The Ohio Quizbowl Forum
http://ohioqbforum.net/

Statistical Rankings
http://ohioqbforum.net/viewtopic.php?f=7&t=541
Page 1 of 1

Author:  eckj [ Tue Oct 31, 2017 11:11 am ]
Post subject:  Statistical Rankings

I was bored and wanted to find a way to rank teams statistically. This is obviously limited to the TU/B format. The data I use makes this “useful” for comparison of Ohio teams only. I was inspired by Tyler Benedicts work last year and used some similar data. I ranked teams based a formula using the adjusted PPB (aPPB) and adjusted power percentage (aP%). I adjusted the PPB and P% by using the average of all IS sets.

The formula I used to calculated the ARP (average ranked points ) number I ranked the teams on was:

(aPPB*aP%) + aPPB = ARP

The reason I chose to add aPPB was to prevent teams from having a zero ARP if their power percentage was 0.

Feel feel to critique and add anything you think would make it better! I am open to lots of suggestions.

I ranked middle school teams and high school teams. I did not use the stats for Solon A and Beavercreek A from their hosted tournaments because I know their real A teams did not play. Let me know if I missed stats anywhere!

Tournaments used:
-High School: Solon Fall, Scottie (KY), Brookfield Gathering, Miami Valley Fall, Beavercreek. Mellon Bowl (PA), Waynesdale, Columbiana
-Middle School: Northmont, Miami Valley Fall

Middle School
Attachment:
ms rank nov 1.PNG
ms rank nov 1.PNG [ 16.37 KiB | Viewed 12914 times ]


High School
Attachment:
rankings nov 1.PNG
rankings nov 1.PNG [ 98.45 KiB | Viewed 12914 times ]

Author:  trbenedict [ Wed Nov 01, 2017 12:06 pm ]
Post subject:  Re: Statistical Rankings

This looks really cool, Josh (and much more nicely conceived than my brute-force attempts last year)...good to have someone with some quantitative acumen crunching numbers for a change!

In particular, the treatment of power percentage seems better than my ham-fisted way of accounting for it last year, which was to do one set of rankings entirely based on aPPB, and another entirely based on aP%, then add the relative positions of both together (so a team who was 1st on the aP% list and 3rd in aPPB would have a combined ranking of 1+3=4, with 2 then being the best possible number). I think your way here, Josh, is a nice solution to get sound rankings without all that double-listing, position-adding and other nonsense. Kudos to you for calculating all those adjustments!

Author:  ppadmanabhan [ Wed Nov 01, 2017 7:46 pm ]
Post subject:  Re: Statistical Rankings

Thanks for doing this! I always love looking at different ways of ranking teams. Looking forward to seeing how more Ohio teams do as the season progresses!

Author:  eckj [ Thu Nov 02, 2017 6:07 am ]
Post subject:  Re: Statistical Rankings

Thank you both. I will continue to do it just for just for fun. I already made a slight adjustment to the stats to remove outliers when calculating the adjustment for the sets.

Author:  eckj [ Sun Nov 05, 2017 6:10 pm ]
Post subject:  Re: Statistical Rankings

Updated High School after Tippecanoe and Gahanna-Lincoln this weekend.

https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=260751617

Author:  eckj [ Sun Nov 19, 2017 2:24 pm ]
Post subject:  Re: Statistical Rankings

Here is an updated: https://docs.google.com/spreadsheets/d/ ... =227305641

I made a few changes:
1. I pulled data for all of the sets from tournaments outside of Ohio to calculate set adjustments. I was getting some weird adjustments because I did not have enough data points.

2. I decided to calculate APR using the teams BEST adjusted PPB and P% instead of using their CUMULATIVE adjusted PPB and P%. This seems more fair because as teams improve their ranking would be held down by their stats from earlier in the season.

Author:  eckj [ Thu Dec 14, 2017 10:10 am ]
Post subject:  Re: Statistical Rankings

Update:

MS: https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=1060668002
HS: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1418089100

Author:  eckj [ Sun Jan 14, 2018 8:25 pm ]
Post subject:  Re: Statistical Rankings

Major update.

I changed how I calculate adjustments and I am now using the average of a teams top two performances in P% and PPB.

I also labeled each teams region. I used this map
Image

My calculations rely on teams playing IS or MS sets to compare to house writes, I do not have enough data to calculate adjustments so teams that have played CFMS 5/6, CFMS 7/8, SCOP HS, SCOP MS may have a score of 0.

