.15 3.7 23 13
Reagan Haley REDWNG 2.79 3.6 8 14
Caitlin Reilly BSM 2.17 3.6 21 15
Nina Rodgers HPKNS 2.00 3.5 33 16
Kate Flug RSVL 2.02 3.5 31 17
Jessica Aney RCHCEN 3.17 3.4 4 18
Emilie Brigham ANOKA 1.98 3.4 35 19
Lauren Hespenheide SHAK 2.17 3.4 21 20
Taylor Williamson EDINA 1.51 3.3 73 21
Carly Moran WINONA 3.02 3.3 6 22
Kate Schipper* BRECK 2.10 3.3 25 23
Brianna Breiland CRKSTN 2.54 3.2 11 24
Dana Rasmussen DODGE 2.50 3.2 12 25
Demi Gardner WRRD 2.02 3.1 30 26
Lindsay Roethke BUFF 1.89 3.1 39 27
Charly Dahlquist EDENPR 1.44 3.1 83 28
Karlie Lund BLAKE 2.15 3.0 24 29
Katherine Aney* RCHCEN 2.75 3.0 9 30
Emily Stegora REDWNG 2.28 3.0 18 31
Makayla Sterrett LUVRNE 3.12 2.9 5 32
Marissa Odell ARMCOO 1.74 2.9 50 33
Darby Dodds DODGE 2.28 2.9 17 34
Dani Sibley NWRCTY 2.08 2.9 26 35
Angie Heppelmann EDENPR 1.34 2.8 100 36
Lindsey Coleman BURNS 1.68 2.8 53 37
Briita Nelson BURNS 1.68 2.8 54 38
Emma Terres ARMCOO 1.64 2.7 57 39
Samantha Donovan IRON 1.95 2.7 36 40
Paige Haley REDWNG 2.05 2.7 28 41
Kelsey Cline BJEFF 1.53 2.6 68 42
Paige Skaja BURNS 1.57 2.6 61 43
Emily Bergland T.R.F. 2.08 2.6 27 44
Mikayla Goodin ANDVR 1.59 2.6 59 45
Kaitlin Storo CHASKA 1.44 2.5 83 46
Haley Mack E.G.F. 2.03 2.5 29 47
Brooke Madsen EAGAN 1.53 2.5 70 48
Kathryn Larson MNDSVW 1.56 2.5 63 49
Alexis Joyce LKVL-N 1.50 2.5 74 50
Blair Parent ANOKA 1.43 2.5 88 51
Hannah Okerstrom STLLW 1.43 2.5 86 52
Christi Vetter LKVL-N 1.47 2.4 80 53
Sierra Hanowski WRRD 1.58 2.4 60 54
Jordan McLaughlin GRG 1.62 2.4 58 55
Sam Swanstrom BLAINE 1.49 2.4 77 56
Catie Skaja N. P. 2.00 2.4 32 57
Lindsay Paschke N. P. 1.99 2.4 34 58
Riley Viner SSP 1.64 2.4 55 59
Sylvia Marolt T.R.F. 1.86 2.3 42 60
Bandy
Posts: 153
Joined: Wed Dec 01, 2010 3:35 pm

Post by Bandy »

Top 60 KRAPGI (Goals index) ranking players, based on MN Girls Hockey Hub stats as of 2/18/2013 and KRACH SOS as of 2/16/2013.


