Cross classification wins and losses

A recent commentary by Tim Martinez in the Vancouver Columbian whined about the Prairie girls being ripped off by the RPI rating the third week into the season. I agree that Prairie is a lot better than the RPI rank. However the column wandered from there into ‘fixing’ RPI by including factors for classification. So is this a problem? Are crappy teams in a higher division gaming the system (intentionally or not) by beating up on even crappier teams in lower divisions? When a 4A team plays a 1A team should the 4A team be penalized?

Looking at last season here are the cross-classification W-L record for girls.

v 4A v 3A v 2A v 1A v 2B v 1B
4A 115-100 26-24 1-6 1-0 1-0
3A 100-115 81-60 11-21 2-3 0-0
2A 26-24 60-81 73-77 8-10 3-2
1A 6-1 21-11 77-73 56-38 25-17
2B 0-1 3-2 10-8 38-56 73-55
1B 0-1 0-0 2-3 17-25 55-73

Let’s look at those 1A v 4A games. Only seven, but non-league games are not scheduled according to randomized experimental design anyway. La Center had two wins over bad Heritage and Battle Ground teams. I guess those crappy 4A teams in Clark county should look for an real 1A pushover instead of La Center. Or Zillah beat Davis. Seattle Christian beat Auburn and Mount Rainier. Good 1A teams beating not good 4A teams. Cascade Christian beat Federal Way: a mediocre 1A team over a bad 4A team. And then Sunnyside beat Zillah. A good non-league matchup between two good teams, one is 4A and the other 1A, but both good.

How about this year? Here’s the table for games so far:

v 4A v 3A v 2A v 1A v 2B v 1B
4A 69-62 27-22 4-4 0-0 0-0
3A 62-69 65-47 14-13 1-1 0-0
2A 22-27 47-65 42-45 9-3 1-2
1A 4-4 13-14 45-42 62-25 16-12
2B 0-0 1-1 3-9 25-62 33-42
1B 0-0 0-0 2-1 12-16 42-33

Best game I’ve seen so far was Sunnyside Christian (1B) at Lynden (2A). Sunnyside Christian won, not an upset. Not a black mark against Lynden for losing. Tim Martinez would put the ‘scarlet RPI’ on Lynden’s warmups for even scheduling that game. Maybe Hudson’s Bay has played several lower classification teams. But they look to be teams roughly well matched to Hudson’s Bay. You want to schedule some games that are likely wins where you can successfully exercise your skills against a live opponent. You want some games that will be challenges. And other games where you’re well matched. Regardless of classification.

Should a ranking method specifically include classification as a factor? There is a wide spread between good and bad teams within a classification, and a great deal of overlap between capabilities of teams in differing classifications. RPI is already a Rube Goldberg ranking. Hammering in a classification factor that doesn’t reflect reality would just make it worse.

Russians Go Gaga for Girls’ Basketball

The TeamBrunnhilde blog, which you are reading, is a moderated blog. Thus even tough the number of readers may exceed four, the profusion of comments generated by this multitude must be approved by the Blog Czar before they appear in the comments for the blog. So far only one (1) comment has made it through the gauntlet. I wonder what those other comments had to say.

Somebody, probably a Russian—at least his email domain name is Russian, wrote in Latin that he protested this. Exactly what ‘this’ is undefined. The kind of post where the poster is just checking to see if the post appears in the site. Included is a URL. Then I get a post, also from a Russian, letting me know that I am wrong and he can prove it just give him a call. That is basically all it said, the kind of reply that fits anywhere. You can write posting program to puke it out where it finds a niche. The Russian version of the Nigerian scam. A bunch of URLs follow. Most of those from Russian domains. My guess is they are all vectors for malware.

How could I resist this adoring comment from February? (also from a Russian):

Simply desire to say your article is as astonishing. The clearness in your post is just cool and i could assume you are an expert on this subject. Well with your permission let me to grab your feed to keep up to date with forthcoming post. Thanks a million and please keep up the rewarding work.

