Individual player stat v. 29192

Old college threads.
Locked
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Individual player stat v. 29192

Post by theMoMA »

I literally came up with the idea for this while asleep flying out to VCU Open, so feel free to merge this with the "More Quizbowl Dreams" thread if you want.

The theory behind this stat is somewhat similar to PATH and another stat that I've previously worked out, but a little more comprehensive. It's based on a simple measure:
Equation 1) individual tossup percentage

player tossups converted / (tossups heard - teammate buzzes)
Teammate buzzes are equal to the number of teammate 15s + 10s + -5s. The above measure can be conceptualized as player tossups converted when the player is playing one-on-four against the other team and the packet (the denominator is equal to opponent tossups converted plus dead tossups). I think this is a reasonable way to isolate individual performance, and it's very similar to the idea behind PATH.

From there, I thought to compute four measures, each normalized to a twenty-tossup sample. Two are very simple: The number of points added for getting 15s, and the number of points subtracted for getting -5s (only accounting for the marginal point value of the two).
Equation 2) value of 15s

5 * (number of 15s / (tossups heard - teammate buzzes)) * 20
This computes the value of all of those five-point bonuses that come along with a 15.
Equation 3) deduction of -5s

-5 * (number of -5s / (tossups heard - teammate buzzes)) * 20
This computes the value of all of those five-point penalties that players incur on -5s.

The third value is also fairly simple: It simply computes the raw positive value of getting a tossup:
Equation 4) value of raw tossups

(10 + team PPB) * (player tossups converted / (tossups heard - teammate buzzes)) * 20
The logic here is that a player's buzz (with the five-point power reward already taken care of above) is worth 10 points plus the expected PPB.

The final calculation measures the value of expected points taken away by a player's negs. The better a team, and the better a player's teammates, the more detrimental a player's negs are.
Equation 5) deduction of -5 opportunity cost

(team tossups answered / (tossups heard - player buzzes)) * (number of -5s / (tossups heard - teammate buzzes)) * (10 + team PPB) * 20
The four terms in this equation can basically be thought of as:
chance that teammates get the tossup if other players don't buzz * player -5 percentage * (10 + expected PPB) * 20
The final stat sums the results of Equations 2 through 5 above to get what I would say amounts to "expected player points contributed against four average players in the field over twenty tossups, if the player in question were the only one able to buzz on his/her team but the rest of the teammates could consult on bonuses."
Andrew Hart
Minnesota alum
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

For discussion, here's this measure as applied to CO stats:

204.96 Jonathan Magin
189.10 Seth Teitler
174.42 Selene Koo
172.91 Jerry Vinokurov
169.71 Matt Weiner
166.56 Jeff Hoppes
165.68 John Lawrence
164.46 Matt Bollinger
160.76 Ted Gioia
160.48 Brendan Byrne
151.57 Ike Jose
149.31 Auroni Gupta
137.78 Chris Ray
134.36 Rob Carson
130.18 Andrew Hart
128.80 Gautam Kandlikar
127.62 Matt Lafer
118.94 Matt Jackson
117.08 Kurtis Droge
111.34 Dallas Simons
111.11 Mike Sorice
106.24 Kevin Koai
101.86 Mik Larsen
98.20 Will Butler
92.26 Guy Tabachnick
83.05 Chris Borglum
77.50 Mike Cheyne
72.14 Bruce Arthur
70.65 Marshall Steinbaum
69.60 Michael Arnold
69.41 Andrew Ullsperger
62.02 Billy Beyer
61.49 Paul Gauthier
57.96 Charlie Dees
53.34 Ahmad Ragab
50.90 Charles Hang
48.92 Greg Peterson
45.73 Michael Hausinger
45.61 Libo Zeng
45.49 Seth Kendall
44.35 Surya Sabhapathy
40.29 Trevor Davis
36.31 Carsten Gehring
32.49 Mike Bentley
29.22 Richard Mason
26.68 Jimmy Ready
24.27 Rom Masrour
19.63 David Seal
18.76 Nick Polk
16.10 Bryan Berend
15.54 Joe Hansen
14.86 Dan Passner
13.75 Emily Koenig
11.66 Charles Martin
7.77 Michael Schreiber
7.59 Margo Emont
-0.68 Adrian Blaszkiewicz
Andrew Hart
Minnesota alum
Windows ME
Wakka
Posts: 177
Joined: Fri Dec 19, 2008 9:06 am