Player | Team | Goals | GoalsPerGame_PIMadjusted | KRAPGI | Rank_GPGadjusted | Rank_KRAPGI
Nicole Schammel REDWNG 61 2.39 3.1 1 1
Dani Cameranesi BLAKE 49 2.17 3.1 3 2
Amy Menke SHAK 40 1.71 2.7 5 3
Brianna Breiland CRKSTN 49 2.05 2.6 4 4
Carly Moran WINONA 57 2.34 2.6 2 5
Laura Bowman MNTKA 26 1.02 2.4 36 6
Katie Rooney BUFF 37 1.44 2.4 10 7
Kelly Pannek BSM 27 1.34 2.2 12 8
Megan Wolfe EAGAN 26 1.25 2.1 15 9
Nina Rodgers HPKNS 27 1.16 2.0 22 10
Taylor Williamson EDINA 24 0.89 2.0 48 11
Dana Rasmussen DODGE 39 1.56 2.0 8 12
Amy Petersen MNTKA 21 0.83 1.9 57 13
Kayla Gardner WRRD 33 1.22 1.9 18 14
Kate Flug RSVL 28 1.09 1.9 28 15
Emilie Brigham ANOKA 26 1.07 1.9 30 16
Brittany Wheeler BSM 28 1.10 1.8 25 17
Jessica Aney RCHCEN 40 1.67 1.8 6 18
Paige Skaja BURNS 27 1.03 1.7 34 19
Caitlin Reilly BSM 27 1.03 1.7 34 20
Lauren Hespenheide SHAK 30 1.09 1.7 29 21
Charly Dahlquist EDENPR 23 0.80 1.7 70 22
Haley Mack E.G.F. 35 1.34 1.7 11 23
Rebekah Smith ORONO 33 1.24 1.6 16 24
Dani Sibley NWRCTY 31 1.16 1.6 21 25
Brittany Sticha N. P. 31 1.33 1.6 13 26
Kaitlin Storo CHASKA 24 0.90 1.6 46 27
Makayla Sterrett LUVRNE 30 1.64 1.6 7 28
Hannah Okerstrom STLLW 23 0.90 1.5 47 29
CoCo Piche E.G.F. 32 1.23 1.5 17 30
Taryn Juberien WASECA 35 1.48 1.5 9 31
Lisa Marvin WRRD 25 0.98 1.5 39 32
Lindsey Coleman BURNS 23 0.90 1.5 45 33
Sylvia Marolt T.R.F. 33 1.19 1.5 19 34
Kaitlyn Klein WYZT 19 0.73 1.4 89 35
Karleigh Wolkerstorfer BMDJ 27 1.09 1.4 27 36
Darby Dodds DODGE 28 1.14 1.4 23 37
Angie Heppelmann EDENPR 18 0.67 1.4 104 38
Kate Schipper* BRECK 21 0.91 1.4 44 39
Samantha Donovan IRON 26 1.02 1.4 36 40
Annie Pumper N-FLD 30 1.19 1.4 20 41
Carley Grunewald AUSTIN 32 1.27 1.4 14 42
Mary Turitto PRK-CG 21 0.80 1.4 71 43
Kathryn Larson MNDSVW 23 0.86 1.4 54 44
Mikayla Goodin ANDVR 21 0.82 1.3 63 45
Demi Gardner WRRD 23 0.87 1.3 52 46
Marissa Odell ARMCOO 20 0.80 1.3 67 47
Kiersten Falck BLAINE 23 0.82 1.3 59 48
Katie Swanstrom BLAINE 23 0.82 1.3 61 49
Reagan Haley REDWNG 27 1.01 1.3 38 50
Corbin Boyd HPKNS 19 0.75 1.3 80 51
Reilly Fawcett PCTHRM 26 1.07 1.3 32 52
Kelsey Cline BJEFF 20 0.75 1.3 84 53
Sam Swanstrom BLAINE 24 0.78 1.3 74 54
Sarah Bobrowski H-M 19 0.75 1.2 82 55
Blair Parent ANOKA 20 0.71 1.2 92 56
Sena Hanson IRON 23 0.89 1.2 49 57
Kendra Goodrich RSMNT 20 0.75 1.2 84 58
Katherine Aney* RCHCEN 18 1.13 1.2 24 59
Leah Elledge CH PRK 19 0.71 1.2 93 60
Bandy
Posts: 153
Joined: Wed Dec 01, 2010 3:35 pm

Post by Bandy »

This post addresses the KRACH RATING vs KRACH SOS RANK discussion LuckeyEP & I had a while back.

First, AVG KRACH RANK is what I used. It's the average KRACH rank of opponents. Lower number = more highly ranked team. Find them here, under the column heading AVG: http://www.bgoski.com/KRACH_SOS.htm

Here are some stats for Minnetonka (toughest schedule), and 3 teams with schedules that were 'twice as easy' (AVG RANK = twice that of Tonka) (Eagan, BSM, and Robbinsdale-Armstrong-Cooper); and one team with a schedule 3.2x as 'easy' as Tonka's, based on AVG RANK of opponents

