Here’s the announcement from the WIAA: http://wiaa.com/news.aspx?ID=493&Mon=9&Yr=2016
There will also be changes to state tournament formats, allowing a good team to lose the first game and still compete for the championship. That was the route Colton girls took in 2011 winning their third of (so far) eight straight championships.
RPI is just a calculation dreamed up by somebody. There is nothing inherently ‘correct’ about it. In fact there isn’t even a standard definition of what RPI is. I use a formula similar to what Wikipedia has. The WIAA formula will be the same as the RPI I have calculated for teams 1988-2016. The WIAA anticipates that data will be available for out of state teams, and would use a win percentage of .500 when not available. There will be no home/away adjustment with the WIAA formula.
It looks like MaxPreps will be the authoritative database for the WIAA RPI calculation. MaxPreps has continuing accuracy and completeness problems. Just today (October 3) I looked at their 2016-2017 schedule for Cascade (Leavenworth) boys. Lots of games for Cascade (Everett) are assigned to Cascade (Leavenworth). I let the WIAA know. Maybe they’ll get it fixed. I’m sure there are and will be other mistakes. It should not be the responsibility of coaches to fix the data on a site owned by CBS, a corporation with a 24 billion dollar market cap.
When you start rewarding teams with a metric, teams will seek to maximize their scoring for that metric. Since RPI considers only wins and losses, that becomes the behavior the metric encourages. If the linear model were used, since that is based on margin, one might see teams ahead by 30 after three quarters seek to boost the final margin to 40 instead of playing deep reserves or giving JV players varsity experience. So for the goals of high school sports: personal development, team building, good sportsmanship, using RPI as a rewarding metric works better than using the linear model. Using any metric does change the way teams will use non-league games since the outcome of these can actually affect where the team is positioned at state.
The use of RPI is not the end of crappy teams in the state tournament. The WIAA allocation method still assumes a uniform distribution of good teams across the districts. It is a geographical quota system whereby weak districts still get allocations based on number of teams, not numbers of good teams. So a number of state berths will be filled with mediocre or mediocre-wannabe teams. But that’s a subject for another post.
HOW DOES THIS WORK IN OREGON?
Oregon uses RPI and another rating not only to seed teams in the state tournament but to also qualify some of the teams into the state tournament. The Oregon RPI formula is simpler, using only a team’s winning percentage (0.35 weight) and the winning percentages of the team’s opponents (0.65 weight). The opponents’ opponents’ win percentages are not used in the calculation. This makes for easier calculation and reduces the need for extensive data on out-of-state opponents. Oregon does adjust for home/away although the premium they put on road wins is too high by historical data at least for Washington. A game database is available through the OSAA site. I don’t know if that has been outsourced. There is a capability to report out-of-state records. I don’t know how up-to-date those are kept. The Oregon selection criteria can be found at http://www.osaa.org/governance/handbooks/osaa#_Toc456100456
Use of ratings to select which teams qualify for state tournament puts importance on some of the arbitrary elements in the RPI rating. This isn’t just for fun. Somebody is going to stay home because the calculation had their RPI .0003 behind another team. Perhaps if the home/away adjustment used correct weightings, the RPI rankings would flip and they’d be in. Or if the arbitrary weightings for team winning percentage and opponent’s winning percentage were different (and who is to say what the ‘right’ weights are). Or if the winning percentage for that Washington team they played was kept up to date and not what the coach entered back in December when they played. I’ll have more to say on how small differences in formula or in games included in the analysis can change the RPI numbers.
I don’t necessarily like RPI as a rating. It is good that the WIAA will be using some method to seed tournament fields and not just relying on placements from district tournaments where a single upset can flip a high seed down to the bottom. It is also good that the WIAA did not go whole hog and embrace a calculation for state tournament qualification. Although I’m sure the districts would have a lot to say about killing their cash cow district tournaments with such a move.
PROBLEMS with RPI
The topological problem.
Washington state is geographically divided by the Cascade mountains. At least in girls’ basketball the Cascades are a huge divide as well. The state is divided into two large sets of teams: WOC and EOC (west and east of cascades). Most of the games, prior to state tournaments, are against teams from the same set. For RPI, by the nature of the calculation, games between the two sets are given no more significance than games within a set. Since there are so many more intra-set games than inter-set games, RPI can’t tell the significance of differences between the sets. The inter-set games are just overwhelmed. A linear model would be better able to differentiate the sets based on the inter-set games.
Sensitivity to arbitrary elements and bad data.