Sometimes I can tell at least one reason for the comment is to trick search ranking rules. Somebody wants to get a particular URL referred from as many sites as possible. Ergo, stick the URL into the comments on blogs, all over. Russians, yeah. Others too, maybe.

Funny also is the amount of hits on the site from Russia. Now I don’t get a lot of hits anyway. And a lot of hit are just bots. But even the bots—a bunch are from Russia. And the 404 errors. Somebody is trying to hit likely unpublished directories, hoping for a hit on a hidden login or something. And a bunch of these are from Russia.

How cum? They might not be Vladimir Putin, but they’re at least a bunch of crooks.

Rating Debut

Mondays are a light day at TeamBrunnhilde. No slate of Sunday games to enter into the database. Time to catch up on missing games from the previous weeks.

A new feature of the schedule pages are TeamBrunnhilde point rating differences of the teams playing. Early in the season, before this season’s points ratings are calculated and published, that difference is based on last season. Consider Dayton girls, 4th in state last year and ranked #1 for part of the season by AP. They scored two (2) points in their opener against River View. That’s falling off a cliff. Mercer Island girls, last year’s 3A champion, who regularly beat hapless teams last season finds itself on the other side of the ‘hap’ ratio, losing their last three by 41 point average. Bothell girls ranked #1 by WIAA RPI are also in dumps, although that slide may have begun before last season ended. And on the boys side, one-and-done Nathan Hale finds itself descending to customary status for that program now that Brandon Roy has set up shop at Garfield. So last year’s ratings may be misleading.

So when to publish ratings based on this season? The TeamBrunnhilde points rating needs teams to be ‘connected’ into a continuous set. Say we measure the distance between teams by games. The distance of teamA to teamA is 0. You probably would have guessed that. From teamA to teamB, which have played each other, are 1 game apart. TeamB’s opponents are 2 games away from teamA (provided that teamA has also not played them). Imagine then a tinker-toy structure. Teams are the hubs and games are the colored rods connecting them. Early in the season, there are few games per hub. At first, there are pairs of hubs connected by single rod. As more games are played more and more connections are made, and finally every hub is connected, via other hubs and games, to every other hub. But the whole structure is pretty rickety. Measuring the width of the whole set, it might be 15 games or so between some teams. And the number of path ways for that distance will be few. By the end of the season the distance between teams will be well under 10 with multiple paths of comparison.

Last Tuesday (December 5), complete connection of teams was achieved; teams that hadn’t played sufficient games to connect to anybody not included. Wednesday, Thursday, Friday, and Saturday added more games. So by now, after two weeks of data, one can pick up the tinker-toy construction of games and it doesn’t sag too badly. It will get better. It is likely better than applying last year’s ratings to this year’s teams.

Here are the initial rankings, broken out by gender and class. This week’s complete rankings are available for boys and girls.

Girls 4A

1 Kentridge 55.5
2 Moses Lake 52.5
3 Eastlake 49.2
4 Lake Stevens 46.2
5 Bellarmine Prep 46.0
6 Woodinville 43.3
7 University 43.1
8 Chiawana 41.4
9 Lewis and Clark 37.7
10 Skyline 35.5

Girls 3A

1 Kamiakin 53.0
2 Garfield 51.5
3 Lincoln 46.6
4 Bethel 46.0
5 West Seattle 45.1
6 Seattle Prep 39.0
7 Rainier Beach 37.4
8 Bellevue 35.5
9 Prairie 34.4
10 Snohomish 34.3

Girls 2A

1 W.F. West 49.0
2 East Valley (Spokane) 41.8
3 Archbishop Murphy 40.5
4 Wapato 40.2
5 Clarkston 31.0
6 Black Hills 30.0
7 Burlington-Edison 29.5
8 White River 28.2
9 Woodland 25.0
10 Washougal 24.7

Girls 1A

1 Lynden Christian 52.5
2 Zillah 52.4
3 Cashmere 44.8
4 La Salle 35.6
5 Medical Lake 35.0
6 Cle Elum-Roslyn 34.8
7 La Center 31.1
8 Meridian 26.4
9 Connell 24.8
10 Seattle Christian 23.9