Stat | Tonka's opponents | Eagan's opponents | BSM's opponents | RobAC's Opponents | RW's opponents
avg krach RANK of opponents | 18.32 | 36.70 | 36.25 | 36.48 | 59.32
Ratio to Tonka: avg krach RANK | 1 | 2.00 | 1.98 | 1.99 | 3.24
Square root, ratio to Tonka RANK | 1 | 1.42 | 1.41 | 1.41 | 1.80

A straight-up SOS adjustment using AVG RANK would mean that Wolf, Panic, or Terres would have to score two points to equal a Peterson point. Schammel would have to scorer 3.2 for every one of Peterson's. KRAPPI is a bit more conservative; the square root adjustment means Wolf, Panic, or Terres would have to score about 1.41 to every one of Peterson's to match her scoring prowess. For Schammel, that ratio is 1.8.

Next, KRACH AVG RATING for these five teams.

Stat | Tonka's opponents | Eagan's opponents | BSM's opponents | RobAC's Opponents | RW's opponents
avg krach RATING of opponents | 198.36 | 83.14 | 138.59 | 59.88 | 24.91
Ratio to Tonka: avg krach RATING | 1 | 2.4 | 1.4 | 3.3 | 8.0
Square root, ratio to Tonka RATING | 1 | 1.5 | 1.2 | 1.8 | 2.8

The results are a little noisier. The three teams identified as 'twice as easy' schedules by RANK come in at 2.4, 1.4, and 3.3 times as easy by RATING. A big factor here is how many times a team plays Tonka, which has a KRACH Rating of 916 (a high outlier). Eagan played them once; BSM twice; and Robbinsdale zero times. Quite a spread in the RATING-based SOS as a result. And Red Wing = 8 times 'easier' of schedule, meaning Schammel would have to score 8 points for every Peterson point. I honestly thought it would be even more biased than this. Taking square roots, again, levels the ice rink somewhat, so it's not so skewed toward Tonka.

I'd say the nonparametric (RANK-based) model is less subject to the influence of outliers and skewed distributions. Indeed, that's why the field was invented. The parametric model could be made to behave, with a transformation, but then it wouldn't preserve the magnitude of SOS RATING differnces.

This was a fun exercise. I'm turning my attention to the D1 landscape next year--as a fan, not a statistician. If anyone wants to adopt KRAPPI, it's in the public domain. :wink:
luckyEPDad
Posts: 416
Joined: Wed Nov 25, 2009 3:31 pm

Post by luckyEPDad »

A test of your assertion is to compute the KRAPPI rating for subsets of a players games. Using Amy Peterson from Minnetonka as an example. Compute here KRAPPI rating against BSM, Warroad, EP, Hill-Murray, Buffalo, Edina, Blake and Hopkins (13 games) and compare the result to games against Apple Valley, Roseau, Duluth, Eagan, Andover, Anoka, Wayzata, Centennial, Breck and Chaska (12 games). If the SOS adjustment is correct the results should be similar.

To get reasonable results you'd have to do this for dozens of players. If game results are in a workable digital format that wouldn't be a problem, but doing the work by hand is a daunting task.
Bandy
Posts: 153
Joined: Wed Dec 01, 2010 3:35 pm

Post by Bandy »

luckyEPDad wrote:A test of your assertion is to compute the KRAPPI rating for subsets of a players games. Using Amy Peterson from Minnetonka as an example. Compute here KRAPPI rating against BSM, Warroad, EP, Hill-Murray, Buffalo, Edina, Blake and Hopkins (13 games) and compare the result to games against Apple Valley, Roseau, Duluth, Eagan, Andover, Anoka, Wayzata, Centennial, Breck and Chaska (12 games). If the SOS adjustment is correct the results should be similar.

To get reasonable results you'd have to do this for dozens of players. If game results are in a workable digital format that wouldn't be a problem, but doing the work by hand is a daunting task.
Thanks for your comments, Lucky. Another way of thinking about this is 'model skill.' Algorithms aside, how 'skilled' is KRAPPI in terms of identifying offensive prowess. Having watched many of these players in the Advanced / HP process, Fall Elite Leauge, and regular season, I'd say its doing pretty good. But I'm not the state's biggest hockey hound. If a season-aggregate model does pretty good, is it worth the effort of compiling a game-specific model?

While I agree that game-specific strength of opponent would be valuable, I'll leave that test for my successor. I'm retiring.

Enjoy the Tourney!
Post Reply