Since the formula for RPI is somewhat arbitrary and the calculation depends on having accurate information, small changes in RPI ratings don’t really mean much. Consider the difference between #8 and #9 in RPI ratings (this would be the split between playing a ‘both to Hardwood Classic’ game and playing a loser out). Using RPI ratings prior to last year’s state tournament
| 1B Girls | 0.5584 | 0.5523 |
| 2B Girls | 0.5914 | 0.5855 |
| 1A Girls | 0.5881 | 0.5880 |
| 2A Girls | 0.5832 | 0.5707 |
| 3A Girls | 0.6250 | 0.6159 |
| 4A Girls | 0.6000 | 0.5995 |
| 1B Boys | 0.5523 | 0.5520 |
| 2B Boys | 0.5649 | 0.5646 |
| 1A Boys | 0.5659 | 0.5653 |
| 2A Boys | 0.5716 | 0.5694 |
| 3A Boys | 0.6019 | 0.5974 |
| 4A Boys | 0.6010 | 0.6000 |
For half of the classifications, 8 and 9 are separated by 0.001 or less. Varying the exact parameters of the RPI calculation could well produce changes larger than this. So would changing the games used for the calculation, whether by administrative decision (no games against non-varisity opponents), or by neglect (MaxPreps lousy data).
HOW WOULD this have played out LAST YEAR?
It’s good news that regional games won’t pit two top teams against each other in a loser out while in another game two mediocre teams play so that one will go on to the quarter finals. Using the RPI I calculated (very similar to the WIAA computation), here is how last year’s tournaments could have been re-seeded. The possible match ups just use the RPI rating for seeding and do not adjust to avoid games between teams from the same district or league. The number in parentheses is the RPI rank of all teams within the classification. Missing numbers did not qualify for the state tournament. A red cell indicates an elimination game between two teams that would have been protected seeds under the new rules. A green cell indicates an advancement game to the quarter finals between two teams that would have at best played a first round game on Wednesday under the new rules.
1B Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Lyle/Wishram (22) v Colton (1) | Evergreen Lutheran (8) v Colton (1) |
| Touchet (3) v Sunnyside Christian (2) | Mt. Vernon Christian (7) v Sunnyside Christian (2) |
| Mary M. Knight (11) v Tulalip Heritage (4) | Almira/Coulee-Hartline (6) v Touchet (3) |
| Mt Rainier Lutheran (15) v Republic (5) | Republic (5) v Tulalip Heritage (4) |
| Lower seeds, winner to Wednesday loser out | |
| Entiat (19) v Almira/Coulee-Hartline (6) | Seattle Lutheran (28) v Mary M. Knight (11) |
| Seattle Lutheran (28) v Mt Vernon Christian (7) | Taholah (25) v Mt Rainier Lutheran (15) |
| Selkirk (17) v Evergreen Lutheran (8) | Neah Bay (23) v Selkirk (17) |
| Taholah (25) v Neah Bay (23) | Lyle/Wishram (22) v Entiat (19) |
2B Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Raymond (11) v Okanogan (1) | Waterville (8) v Okanogan (1) |
| Dayton (16) v Mabton (2) | Friday Harbor (7) v Mabton (2) |
| Ilwaco (4) v Toutle Lake (3) | Napavine (6) v Toutle Lake (3) |
| Lind-Ritzville/Sprague (10) v St. George’s (5) | St. George’s (5) v Ilwaco (4) |
| Lower seeds, winner to Wednesday loser out | |
| Davenport (15) v Napavine (6) | Tonasket (32) v Colfax (9) |
| Wahkiakum (14) v Friday Harbor (7) | Dayton (16) v Lind-Ritzville/Sprague (10) |
| Colfax (9) v Waterville (8) | Davenport (15) v Raymond (11) |
| Tonasket (32) v Adna (12) | Wahkiakum (14) v Adna (12) |
1A Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Charles Wright (14) v Lynden Christian (1) | Lakeside (NMF) (8) v Lynden Christian (1) |
| Montesano (12) v Granger (2) | King’s (7) v Granger (2) |
| Chewelah (24) v Zillah (3) | Kalama (6) v Zillah (3) |
| Nooksack Valley (15) v La Center (4) | Bellevue Christian (5) v La Center (4) |
| Lower seeds, winner to Wednesday loser out | |