Girls 2B

1 Davenport 37.3
2 Northwest Christian (Colbert) 29.7
3 Mabton 28.9
4 St. George’s 25.7
5 Ilwaco 25.3
6 Wahkiakum 20.1
7 Lind-Ritzville/Sprague 19.7
8 Colfax 18.6
9 Napavine 16.9
10 White Swan 14.4

Girls 1B

1 Sunnyside Christian 56.1
2 Colton 23.7
3 Pomeroy 17.8
4 Inchelium 14.6
5 Garfield-Palouse 14.4
6 Oakesdale 9.2
7 Touchet 8.9
8 Selkirk 7.4
9 Wellpinit 5.5
10 Neah Bay 2.0

Boys 4A

1 Federal Way 57.8
2 Gonzaga Prep 45.1
3 Enumclaw 44.9
4 Richland 41.0
5 Bothell 39.2
6 Kentwood 38.2
7 Auburn 37.2
8 Curtis 36.4
9 Kamiak 36.1
10 Union 34.6

Boys 3A

1 Seattle Prep 61.2
2 Garfield 60.2
3 Eastside Catholic 52.7
4 Wilson 52.6
5 Rainier Beach 51.9
6 O’Dea 49.5
7 Lincoln 48.1
8 Franklin 46.9
9 Cleveland 42.0
10 Lakeside (Seattle) 38.5

Boys 2A

1 Mountlake Terrace 50.1
2 Anacortes 36.4
3 Columbia River 35.6
4 Foss 33.6
5 Selah 31.8
6 Sedro-Woolley 31.4
7 Lynden 31.1
8 Mark Morris 28.4
9 Lakewood 27.3
10 Renton 27.0

Boys 1A

1 Lynden Christian 35.1
2 Bellevue Christian 21.0
3 Montesano 19.6
4 King’s 17.4
5 Northwest School 13.4
6 Nooksack Valley 11.2
7 Freeman 10.1
8 Okanogan 9.6
9 South Whidbey 8.9
10 Mount Baker 8.4

Boys 2B

1 Winlock 23.0
2 Toutle Lake 22.1
3 Life Christian 21.5
4 Adna 14.6
5 Napavine 13.9
6 St. George’s 13.7
7 Brewster 13.1
8 Morton-White Pass 7.0
9 Toledo 3.7
10 Wahkiakum -3.3

Boys 1B

1 Sunnyside Christian 12.9
2 Muckleshoot 9.6
3 Yakama Tribal 8.6
4 Odessa 3.7
5 Cedar Park Christian (Mountlake Terrace) 0.7
6 Pomeroy -1.9
7 Naselle -5.5
8 Wellpinit -6.2
9 Tacoma Baptist -8.3
10 Mount Rainier Lutheran -8.4

On oddity is that the overall top girls team is Sunnyside Christian, a 1B school. This is likely an early-season artifact due to the paucity of games played. Sunnyside Christian had played only White Swan and Mabton. The latter two have only played, besides Sunnyside Christian, 1A teams from the SCAC. The SCAC has #6, #26, #31 overall (out of 357 teams with point ratings). This is likely due to a very small sample size of game connecting that set of teams with the rest of the set. So if one is inferring that Sunnyside Christian would beat Kentridge, the initial #2 team, don’t bet the mortgage money on that. However, within the SCAC the estimates are likely more reliable. Note also that Central Valley girls have played only one game so is not included in the TB point rating. Several other teams also fall into that category.

Getting Ready for the Season

In three weeks the high school basketball season will be well underway.

November is the month I gather schedules. For 2016-2017, there were 7668 scheduled games for 764 teams. With a new high school, but other schools taking a hiatus, likely about the same number this year.

I found that Richland boys are playing in an event: “Free Tax USA Shoot out” in Orem Utah. Is that where all the tax-dodging American Oligarchs are given firearms, put in a pen, and have a gun fight? No. It’s just a tournament sponsored by a company that offers tax filing for free. Either possibility has entertainment value.