Re: Individual player stat v. 29192

Post by Windows ME »

theMoMA wrote:I literally came up with the idea for this while asleep flying out to VCU Open, so feel free to merge this with the "More Quizbowl Dreams" thread if you want.

The theory behind this stat is somewhat similar to PATH and another stat that I've previously worked out, but a little more comprehensive. It's based on a simple measure:
Equation 1) individual tossup percentage

player tossups converted / (tossups heard - teammate buzzes)
Teammate buzzes are equal to the number of teammate 15s + 10s + -5s. The above measure can be conceptualized as player tossups converted when the player is playing one-on-four against the other team and the packet (the denominator is equal to opponent tossups converted plus dead tossups). I think this is a reasonable way to isolate individual performance, and it's very similar to the idea behind PATH.
What if you're a specialist and someone on your team is negging several of your tossups?

Not sure there is any stat ever that can account for that. -_-
Sinan U.
U of Toronto 2010
U of Alberta 2015
User avatar
Mechanical Beasts
Banned Cheater
Posts: 5673
Joined: Thu Jun 08, 2006 10:50 pm

Re: Individual player stat v. 29192

Post by Mechanical Beasts »

Well, this stat is an individual player stat but it does not attempt to be very team-independent. It tries to describe a player's total value contribution (by incorporating team PPB) and obviously if you're on a team where you can't answer tossups because your teammate negs all the science (or gets all the science, and you only know... science) then you aren't actually contributing much. You might be great if you were on another team, but you're not--and I don't think this stat is supposed to try to measure that.
Andrew Watkins
Black-throated Antshrike
Rikku
Posts: 379
Joined: Fri Jan 22, 2010 12:47 am

Re: Individual player stat v. 29192

Post by Black-throated Antshrike »

Mechanical Beasts wrote:Well, this stat is an individual player stat but it does not attempt to be very team-independent. It tries to describe a player's total value contribution (by incorporating team PPB) and obviously if you're on a team where you can't answer tossups because your teammate negs all the science (or gets all the science, and you only know... science) then you aren't actually contributing much. You might be great if you were on another team, but you're not--and I don't think this stat is supposed to try to measure that.
This also gives players a boost if their team is carrying them on bonuses
Joe
Delaware
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Andrew Jackson's Compatriot wrote:This also gives players a boost if their team is carrying them on bonuses
That's true, though realize that the better a team is, the more point are on the line with each buzz. Even if you're contributing nothing to your team's bonus efforts, your team can't answer bonuses without someone getting the tossup first, so in that sense it's reasonable to say that the player who buzzes should get credit for the team's efforts that stem from the player's buzzes.
Andrew Hart
Minnesota alum
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Mechanical Beasts wrote:Well, this stat is an individual player stat but it does not attempt to be very team-independent. It tries to describe a player's total value contribution (by incorporating team PPB) and obviously if you're on a team where you can't answer tossups because your teammate negs all the science (or gets all the science, and you only know... science) then you aren't actually contributing much. You might be great if you were on another team, but you're not--and I don't think this stat is supposed to try to measure that.
Andy is somewhat right, but it's true that I did claim that this measure would show how a player does playing alone against an average opponent, and it would underrate a player whose specialties the rest of the team was consistently negging. It doesn't seem like this is correctable given basic input data. I'll also note that this measure assumes that, absent the player's own neg, the rest of the team had the probability of (rest of team tossups answered / rest of team tossups heard) of picking up the tossup. This assumption isn't perfect because players tend to buzz on their own specialty categories, and the rest of the team likely has a lesser chance of picking up questions in those categories. It also does not account for the fact that, had the player in question not negged, that player still would have had a chance to pick up the tossup later. These two non-factors work in opposite directions, so my hope is that they cancel out most of each other's effects.
Andrew Hart
Minnesota alum
User avatar
grapesmoker
Sin
Posts: 6345
Joined: Sat Oct 25, 2003 5:23 pm
Location: NYC
Contact:

Re: Individual player stat v. 29192

Post by grapesmoker »

Finally, a stat that shows Selene Koo to be the elite player we all knew she was.

edit: in case this isn't clear, I am 100% serious.
Jerry Vinokurov
ex-LJHS, ex-Berkeley, ex-Brown, sorta-ex-CMU
presently: John Jay College Economics
code ape, loud voice, general nuissance
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

I admittedly haven't taken the time (probably 10 minutes) to write this out and see fully why it's true, but it seems like this stat is really just a convenient way to capture power to neg ratio (actually 10s-and-15s-to-negs ratio) alongside player PPG/team PPG in one stat. For example, Selene shows up much higher than she did on the regular PPG-based rankings because of her amazing 0 negs while contributing 40 questions, 7 of which are powered. On the other side of things, Chris Ray and Ike Jose get pushed down from the standard individual rankings because their large neg totals (26 and 25) eliminated chances for their teammates to buzz, therefore lowering their "contribution" as Andy put it. I think this stat would be especially useful for high school coaches looking to pick the optimal 4-person line-up out of 5 or 6 people, since it's a quick way to quantify how much each possible third and fourth chair helps the primary scorers.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
Matt Weiner
Sin
Posts: 8145
Joined: Fri Apr 11, 2003 8:34 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Matt Weiner »

I remain convinced that no amount of algebraic wizardry can compensate for the fact that the underlying data of quizbowl statistics are rather sparse, and will remain so until some sort of integrated reading/scorekeeping system that captures buzzpoints and subjects is developed.
Matt Weiner
Advisor to Quizbowl at Virginia Commonwealth University / Founder of hsquizbowl.org
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

Matt Weiner wrote:I remain convinced that no amount of algebraic wizardry can compensate for the fact that the underlying data of quizbowl statistics are rather sparse, and will remain so until some sort of integrated reading/scorekeeping system that captures buzzpoints and subjects is developed.
This is definitely true to the extent that stats like this will never be able to demonstrate winners of hypothetical one-on-ones unless we have a huge singles event played in that "goldfish" style of noting buzzpoints on identical questions in each subject (and even with this scenario stats like this one will be necessary to decide how to of the "best" individuals would play together). However, I totally support the idea of using them internally to judge what lineups will work best. While this can't be called an "individual player stat" in the sense that the holder of the highest one is the "best" individual, it can tell you in the right context who will work out best on a given team.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
cvdwightw
Auron
Posts: 3291
Joined: Tue May 13, 2003 12:46 am
Location: Southern CA
Contact:

Re: Individual player stat v. 29192

Post by cvdwightw »

My favorite thing about this stat is that you get an indefinite player rating if you put up zero tossup points on a team that grails.
Dwight Wynne
socalquizbowl.org
UC Irvine 2008-2013; UCLA 2004-2007; Capistrano Valley High School 2000-2003

"It's a competition, but it's not a sport. On a scale, if football is a 10, then rowing would be a two. One would be Quiz Bowl." --Matt Birk on rowing, SI On Campus, 10/21/03

"If you were my teammate, I would have tossed your ass out the door so fast you'd be emitting Cerenkov radiation, but I'm not classy like Dwight." --Jerry
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Another way to think of this stat is essentially asking how well a player maximizes his/her niche on a team. Ideally, you'd have four players who, when pitted against the other team, each maximize their buzzing percentage to a similar extent. For example, it looks like the cutoff for being well above-average at CO was getting about 24% of the tossups that your teammates don't buzz on. So if you've got really good teammates who are buzzing on 70% of the questions instead of 40%, you only have to pick up 7.2% of total questions (instead of 14.4% of total questions) to be in that well-above-average category. That makes some empirical sense to me; this measure is essentially looking for players who can, like I said, maximize their niche.
Andrew Hart
Minnesota alum
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