| King’s (7) v Bellevue Christian (5) | Chewelah (24) v Cashmere (10) |
| Seattle Academy (11) v Kalama (6) | Coupeville (21) v Seattle Academy (11) |
| Columbia (Burbank) 13 v Lakeside (NMF) (8) | Nooksack Valley (15) v Montesano (12) |
| Coupeville (21) v Cashmere (10) | Charles Wright (14) v Columbia (Burbank) (13) |
2A Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| North Kitsap (17) v East Valley (Spokane) (1) | Shorecrest (9) v East Valley (Spokane) (1) |
| Liberty (Issaquah) (20) v Lynden (2) | Clarkston (8) v Lynden (2) |
| Clarkston (8) v Black Hills (3) | White River (7) v Black Hills (3) |
| River Ridge (25) v Ellensburg (4) | Anacortes (5) v Ellensburg (4) |
| Lower seeds, winner to Wednesday loser out | |
| Port Angeles (12) v Anacortes (5) | Toppenish (30) v Port Angeles (12) |
| Washougal (14) v White River (7) | River Ridge (25) v Washougal (14) |
| Franklin Pierce (22) v Shorecrest (9) | Franklin Pierce (22) v Mark Morris (15) |
| Toppenish (30) v Mark Morris (15) | Liberty (Issaquah) (20) v North Kitsap (17) |
3A Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Wilson (11) v Bellevue (1) | Auburn Riverside (8) v Bellevue (1) |
| West Seattle (10) v Lynnwood (2) | Lincoln (7) v Lynnwood (2) |
| Prairie (6) v Mercer Island (3) | Prairie (6) v Mercer Island (3) |
| Arlington (5) v Bishop Blanchet (4) | Arlington (5) v Bishop Blanchet (4) |
| Lower seeds, winner to Wednesday loser out | |
| Rainier Beach (15) v Lincoln (7) | Mount Spokane (32) v West Seattle (10) |
| Edmonds-Woodway (14) v Auburn Riverside (8) | Kamiakin (26) v Wilson (11) |
| Mount Spokane (32) v Auburn Mountainview (17) | Columbia River (21) v Edmonds-Woodway (14) |
| Kamiakin (26) v Columbia River (21) | Auburn Mountainview (17) v Rainier Beach (15) |
This was the worst seeded bracket overall. Two high seeded loser outs and two low seeded winners advance. That doesn’t include the Lynnwood/West Seattle game. Most other rankings had West Seattle ranked well above the RPI rank. Note that the two high seeded loser out games would still have been played, but both teams would be going on, the game just to determine who would get the bye to Thursday.
4A Girls:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Kamiak (33) v Kentlake (1) | Chiawana (9) v Kentlake (1) |
| Bellarmine Prep (10) v Skyview (2) | Bothell (7) v Skyview (2) |
| Rogers (Puyallup) (34) v Central Valley (3) | Moses Lake (6) v Central Valley (3) |
| Eastlake (14) v Beamer (4) | Snohomish (5) v Beamer (4) |
| Lower seeds, winner to Wednesday loser out | |
| Gig Harbor (15) v Snohomish (5) | Rogers (Puyallup) (34) v Bellarmine Prep (10) |
| Chiawana (9) v Moses Lake (6) | Kamiak (33) v Lewis and Clark (11) |
| Curtis (12) v Bothell (7) | Sunnyside (24) v Curtis (12) |
| Sunnyside (24) v Lewis and Clark (11) | Gig Harbor (15) v Eastlake (14) |
1B Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Columbia Adventist (22) v Seattle Lutheran (1) | Wellpinit (8) v Seattle Lutheran (1) |
| Liberty Christian (7) v Almira/Coulee Hartline (2) | Liberty Christian (7) v Almira/Coulee-Hartline (2) |
| Naselle (41) v Shorewood Christian (3) | Sunnyside Christian (6) v Shorewood Christian (3) |
| Wellpinit (8) v Garfield-Palouse (4) | Neah Bay (5) v Garfield-Palouse (4) |
| Lower seeds, winner to Wednesday loser out | |
| Riverside Christian (20) v Neah Bay (5) | Naselle (41) v Evergreen Lutheran (9) |
| Lummi Nation (14) v Sunnyside Christian (6) | Taholah (24) v Yakama Tribal (11) |
| Taholah (24) v Evergreen Lutheran (9) | Columbia Adventist (22) v Lummi Nation (14) |
| Entiat (19) v Yakama Tribal (11) | Riverside Christian (20) v Entiat (19) |
Another screwed up bracket. WIAA obsession with travel distances causes eastern teams to eliminate each other. Liberty Christian and ACH would still have played but both would have advanced.