Athletic co-operatives continue to change. After losing longtime partnerships Tekoa-Oakesdale and LaCrosse-Washtucna in recent years, Odessa-Harrington has split up. Odessa retains the league team for NE1B while Harrington appears to have the limited schedule of an independent, or even quasi-JV, team. LaCrosse has resurfaced in an enlarged co-op with St. John-Endicott. Kittitas has combined with Thorp. What about Easton? Are they this season’s Washtucna?

MaxPreps does it again.

If you really want something screwed up, you can depend on MaxPreps. MaxPreps index pages for girls basketball and boys basketball are missing a lot of teams. They had them last season. What happened? How did it get to publication? Is anybody checking?

There are schools I didn’t know had basketball teams, even schools I didn’t know existed. ‘Harbor’ high school in Aberdeen for instance. I do recall years ago seeing ‘Harbor’ in the teams with results at a state swim meet. If appeared to be several of the schools around Grays Harbor combining to offer a swim program. On MaxPreps, ‘Harbor’ has five boys games over three seasons (2010-2013). None of the games have scores. All are actually scheduled games for FRIDAY HARBOR, which is not near Aberdeen by many measures. First of all MaxPreps database is wrong for having these games given to a fictitious high school. But now, they have resurrected that school, without even a spurious result for four years, and proclaimed it to be an active basketball school in Washington state, while dropping many other schools, like Rainier Beach. Central Valley and University apparently have combined into one boys’ team and Bothell, Inglemoor, and Woodinville have combined for a girls’ team. Cascade (Leavenworth) is there but Cascade (Everett) is not. You can look at the page and can likely pick out dozens of schools you know are playing basketball this season, who MaxPreps has dropped. While MaxPreps has elevated ‘Harbor’ to actuality as a school in Aberdeen, Aberdeen high school itself has disappeared from the list.

These are the guys running the database upon which WIAA RPI calculations depend.

Upsets

I’ll admit to being a Liberty HS (Issaquah) girls basketball fan. Long ago, although I can remember, when Liberty made the state tournament in 1999, they played their winner-to-state versus Hazen, the next high school over.

Now Hazen finished second in Seamount that season, which doesn’t sound so bad. In the district tournament they played Holy Names, considered one of the hot teams in the state that season. Holy Names was Metro #1 and Hazen was Semount #2. Hazen won. “What a huge upset!” was my thought. But other than my opinion that Seamount girls basketball was mostly crappy and Metro mostly not (at least in the top half of the league) I had meager data to back it up. I needed to formulate criterion to determine what constituted an upset.

It took nearly fifteen seasons and acquisition of lots of data that I didn’t have in 1999, but I came up the the TeamBrunnhilde points rating, calculated historically now from 1988-1989 on. Using the TB Points rating calculated after the season, one can determine what games were won by teams with lower ratings. These can be classified as ‘upsets’.

Not every such ‘upset’ is much of a surprise, however. If Richard Nixon high school has a 23.80 points rating, and Lance Armstrong high school is rated 23.77, is an Armstrong win over Nixon an upset? Technically, yes, but practically, no. One would say Nixon and Armstrong were pretty well matched. But the points rating, shown to have a track record better than most for picking winners in state games correctly, is a reasonable start to assess upset magnitude.

Nowadays, when I attend a game, I consult my TB Points rating to estimate the final margin (point spread for when betting on girls’ high school games is legalized). What does that mean if Richard Nixon HS is rated +7.83 over Harvey Weinstein HS? How likely is Nixon to kick Weinstein’s ass? Or just win?

So I looked at my database for girls’ games 1988-2017 (as of October 10, 2017). In all, I’ve recorded 103,779 girls’ game results where I have ratings for both teams. The lower rated team won 14.07% of those games. Upsets if you will. However, most of those upsets occurred at fairly low ratings differences. For very closely rated games (ratings differences less that 1 point), the lower rated team won 46.97% of the time. Out of 4245 games, that beats flipping a coin. Not until one gets to about a 12 point difference does the upset percentage drop to 10%. That is for games with TB Points ratings differences of between 11 and 12 points, the lower rated team wins 10.79% of the time.

