Handicapping Nationals via spread-based ratings

Come here when you have absolutely nothing to say and the need to say it! :-) Off-topic discussions are welcome.

Handicapping Nationals via spread-based ratings

Postby arenasnow » Tue Jul 16, 2013 8:43 pm

A couple months ago on this thread (viewtopic.php?f=25&t=466) I suggested the idea of spread-based ratings to complement the official Elo style ratings. Following the ideas of Jeff Sagarin who calculates his sports ratings by combining Elo-style ratings (which only take wins and losses into account, like the Scrabble ratings system) and "pure points" or "predictor" ratings (which only take point spread into account and are considered more accurate in predicting future games), I inferred that perhaps creating a spread-based rating would more accurately predict future games than the official NASPA ratings do.

I admit this project was largely out of my own self-interest because I was rather flabbergasted to observe that I have a career winning record in tournaments and average a 35 point win no less, but also notice that my average opponent rating is higher than my own rating. I also noticed Siddharth Murali almost instantly go up to 1600 even though we pretty much split our games (although I would definitely give a slight edge to him, especially lately). This all seemed wrong to me, so I wanted to calculate what my "real" ability was. I know that the reason for this is I notoriously choke games all the time against everyone (if I ever upload my game against Morris Greenberg from last weekend onto cross-tables, you will see me choke a game that even a 700 player would not have lost, but alas). I do not play like a 1300. I play like a 1500 who chokes. I blow out intermediates and novices and am rarely blown out by experts. While I do not argue that I should have a higher rating, and do not argue that ratings SHOULD be spread-based, I do believe that my increasing rating and matching or beating my seed at every tournament (albeit slightly) and my ability to beat most top 100 players on ISC at least 1/4 times and relatively rarely be blown out by them would seem to indicate I am a high, not low, intermediate. Would I be expected to beat a career 1400 player in my next game? Yes, I believe that I would, even though my rating is not 1400 yet. While my current rating is what I deserve due to my perpetual choking, I'd still be expected to beat many players rated higher than I am IN THE NEXT GAME. And when forecasting tournaments (for a roto, etc...) this might provide a more accurate picture for many players.

I realize the formula I used to do this was quite simplistic. When I was statistician of the Scrabble-by-Mail SIG, I calculated what was essentially a performance rating (but I had some stupider term for it, I believe) to measure performance over a given game series: avg. opponent rating + 5*(spread/game). I figured that the ratio between the ratings difference and point spread would be around 4 or 5. It was in fact higher. From a complete set of games Seth Lipkin provided that included at least one 1600+ rated player from 2010 to early July 2013, I calculated ratings difference to spread ratios for TWL and Collins, which were 6.32 and 5.83, respectively. I assume the Collins ratio is lower since it is easier for players to make comebacks, and harder for a higher-rated player to control a game and shut the board down.

I calculated spread based ratings in this manner for every player entered at Nationals. I included only games from July 5, 2012-July 7, 2013 (to include both the Wilmington and Oakland tournaments in 2012 and both the Albany and Las Vegas tournaments in 2013), included only players with 30 or more games (so that everyone entered at Nationals last year will be listed and also because 30 is often considered a large enough sample size to approximate a normal distribution), and only included games in the lexicon that players will be playing at Nationals. I have listed all players in their roto groupings in case you want to more closely ogle the profiles of certain dark horses that I may have finally inadvertently given attention to for your own roto teams. I really think this is capturing something since many of the hottest players (Mike Johnson, Jack Norman, Adam Ginzberg, etc...) have significantly higher spread-based ratings than actual ratings, indicating that their current ratings may not have caught up to their actual ability, but in their cases, it's because they're continuing to improve, not because they choke games like I do. Having said that, Johnson and Norman did not play 30 games in the period so that could have something to do with it too. As a result, I have listed players with extremely high spread-based ratings for each group with less than 30 games as well since many of these players may indeed be dark horse contenders. Some of these may be accurate, some of these may not. I am only listing the top ten for each group in divisions 3 and 4 along with others who have higher ratings than the 10th place player in those groups for brevity.