MS: https://docs.google.com/spreadsheets/d/ ... 1766095557
-No major changes

HS: https://docs.google.com/spreadsheets/d/ ... 1684871096
-Bexley debuts at #9
-Dublin Scioto leapfrogs Solon into the #3 spot.

Breakdown By Region (Remember I am ranking every team I have TU/B stats for)
Northeast: Number of Teams (76), Avg. ARP (11.61)
Northwest: Number of Teams (61), Avg. ARP (9.80)
Central: Number of Teams (32), Avg. ARP (15.76)
Southeast: Number of Teams (15), Avg. ARP (11.61)
Southwest: Number of Teams (43), Avg. ARP (12.75)

Also, if any TD wants the raw data after/before adjustments for bracket placement I would be happy to share.

Author:  Get Lynned [ Sun Jan 14, 2018 9:07 pm ]
Post subject:  Re: Statistical Rankings

Damn, another impressive compilation. Thanks for taking the time to do this, Josh. I’m a believer that these quantitative efforts help legitimize Ohio’s Quiz Bowl circuit.

Author:  eckj [ Tue Jan 16, 2018 7:47 am ]
Post subject:  Re: Statistical Rankings

Thank you Tom. I like doing this and I have gotten some good feedback so far.

Author:  ThePocketProtector [ Tue Jan 16, 2018 9:50 am ]
Post subject:  Re: Statistical Rankings

Warning: My terrible sense of humor incoming.

Hey, look at that map! Columbus, Delaware, and Mount Vernon are in one region and Portmouth, Wheelersburg, and Ironton are in a different one. That doesn't seem right...

Okay, I got that out of my system. I know that the map isn't for the same purpose and there are many factors involved in our current regional map.

Josh, this is amazing. I'm greatly enjoying it. I'm glad that someone is doing this and there is actual data we can point to for things we have thought in the past.

Author:  eckj [ Tue Jan 16, 2018 1:10 pm ]
Post subject:  Re: Statistical Rankings

Thanks! I think the data could be useful once we have a whole season.

One thing I just looked at was numbers of the top 16 in each region. The southeast is lacking data right now so I only used the 8 teams that have data, but once your tournament happens I think I will be all set.

Attachment:
data.PNG
data.PNG [ 8.47 KiB | Viewed 11708 times ]

Author:  rmaupin [ Tue Jan 16, 2018 4:48 pm ]
Post subject:  Re: Statistical Rankings

Josh,

I really appreciate all of the work that you have put into the statewide ranking data, it really does do a lot to help promote our sport. It was also interesting to see the state divided up into regions for comparing data. However, why did you organize the state into 5 regions instead of 6 regions (like what we have for OAC). I know that your data is based on NAQT results (and not OAC), however the six region OAC system is the only way that we have organized the state (because as far as I know, we have never had NAQT regional tournaments, just a state tourney). In our case at Little Miami we are in the SW region for OAC, and it would not include teams such as Beavercreek and Northmont because they are in the WC region. Hopefully this doesn't come off as nitpicking, rather just curious about the methodology used. Once again, thanks for all the work you do in putting together state wide data.

Author:  eckj [ Tue Jan 16, 2018 8:42 pm ]
Post subject:  Re: Statistical Rankings

It does not come back as nitpicky at all. I just chose these based on what most maps on Google were haha. I will make one with sub regions that fit the usual OAC regions here in the next couple days.

Author:  eckj [ Wed Jan 17, 2018 10:31 am ]
Post subject:  Re: Statistical Rankings

Okay. I took a very quick guess at where most teams would go for OAC Regionals. If there any teams you think should switch regions let me know. I may have misplaced a few.
Attachment:
SC TO NC.PNG
SC TO NC.PNG [ 36.56 KiB | Viewed 11624 times ]

Attachment:
CN TO SW.PNG
CN TO SW.PNG [ 73.34 KiB | Viewed 11624 times ]


Top 16 teams in each regional summary
Attachment:
summm.PNG
summm.PNG [ 21.34 KiB | Viewed 11624 times ]

Author:  eckj [ Sat Jan 20, 2018 5:58 pm ]
Post subject:  Re: Statistical Rankings

1/20: Small update after Chillicothe today. The data from Chillicothe allowed me to get enough data for an adjustment for SCOP and tweak all adjustments a little.