The biggest upset was Thorp over Waterville on January 26, 2001. Waterville was a 52.72 point favorite, but lost 32-26.   But that Thorp Waterville upset is suspect since Waterville may have counted it as JV and not brought their best team.  Waterville had too many games: two late season non-league games (including that one) do not appear in their State Tournament program game listing.  (Disclaimer added after initial post).

Upsets with TB Points differences over 30 are quite rare. Of 15,487 such games only eight times was an upset recorded. Better odds than PowerBall, though.

For playoff games, the percentage of upsets is higher, 19.80% of 14,968 games. But that is mostly due to the worst teams getting eliminated quickly, leaving more closely matched games as the tournaments progress. For instance, for all games 79.43% of the games have TB Points rating differences of 26 points or less. For Playoff games 92.38% games are under 26 points. To get under 80% for playoff games, the TB Point margin drops to 17. The actual rate of upsets for various margins is roughly equivalent. The biggest playoff upset was Lopez over Neah Bay on February 18, 2011. Lopez, a 24.88 point underdog won 45-33.

Oh yeah. Hazen was a 22.8 point underdog when it beat Holy Names in 1999: the fifth biggest girls’ playoff upset in my database.

Following is the summary for upsets in games based on the differences of end-of-season TB points rating (so it may not agree with a similar analysis taken of differences during the season). The first column indicates the difference between the two teams point rating (stated in absolute terms). One can see in the last column the percentage of games where the lower rated won steadily decreases, until the ‘rare random event’ portion of the distribution takes over. Girls’ data 1988-2017.

range

All Games

upsets

games

[0,100)

14603

103779

14.07%

[0,1)

1994

4245

46.97%

[1,2)

1899

4195

45.27%

[2,3)

1680

4271

39.34%

[3,4)

1456

4115

35.38%

[4,5)

1270

4041

31.43%

[5,6)

1096

4019

27.27%

[6,7)

963

3937

24.46%

[7,8)

810

3894

20.80%

[8,9)

657

3642

18.04%

[9,10)

560

3745

14.95%

[10,11)

471

3769

12.50%

[11,12)

368

3410

10.79%

[12,13)

294

3323

8.85%

[13,14)

226

3162

7.15%

[14,15)

206

2979

6.92%

[15,16)

155

2874

5.39%

[16,17)

117

2806

4.17%

[17,18)

96

2726

3.52%

[18,19)

88

2513

3.50%

[19,20)

55

2466

2.23%

[20,21)

40

2409

1.66%

[21,22)

23

2133

1.08%

[22,23)

22

2083

1.06%

[23,24)

14

2036

0.69%

[24,25)

15

1811

0.83%

[25,26)

6

1827

0.33%

[26,27)

8

1609

0.50%

[27,28)

4

1471

0.27%

[28,29)

2

1438

0.14%

[29,30)

0

1343

0.00%

[30,31)

1

1282

0.08%

[31,32)

0

1186

0.00%

[32,33)

2

1087

0.18%

[33,34)

0

1068

0.00%

[34,35)

1

960

0.10%

[35,36)

0

903

0.00%

[36,37)

0

844

0.00%

[37,38)

1

678

0.15%

[38,39)

2

696

0.29%

[39,40)

0

591

0.00%

[40,41)

0

618

0.00%

[41,42)

0

495

0.00%

[42,43)

0

550

0.00%

[43,44)

0

482

0.00%

[44,45)

0

430

0.00%

[45,46)

0

380

0.00%

[46,47)

0

341

0.00%

[47,48)

0

297

0.00%

[48,49)

0

287

0.00%

[49,50)

0

252

0.00%

[50,51)

0

244

0.00%

[51,52)

0

195

0.00%

[52,53)

1

179

0.56%

[53,54)

0

160

0.00%

[54,55)

0

140

0.00%

[55,56)

0

106

0.00%

[56,57)

0

116

0.00%

[57,58)

0

102

0.00%

[58,59)