I do think this is quite a breakthrough, because now you can make predictions based on this to see how many points a player would be expected to win by. For TWL, a player with a rating 623 points higher would be expected to win by 100 points. Nigel Richards at 2244 would be expected to beat Kenji Matsumoto at 2118 by 21 points (yes, that means he would be expected to win every single game against everyone entered by over 20 points...obviously he's not going to, but that is amazing).

Obviously some of these ratings do need to be taken with a grain of salt. My own spread-based rating of 1585 would indicate that I'm better than Mike Baron (and several other clearly superior players) or have been better lately. That's obviously not true, but the same could be said in many cases with the actual ratings system. While I think 1585 overstates it due to my choking, I do think my ability is closer to that than to 1360. Winning is obviously the object of the game, and I have not done a good job of that, but I do think this does a better job of reflecting who's on a roll lately, which would likely be helpful for roto picks.

I won't be attending Nationals myself (I will be attending a typing competition in Cincinnati instead) but good luck to everyone who will be going. Now, on to the ratings (which I entered largely by hand, using only a word processor, spreadsheet, and calculator program - no database queries at all...as a result it's taken me over a week off and on, but I thought some people might be interested).

I also wish there was a board here for Statistics so I could think of somewhere better to put these posts than either here or Strategy, neither of which is a good choice.

TWL Division 1

Nigel Richards - 2244

Group A

Kenji Matsumoto - 2118
Jesse Day - 2116
Mack Meller - 2093
Will Anderson - 2088
Dave Wiegand - 2081
Rafi Stern - 2054
Noah Walton - 2053
Ian Weinstein - 2024
Conrad Bassett-Bouchard - 2005
Cesar del Solar - 2005
Robin Pollock Daniel - 1984
Komol Panyasophonlert - 1971

Group B

Jim Kramer - 2024
Chris Cree - 2024
Joey Mallick - 2018
John Luebkemann - 2007
Carl Johnson - 2007
Joel Sherman - 1995
Joe Edley - 1970
Jason Keller - 1965
Stefan Rau - 1948
Peter Armstrong - 1943
Ben Schoenbrun - 1939
Orry Swift - 1927
Matthew Tunnicliffe - 1901
Mike Frentz - 1877
Lisa Odom - 1811
Scott Appel - 1731

Group C

Paul Holser - 1995
Rob Robinsky - 1975
Leesa Berahovich - 1947
Adam Ginzberg - 1919
Steve Glass - 1901
Jerry Lerman - 1894
John Stardom - 1893
Kevin Rickhoff - 1893
Cecilia Le - 1891
Maddy Kamen - 1888
Chris Williams - 1886
Kate Fukawa-Connolly - 1883
Mark Przybyszewski - 1876
Edward de Guzman - 1871
Jason Li - 1870
Alan Stern - 1865
Nigel Peltier - 1864
Matt Canik - 1861
Darrell Day - 1851
Erickson Smith - 1849
Scott Jackson - 1845
David Whitley - 1842
Zev Kaufman - 1826
Lou Cornelis - 1812
Ross Brown - 1811
Charles Reinke - 1808
Chris Patrick Morgan - 1807
Greg Heidler - 1795
Trip Payne - 1784
Scott Garner - 1773
Joey Krafchick - 1770
Dielle Saldanha - 1767
Ryan Fischer - 1764
Michael Thelen - 1762
Mike Howlett - 1734
Mark Milan - 1730
Thomas Reinke - 1722
Kolton Koehler - 1718
Jeremy Hall - 1695
Jeff Fiszbein - 1670

Noteworthy with less than 30 games:

Amit Chakrabarti - 1983

Group D

Joseph Gaspard - 1833
Jason Ubeika - 1792
Jesse Matthews - 1752
Jan Cardia - 1743
Brian Galebach - 1735
Timothy Bottorff - 1731
Michael Baker - 1728
Ken Kasney - 1692
Darin True - 1689
Robert Kahn - 1688
Lawren Freebody - 1686
Michael Early - 1684
Matthew Ridout - 1684
Jim Burlant - 1673
Juraj Pivovarov - 1658
Dave Leifer - 1658
Richard Popper - 1657
Emely Weissman - 1656
Roy Kamen - 1648
Purinth Thitiyagoon - 1646
Gunther Jacobi - 1644
Wallace Schultz - 1630
Terry Kang Rau - 1629
Melissa Routzahn - 1621
Howard Pistol - 1614
Amnuay Ploysangngam - 1609
KC Frodyma - 1579
Avery Mojica - 1565
Yukiko Loritz - 1564
Mike Baron - 1553
Win Inthara - 1550
Liz Gottlin - 1522
Jeffrey Nelson - 1514
Cheryl Tyler - 1512
Wayne Clifford - 1500
Bruce D'Ambrosio - 1489
Siri Tillekeratne - 1488
Judy Newhouse - 1470
Noella Ward - 1469
Scott Leifer - 1467
Robert Felman - 1340
Danny Panganiban - 1317

Collins Division

Group A

Geoff Thevenot - 2063
John O'Laughlin - 2056
Evan Berofsky - 2051
Mark Kenas - 2025
Sam Kantimathi - 2020
Brian Bowman - 1994
Joel Wapnick - 1968
Travis Chaney - 1967
Chris Lipe - 1935
Evans Clinchy - 1931
Marty Gabriel - 1916
Steve Polatnick - 1846
Laurie Cohen - 1802
Geoffrey Newman - 1800

Noteworthy with less than 30 games:

David Eldar - 2346
Quinn James - 2139

Group B

Winter - 1834
Bradley Whitmarsh - 1808
Robert Linn - 1758
Ben Withers - 1740
Becky Dyer - 1733
Tony Leah - 1731
Orlet Bullock - 1709
Carl Durdan - 1704
Martin Demello - 1656
Puneet Sharma - 1612
JIm Lamerand - 1580
Zbigniew Wieckowski - 1494
Sandy Nang - 1469
Daniel Milton - 1411

TWL Division 2 Group A

Luise Shafritz - 1736
Mark Rosenberg - 1731
Doug Lundquist - 1722
Gerry Greenside - 1709
Chris Tallman - 1700
Chris Schneider - 1693
Morris Greenberg - 1685
Sal Piro - 1648
Judy Romann - 1627
Robin Lewis - 1609
Chris Canik - 1584
Carl Stocker - 1564
Siddharth Murali - 1532
David Engelhardt - 1524

Noteworthy with less than 30 games:

Rahn McKeown - 2044

Group B

John Dalton - 1608
Harriette Lakernick - 1603
Michael Kapernaros - 1598
Mark Fidler - 1598
Michael Stevens - 1596
Connie Creed - 1595
Joshua Sokol - 1587
Steve Anstandig - 1561
Glenn Dunlop - 1559
Martin Weisskopf - 1559
Chuck Riordan - 1556
Cynthia Pughsley - 1545
Sam Towne - 1542
Murray Weber - 1537
Arun Methaset - 1535
Joanne Cohen - 1529
James Porter - 1526
Stefan Fatsis - 1511
Kay Patterson - 1510
Ricky Sirois - 1508
Cynthia Seales - 1508
Bob Becker - 1489
Diana Grosman - 1479
Curran Eggertson - 1479
Betty Burton - 1474
Paul Avrin - 1463
Heather Steffy - 1459
Heather McCall - 1458
Gary Perman - 1457
Curtis Lee - 1446
Jeanmarie Youngblood - 1444
Kit Morehead - 1443
Roger Cullman - 1440
Paula Catanese - 1435
Lila Crotty - 1432
Tracy Bowman - 1431
Mariah Smith - 1424
Carole Denton - 1380
Lynda Wise - 1267

Noteworthy with less than 30 games:

Mike Johnson - 1840
Jack Norman - 1816

Group C

Arthur Finkle - 1564
Robert Fenske - 1498
Mike Ebanks - 1494
Nick Fraher - 1481
George Asaka - 1478
Ruchi Gupta - 1476
Jeremy Jeffers - 1470
Frederick Schneider - 1449
Carolyn Easter - 1433
Tobey Roland - 1424
Carl Davis - 1413
Trevor Sealy - 1397
Priya Fernando - 1390
Andrea Michaels - 1386
Marty Fialkow - 1374
Richard Strick - 1371
Supriya Devnani - 1370
Joe Neff - 1367
Will Scott - 1363
Brenda Casey - 1360
Alan Kait - 1357
Lilla Sinanan - 1351
Joan Kelly - 1350
Anna Miransky - 1345
John Fultz - 1344
Ryan Burgess - 1342
Linda Wancel - 1341
Caesar Jaramillo - 1334
Linda Oliva - 1316
Doug Szymkowiak - 1304
Tim Weiss - 1303
Judy Cole - 1299
Betty Cornelison - 1297
Geoff Gibson - 1269
Cheryl Melvin - 1258
Nancy Scott - 1236
Bruce Shuman - 1227
George Bissonnette - 1214
Michael Bassett - 1208
Carole Miller - 1199
Ellen Perr - 1179
Mary Ellen Weisskopf - 1153
Samantha Orleans - 1136
Betsey Wood - 1129

Noteworthy with less than 30 games:

Max Dwyer - 1559

TWL Division 3 Group A

John Price - 1505
Gary Skowronski - 1495
Alex Rodriguez - 1402
Ted Bryde - 1361
Steve Moniz - 1357
Amalan Iyengar - 1355
April McCarley - 1349
Marla Riff - 1347
Huguette Settle - 1342
Stan Khury - 1337

Noteworthy with less than 30 games:

Randy Urist - 1446
Ryan Chepita - 1445
Miriam Green - 1398

Group B

Clay Koziol - 1437
Andy Hoang - 1380
Mary Pastore - 1357
Rhonda Roederer - 1344
Alexander Rivard - 1340
Robb Griffith - 1297
Edward Gowash - 1291
Miriam Gregory - 1286
Guy Ingram - 1283
Ingrid Hawkins - 1273

Noteworthy with less than 30 games:

Jason Steuber - 1450
Andy Murphy - 1434

TWL Division 4

Chris Autio - 1194
Daniel Novinson - 1155
Phil Seitzer - 1142
Michael Krafchick - 1089
Conchita Deprospo - 1064
Marvin Kraus - 1043
Sylvia Polson - 1026
Charles Dupont - 1021
Marie Filandro - 995
Jim Laris - 992

Noteworthy with less than 30 games:

Daniel Moglen - 1476
Kenny Hoang - 1138
David Baran - 1137
Jennifer Zinn - 1090
Diana Bowen - 1088
Kevin Bowerman - 1054
Robert Readle - 1037
Diane Kerner - 1026
Ed Morrison - 1007
Last edited by arenasnow on Tue Jul 16, 2013 9:28 pm, edited 1 time in total.
arenasnow
 
Posts: 71
Joined: Wed Dec 31, 2008 12:42 pm
Location: Syracuse, NY

Re: Handicapping Nationals via spread-based ratings

Postby cesar » Tue Jul 16, 2013 9:22 pm

what's my rating? :(
cesar
 
Posts: 101
Joined: Wed Dec 31, 2008 5:04 pm
Location: Los Angeles

Re: Handicapping Nationals via spread-based ratings

Postby arenasnow » Tue Jul 16, 2013 9:28 pm

You 2005, RPD 1984. I can't believe I somehow combined them into one line like that.
arenasnow
 
Posts: 71
Joined: Wed Dec 31, 2008 12:42 pm
Location: Syracuse, NY

Re: Handicapping Nationals via spread-based ratings

Postby snowplow48 » Wed Jul 17, 2013 12:22 pm

[Edit: Deleted for not fully reading post]
User avatar
snowplow48
 
Posts: 32
Joined: Wed Jul 27, 2011 7:15 pm


Return to From AA to ZZZ

Who is online

Users browsing this forum: No registered users

cron