2B Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Asotin (6) v Northwest Christian (Colbert) (1) | Napavine (10) v Northwest Christian (Colbert) (1) |
| Lake Roosevelt (17) v Morton-White Pass (2) | Mossyrock (8) v Morton-White Pass (2) |
| Adna (16) v Brewster (3) | Warden (7) v Brewster (3) |
| Napavine (10) v Life Christian (4) | Asotin (6) v Life Christian (4) |
| Lower seeds, winner to Wednesday loser out | |
| Lind-Ritzville/Sprague (15) v Warden (7) | DeSales (21) v Ocosta (11) |
| DeSales (21) v Mossyrock (8) | Friday Harbor (19) v Mabton (12) |
| Friday Harbor (19) v Ocosta (11) | Lake Roosevelt (17) v St. George’s (13) |
| St. George’s (13) v Mabton (12) | Adna (16) v Lind-Ritzville/Sprague (15) |
1A Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Zillah (2) v Lynden Christian (1) | Connell (11) v Lynden Christian (1) |
| Vashon Island (18) v King’s (3) | Seattle Academy (7) v Zillah (2) |
| Cascade Christian (Puyallup) (16) v King’s Way Christian (4) | La Salle (6) v King’s (3) |
| Naches Valley (20) v Freeman (5) | Freeman (5) v King’s Way Christian (4) |
| Lower seeds, winner to Wednesday loser out | |
| Medical Lake (14) v La Salle (6) | Seattle Christian (27) v Hoquiam (13) |
| Kalama (19) v Seattle Academy (7) | Naches Valley (20) v Medical Lake (14) |
| Cashmere (15) v Connell (11) | Kalama (19) v Cashmere (15) |
| Seattle Christian (27) v Hoquiam (13) | Vashon Island (18) v Cascade Christian (Puyallup) (16) |
This bracket featured the biggest seeding mistake of the year, pitting the two best teams (all ratings had these both in the top three if not 1 and 2) in a regional loser out game. Part of the problem was caused by D1/2/3 1A Tournament bracketing where the season winner of the Emerald City league is handed a state berth without playing a tournament game. The worst they can finish is second at district, which pushes better teams from other leagues down to third or lower which to the WIAA means they are a poorer team.
2A Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Liberty (Issaquah) (21) v Lynden (1) | Tumwater (9) v Lynden (1) |
| Renton (26) v Clarkston (2) | Shorecrest (8) v Clarkston (2) |
| Tumwater (9) v Selah (4) | Wapato (7) v Selah (4) |
| Squalicum (6) v River Ridge (5) | Squalicum (6) v River Ridge (5) |
| Lower seeds, winner to Wednesday loser out | |
| Lindbergh (23) v Wapato (7) | Renton (26) v Mark Morris (11) |
| Steilacoom (15) v Shorecrest (8) | Lindbergh (23) v North Kitsap (12) |
| Ellensburg (16) v Mark Morris (11) | Liberty (Issaquah) (21) v Steilacoom (15) |
| Woodland (17) v North Kitsap (12) | Woodland (17) v Ellensburg (16) |
3A Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Wilson (11) v Garfield (1) | Lincoln (8) v Garfield (1) |
| Mountlake Terrace (14) v Bellevue (2) | Edmonds-Woodway (7) v Bellevue (2) |
| Rainier Beach (5) v Auburn Mountainview (3) | Cleveland (6) v Auburn Mountainview (3) |
| Cleveland (6) v Peninsula (4) | Rainier Beach (5) v Peninsula (4) |
| Lower seeds, winner to Wednesday loser out | |
| O’Dea (12) v Edmonds-Woodway (7) | Kennedy Catholic (29) v Wilson (11) |
| Kamiakin (25) v Lincoln (8) | Stanwood (26) v O’Dea (12) |
| Stanwood (26) v Enumclaw (18) | Kamiakin (25) v Mountlake Terrace (14) |
| Kennedy Catholic (29) v Shadle Park (24) | Shadle Park (24) v Enumclaw (18) |
4A Boys:
| 2016 As Scheduled (all loser out) | 2016 if RPI seeded |
| All games winner to hardwood classic, loser out | Top seeds, winner to Thursday, loser to Wednesday, both to hardwood classic |
| Woodinville (15) v Federal Way (1) | Issaquah (9) v Federal Way (1) |
| Kamiak (12) v Curtis (2) | Union (7) v Curtis (2) |
| Union (7) v Bellarmine Prep (3) | Cascade (Everett) (6) v Bellarmine Prep (3) |
| Cascade (Everett) (6) v Kentwood (4) | Gonzaga Prep (5) v Kentwood (4) |
| Lower seeds, winner to Wednesday loser out | |
| Wenatchee (13) v Gonzaga Prep (5) | Kentridge (20) v Central Valley (10) |
| Timberline (17) v Issaquah (9) | Timberline (17) v Lewis and Clark (11) |
| Kentridge (20) v Central Valley (10) | Woodinville (15) v Kamiak (12) |
| Davis (14) v Lewis and Clark (11) | Davis (14) v Wenatchee (13) |
Different ratings might not complain about some of the red matchups. I like my linear model rating for accuracy. Here is a full listing of RPI and other ratings for last year’s state tournament teams. I’m kind of surprised that in some brackets, the WIAA only screwed over a pair of good teams. The WIAA seeding based solely on district tournament results was a recipe for bad seeding. That’s exactly what they got for years and years. The regional format with attempts to limit travel just made it worse. Good for the WIAA in finally doing something about it. I hope it works out.
More to come on this subject and rankings in general.