0

91

0.00%

[59,60)

0

99

0.00%

[60,61)

0

111

0.00%

[61,62)

0

67

0.00%

[62,63)

0

72

0.00%

[63,64)

0

47

0.00%

[64,65)

0

55

0.00%

[65,66)

0

42

0.00%

[66,67)

0

28

0.00%

[67,68)

0

27

0.00%

[68,69)

0

26

0.00%

[69,70)

0

23

0.00%

[70,71)

0

20

0.00%

[71,72)

0

17

0.00%

[72,73)

0

17

0.00%

[73,74)

0

4

0.00%

[74,75)

0

14

0.00%

[75,76)

0

11

0.00%

[76,77)

0

10

0.00%

[77,78)

0

7

0.00%

[78,79)

0

4

0.00%

[79,80)

0

4

0.00%

[80,81)

0

2

0.00%

[81,82)

0

4

0.00%

[82,83)

0

2

0.00%

[83,84)

0

3

0.00%

[84,85)

0

2

0.00%

[85,86)

0

0

0.00%

[86,87)

0

0

0.00%

[87,88)

0

3

0.00%

[88,89)

0

0

0.00%

[89,90)

0

0

0.00%

[90,91)

0

3

0.00%

[91,92)

0

1

0.00%

[92,93)

0

0

0.00%

[93,94)

0

0

0.00%

[94,95)

0

1

0.00%

[95,96)

0

0

0.00%

[96,97)

0

1

0.00%

[97,98)

0

0

0.00%

[98,99)

0

0

0.00%

[99,100)

0

0

0.00%

More data

I’ve added more boys data for 2002-2005 seasons (three seasons).  Although not a complete set of games for every team, at least there is some data for all teams, and some of the teams are complete (I think).  Data was gathered from scanned newspapers, WIAA programs, and searchable newspaper archives, which extend back into the end of the period.  This allows calculation and publication of rating numbers for boys for 2002-2005 which I had not previously done.  Some additional and corrected games for girls have been published.  For Greater Spokane League, Bill Pierce provided scores and coaches.  Hopefully I checked these all.  I thought I did.  Maybe I better check again.

More coaches names have been added.  Except for the “golden data” from Bill Pierce for the Greater Spokane League (that data is right, provided I entered it correctly), coaches names are found in various programs, searched articles (particularly the ‘fired/hired’ articles), or mentions of the coach in a scanned newspaper article.  When searching for a coach (search: <coaches name> <school>, but try to be a little more selective than ‘John Smith Jefferson’), one can really start crawling through the web.  A lot of the coaches found are for Puget Sound schools.  There are several searchable newspapers, Seattle Times, Tacoma News Tribune, Kitsap Sun, The Olympian, Everett Herald, Bellingham Herald, lots of minor newspapers.  Fortunately I searched the Seattle Times a while ago before they dumbed down their search engine and lopped off 14 years of searchable content, effectively rendering the search task impossible for their archive.  Coaches names are an afterthought for this project.  A guideline I’ve mostly used is not to publish a coach’s name unless I have evidence that the coach was actually the coach for a particular year.  So if I know the coach for year X and for year X+2, I leave year X+1 blank (mostly).  The site user can fill in.  When an article says thus-and-such has coached here for N years, I usually count back N years and show thus-and-such as the coach.  That is not always the case, as several coaches have left and returned to the same school, newspaper articles aggregating the years without mention of the hiatus.  Sometimes one can figure it out.  Looking back for coaches you come across some pretty interesting cases for coaching replacments: dressing too many girls for a playoff game, not supervising during a water balloon fight on a holiday road trip, smoking marijuana with players, getting arrested, sexting.

As for scores, there are disagreements between sources, even as to the outcome (win/loss).  I have my rules for deciding what to go with.  Every data source has ‘credible errors’, meaning that I don’t just trust the data based on source.  Sometimes the source contradicts itself.

Thanks to Bill Pierce for GSL data, and the WIAA staff for allowing access to their program library.