Some shuffling occurred in the top-10, but nothing major.

https://docs.google.com/spreadsheets/d/ ... 1004929050

Author:  rmaupin [ Mon Jan 22, 2018 2:16 pm ]
Post subject:  Re: Statistical Rankings

Josh,

Thanks for using the 6 region model, it looks great!

Author:  eckj [ Wed Jan 31, 2018 7:53 pm ]
Post subject:  Re: Statistical Rankings

Small Update

HS: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1248831504
-I added a tab for the OAC Regionals Top 16 (Estimated Regions, not guaranteed) and Geographical Regions Top 16
-Highest climber was Waverly
- Hoping for a big update as leagues finish up!

MS:https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=862958647

EDIT: I also created a Small School tab with ranking for public and private. I would appreciate it if anyone would point out errors or omissions for the small schools.

Author:  eckj [ Sat Feb 03, 2018 6:53 pm ]
Post subject:  Re: Statistical Rankings

HS Update: https://docs.google.com/spreadsheets/d/ ... =292419698
-Northmont climbs into the top ten.
-There are now around 300 teams in the data set!

Author:  eckj [ Sat Feb 10, 2018 6:36 pm ]
Post subject:  Re: Statistical Rankings

HS Update after Olmsted Falls:https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1878447459
-I will have another update this week when the stats for the Ashtabula ESC Tournament are posted.

Author:  eckj [ Thu Feb 22, 2018 9:57 am ]
Post subject:  Re: Statistical Rankings

HS update after Solon: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=220681583

Author:  eckj [ Fri Feb 23, 2018 8:04 pm ]
Post subject:  Re: Statistical Rankings

MS Update after Solon https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=767069257

Highlights:
-Solon debuts at 3
-Ratner moves up three spots to 4th

Author:  eckj [ Mon Feb 26, 2018 9:09 am ]
Post subject:  Re: Statistical Rankings

MS update after Copley: https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=1011097484

Author:  eckj [ Mon Feb 26, 2018 7:21 pm ]
Post subject:  Re: Statistical Rankings

On interesting thing I noticed about the middle school stats is that out of 95 NAQT MS set data points there has only been one team to have a PPB over twenty for a tournament all season long. Seem like our MS circuit is a little weak after looking at other MS results.

Author:  eckj [ Tue Mar 06, 2018 3:05 pm ]
Post subject:  Re: Statistical Rankings

HS Update: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=903325002

Author:  eckj [ Sun Mar 11, 2018 3:31 pm ]
Post subject:  Re: Statistical Rankings

HS UPDATE: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1315482273

Here are the adjustments for each set so far.
Attachment:
adj march.PNG
adj march.PNG [ 8.72 KiB | Viewed 9163 times ]

Author:  eckj [ Sun Mar 18, 2018 5:58 pm ]
Post subject:  Re: Statistical Rankings

Tiny update after Boardman https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1917773160

There are 3 pending HS and 3 pending MS tournaments on NAQT right now so I am hoping for a big update soon!

Author:  eckj [ Sun Apr 08, 2018 9:34 am ]
Post subject:  Re: Statistical Rankings

Update After States

HS:https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1352822314
MS:https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=581556396

Overall thoughts
I am rethinking power% a little. Some changes I expected did not happen because of the power%. I really expected Boardman to jump to 4th based on their strong PPB this weekend. Joe Esposito and I talked and he suggested using Powers Per 20TUH. I am open to suggestions. I want to incorporate powers somehow. I was also thinking about maybe just adding the adjusted P% to the adjusted PPB. This would cause it to impact the ranking, but not too much.

Author:  TwoHeaded_Beast [ Sun Apr 08, 2018 10:51 am ]
Post subject:  Re: Statistical Rankings

William and I would like to suggest two different methods of ranking.

Mine: 2*aP%*aPPB

I didn't think about this as much as William, so this was just a crude way to incorporate P% into the ranking less arbitrarily. Multiplying aP% by two ensures that teams with more powers that tossups benefit. This hurts teams like Miami Valley that generally put up low P%, but I think it does so justifiably. As demonstrated by our performance last year, it is just as (if not more) important to get questions early than to get bonus points. Morlan's system does not address this problem. This is an early version that we have used as the basis for filling out HSQB polls, but this was difficult to do without knowing aP% and not wanting to calculate them.