Finalizing the season

I’ve posted final checked data for the season.  It is a month after the championship games, but it takes a while to give the data one more look.  By my count there were 8,834 games played.  There were also a number of canceled games, of which I have retained the canceled league games in the database.  A dozen years hence one may wonder whatever happened to the girls game between Lynden and Ferndale.  Not played; canceled; don’t bother looking for it.

Although the season lasts three months (or a bit more), gathering the data takes five months.  November is for gathering the schedule data–always subject to change and more so this season.  March is spent cleaning up and verifying.  I attempt to get at least two sources for every score I post.  Even a difference of a single point means that only one of those sources can be used and that I need to find some other corroborating source, and then make a call as to which is the score I’ll post.  A little over 6% of the games have a difference in reported score between various sources.  Sometimes there are three different scores reported for a game.  Single sources for a score are used for just over 6% of the games, although that includes the 180 games played March 1-4 (about 2%) for which I just trusted WIAA to have the right information.  Both WPAN and MaxPreps have some claims for official status, but often differed on what the game score was.  Both sources agreed on outcomes, though.  Just a few outlets, notably the Seattle Times, had scores reversed.  The quality Seattle Times game data has declined substantially since that function was outsourced to SportNGIN.  Kitsap Sun has apparently dropped their attempt at a database entirely.

The differences in reported scores reveals a problem with an official rating that uses the actual score.  With RPI, it was only necessary, once one got the schedule right, to know the winner of a game.  By the end of the season, the handful of games where the score was reversed at some point during the season on MaxPreps were all corrected.  But if ratings depended on whether the score was 67-49 or 64-48, there are hundreds of such games where official reporting disagreed as to score.  In some cases, the best evidence indicated that neither official source was correct.

Of the 8,824 games with scores, I have valid quarter scores for nearly 80% of these.  I’ve updated the quarter score report page to include data through 2016-2017.  Likewise the home/away report page has been updated.

Sources for 2016-2017 results are:

Aberdeen Daily World, Longview Daily News, MaxPreps, Oakville HS, Bellingham Herald, Everett Herald, Peninsula Daily News, District 9, Spokane Spokesman-Review, Emerald City League, Seattle Times, Facebook, Twitter, idahosports.com, Manson HS, Mid-Columbia Conference, Tri-Cities Herald, Walla Walla Union-Bulletin, Martin Luther College, District 4, District 3, Goldendale Sentinel, Islands Sounder, Kitsap Daily News, Kitsap Sun, Tacoma News Tribune, Oregon Prep Sports Net, Oregon School Activities Association, Associated Press, Vancouver Columbian, Yakima Herald, Eugene Register-Guard, South Central Athletic Conference, WPAN: Washington Prep Athletics Network, ranchomirageholidayinvitational.com, Ellensburg Daily Record, Northeast A League, WIAA, basketballtravelers.com, Columbia Basin Herald, Douglas County Empire-Press, Gig Harbor HS, myself, Portland Oregonian, Port Townsend Leader, Renton Reporter, surfnslam.com, theholidayclassic.org, Marin Catholic HS, Walla Walla University, Cusick HS, Bear Creek HS, Cle-Elum/Roslyn HS, Davenport HS, Kettle Falls HS, Brewster HS, Methow Valley News, sportsmx.com, Nike Tournament of Champions, Bill Pierce, tourneymachine.com, socalholidayprepclassic.com, Santa Barbara TOC (sbtoc.com), West Coast Jamboree, Whidbey News Times

 

Ranking the rankings: WIAA RPI finishes last

The updated final (I think) ratings for various rankings is done.  Of the eight rankings, the one used to seed the tournament was the very worst at actually picking winners of the games.  It isn’t as easy picking winners are one might think.  You should try it next year.  Rank 16 teams for all twelve tournaments.  But it looks like it is fairly easy to pick winners better than the WIAA RPI did.  Even good old ‘won-loss’ managed to have a higher portion of correct game picks than WIAA RPI.

I don’t know what WIAA has up its sleeve for tournament format.  Don’t know who will be the official ball of the 2018 tournaments.  But I sure hope that the ranking method for seeding gets some rework.  Four of the girls championship games were a repeat of a regional game.  Three of the results switched.  How’s that for making sure the top teams don’t meet until the championship?  At least the regional games weren’t loser out, although the three teams that won the regional and lost the championship are probably wanting a rubber game.

Exciting regional basketball

This year I mapped out six regional games featuring big school girls teams (my general interest).  Two in Bothell on Friday, and on Saturday two at Puyallup HS, and two at Rogers (Puyallup) HS.  Four games decided in the final seconds.  One twice (it went to overtime).  I can’t recall a cluster of games like that at a girls state tournament.  Three of the games were winner to Thursday/loser to Wednesday, so lacked the urgency of a loser out game but competition was still fierce.  The final of the four was the loser out between Wilson and Bethel.  The final quarter with end-to-end rushes, spectacular plays, lead changing back and forth as the team desperately tried to assure that trip to the Tacoma Dome.  Everything one hopes for in a state tournament game.  Thank you, Mercer Island, Bishop Blanchet, Gig Harbor, Kamiakin, Camas, Kentlake, Wilson, and Bethel.  A well spent weekend watching basketball.

Last year there were four games total out of 48 girls regional games that were decided by five points or less (or in overtime).  This year there were 16 in all.  A third of the games.  The average margin dropped from 16.8 to 12.9, nearly four points closer.

Ratings Comparison Update

With 96 of the 276 tournament games in the book, just over a third of the state tournament is complete.  Of those 96 games, using the WIAA RPI, 60 of 96 higher seeds won, 62.5%.  That is about the same as if one were picking winners based on win-loss percentage: 59 of 94 (two games had teams teams with the same w-l records), 62.8%.  With a better RPI that uses playoff games and available out-of-state win-loss records, RPI would have a 67.7% correct mark (65 of 96).  My TeamBrunnhilde points rating picked 73 of 96 games correctly for 76%.  There are still 180 games to go, but the WIAA RPI isn’t looking like God’s gift for seeding teams.

That is not to say the WIAA should just seed based on win-loss and forget the RPI rigmarole.  If seeding was based on won-loss there would be a lot of games scheduled against the Little Sisters of the Poor.  When one announces the basis for ranking, behavior is altered to optimize the ranking.  One can see that in how teams scheduled this year with RPI in mind compared to last year.  It is to say that one should be able to create an open ranking system (not a secret one that nobody knows what goes into it), that can more accurately assess better teams than the WIAA RPI did for 2016-2017.

MaxPreps Update

As soon as the WIAA not need accurate data from MaxPreps, MaxPreps is back to their old tricks.  Playoff game data is badly incomplete.  Look for the regional games and you’ll see many are missing from the schedule entirely.  So for all the effort expended in getting the WIAA RPI data accurate, no lasting reforms are forthcoming at MaxPreps.  It will be the same work next year just to get to the same spot.

2017 Ratings Comparison

Back by popular demand. Well, I wanted to do it anyway, here is the ranking comparison prior to the state tournaments. This year I’ve collected eight rankings:

WIAA RPI

TeamBrunnhilde RPI

Win-loss

Captured Win-Loss

TeamBrunnhilde Points rating

MaxPreps Rating

Evans Rating

Score Czar rating

The two RPI ratings differ in that TeamBrunnhilde RPI uses available out-of-state records and also utilizes playoff results. The latter three ratings use proprietary rating methods. Win-loss and Captured Win-loss are straightforward calculations one could compute by hand. TeamBrunnhilde Points uses a standard statistical technique.

There is consensus for Boys 3A (Nathan Hale), Girls 1A (Cashmere), and Boys 2B (Kittitas). The other classifications have varying degrees of disagreement between the ratings. While some ratings will have better results than others, all ratings will have notable successes and failures. Like last year, I expect that some ‘sure thing’ game picks (where all rankings agree) will be wrong. The failure rate last year was about 18% for these.

I note that this will be the final year for Score Czar doing high school sports. We’ll miss you. Good luck with your further rating endeavors.