William's: P/N+(200*aP%+125*aPPB%)/2

This system is based around a 100-point scale. Teams with an aP% above 0.5 will have above 100 in this area. Teams with aPPB% (aPPB/30) above 0.8 (translating to 24 aPPB) will get above 100 in this area. The two are weighted equally for now, although this can be changed to give more emphasis to one or the other, so they are simply averaged. While P/N is generally not a major statistic, adding it to the above value benefits teams that do not neg, but only slightly. The argument is that negs can be important in close games between teams of similar level—adding a number between 0 and 5 or so should not provide a major boost in the rankings, but it can, theoretically, be crucial in determining the outcome of close games. A very high-caliber team (~25 aPPB, ~70 aP%, ~5 P/N) would get around around 127, whereas a good but slightly less competitive (~22.5 aPPB, 40 aP%, 3 P/N) would have around 90. Even if the lower-caliber team had a P/N of 7, their rank would not change significantly.

Having seen William's method, I now prefer it to mine (as it incorporates three variables rather than two), but I can see how others would object to using negs as a statistic. I will be happy to take suggestions or answer questions about both systems!

Author:  eckj [ Sun Apr 15, 2018 6:47 pm ]
Post subject:  Re: Statistical Rankings

Very Small High School Update
https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1086037147

The season is basically over but I will update when there are tournaments.

Overall
-52 Tournaments worth of data
-353 Teams (Not sure how many schools)

Geographically
-Central: Teams (35), Average ARP (16.40)
-Southwest: Teams (50), Average ARP (13.64)
-Northeast: Teams (172), Average ARP (9.98)
-Southeast: Teams (23), Average ARP (9.33)
-Northwest: Teams (73), Average ARP (9.29)

Author:  eckj [ Sun Apr 29, 2018 4:44 pm ]
Post subject:  Re: Statistical Rankings

Rankings "update" after SSNCT.
-Nothing has changed since the last update.
-Something strange, but promising happened. Teams on average (after outliers were removed) performed 1.2 PPB better at SSNCT than their best IS sets. Since my rankings uses the IS sets as the base (standard difficulty) the adjustment for SSNCT was -1.2 PPB.

https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1257099566

Author:  eckj [ Sun May 13, 2018 6:35 pm ]
Post subject:  Re: Statistical Rankings

Last MS update of the year. https://docs.google.com/spreadsheets/d/1WDnH80VrtPUJiGwiRskHas9-W2pBNAXlN0NoebCBjoM/edit#gid=165552030

Congrats to Miami Valley School A for being the top team the entire season!!!

There were quite a few MS tournaments that only did game stats so sadly I was only able to rank 77 teams over 11 tournaments.

I plan on doing this again next year!

The final overall top 10
1. MVS A
2. Copley A
3. Solon A
4. Heights A
5. Ratner A
6. University School A
7. Oakwood A
8. Bexley A
9. Northmont A
10. St. Brendan A

Just for fun!
Top ten B, C, D, or E teams:
1. University School B
2. Heights B
3. Solon B
4. Ratner B
5. St. Hilary B
6. MVS B
7. Copley C
8. Copley B
9. St. Brendan B
10. Revere B

Author:  eckj [ Sun May 27, 2018 8:32 pm ]
Post subject:  Re: Statistical Rankings

Update after HSNCT: https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=1179609304

Author:  eckj [ Sun Jun 03, 2018 3:56 pm ]
Post subject:  Re: Statistical Rankings

Final HS Rankings https://docs.google.com/spreadsheets/d/1p0thgYk4FXer2QnJuoT8FTrmcYWHQIxjZrkkTafnLqQ/edit#gid=849768187

Final 10 teams
1. Beavercreek A
2. Dublin Scioto A
3. Miami Valley A
4. Solon A
5. Boardman A
6. Copley A
7. Bexley A
8. Walnut Hills A
9. Northmont A
10. Ottawa Hills A


Top 10 B, C, D, E
1. Solon B
2. Copley B
3. Beavercreek B
4. St. Ignatius B
5. Walnut Hills B
6. Olentangy Liberty B
7. Little Miami B
8. Cloverleaf B
9. Miami Valley B
10. Beachwood C

Page 1 of 1 All times are UTC - 5 hours
Powered by phpBB® Forum Software © phpBB Group
https://www.phpbb.com/