The above post just inspired me to find another measure, which is essentially how many standard deviations a player is above the average for the scoring role that they played for their team. In other words, how much of an advantage is it to have Seth Teitler over the average #1 scorer on a CO team, or to have Gautam Kandlikar over the average #4 scorer. Here's what I came up with (after adding in three chairs to fill out the Rom and Hang teams). Apologies for the inadvertent self-aggrandizing; this is no measure of skill, simply a measure of the advantage that a player provides over the average player who has the same scoring role (in fact, for players on the same team, it operates as a sort of anti-skill measure, because the better you do, the tougher it is to do well in this measure). As you can see, it appears that overall, the strongest teams get their biggest advantage from having advantages at the quaternary and tertiary positions than their opponents (which makes a lot of sense).

Code: Select all

Player	SUM	tRANK	stDEV
Andrew Hart	130.18	4	1.91
Gautam Kandlikar	128.80	4	1.88
Matt Bollinger	165.82	3	1.74
Jeff Hoppes	165.42	3	1.73
Selene Koo	174.42	2	1.65
Mike Sorice	115.07	4	1.57
John Lawrence	167.16	2	1.51
Jonathan Magin	204.96	1	1.40
Auroni Gupta	149.31	2	1.19
Seth Teitler	189.10	1	1.13
Matt Lafer	127.62	3	1.02
Jerry Vinokurov	172.91	1	0.86
Matt Weiner	169.71	1	0.80
Dallas Simons	111.34	3	0.72
Matt Jackson	121.39	2	0.69
Ted Gioia	160.76	1	0.65
Brendan Byrne	160.48	1	0.65
Kurtis Droge	117.08	2	0.61
Ike Jose	151.57	1	0.50
Will Butler	98.20	3	0.47
Kevin Koai	106.24	2	0.42
Guy Tabachnick	92.26	3	0.36
Chris Ray	137.78	1	0.27
Rob Carson	134.36	1	0.21
Ahmad Ragab	53.34	4	0.19
Libo Zeng	45.61	4	0.02
Seth Kendall	45.49	4	0.02
Andrew Ullsperger	69.41	3	-0.07
Charlie Dees	67.91	3	-0.10
Mike Cheyne	77.50	2	-0.10
Trevor Davis	40.29	4	-0.10
Bruce Arthur	72.14	2	-0.19
Billy Beyer	62.02	3	-0.21
Marshall Steinbaum	70.65	2	-0.22
Michael Arnold	69.60	2	-0.24
Mike Bentley	32.49	4	-0.27
Mik Larsen	101.86	1	-0.33
Richard Mason	29.22	4	-0.35
Paul Gauthier	61.24	2	-0.39
Michael Hausinger	45.73	3	-0.51
Surya Sabhapathy	44.35	3	-0.54
Nick Polk	18.76	4	-0.58
Bryan Berend	16.10	4	-0.64
Chris Borglum	83.05	1	-0.65
Michael Schreiber	7.77	4	-0.83
Margo Emont	7.59	4	-0.83
David Seal	19.63	3	-1.00
Jimmy Ready	26.68	2	-1.01
Emily Koenig	13.75	3	-1.11
Charles Martin	11.66	3	-1.15
Charles Hang	50.90	1	-1.19
Joe Hansen	15.54	2	-1.21
Dan Passner	14.86	2	-1.22
Greg Peterson	48.92	1	-1.22
Carsten Gehring	36.31	1	-1.43
Adrian Blaszkiewicz	-0.68	2	-1.50
Rom Masrour	24.27	1	-1.64
Also, it's somewhat cool that you can see how the chairs stack up. In these stats, having a chair instead of an average fourth scorer provides -1 standard deviation of value, while having a chair instead of an average third is worth -1.37 stdevs.
Andrew Hart
Minnesota alum
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Okay, I came up with a follow-up to the preceding stat that should normalize things a bit. I found out the percentages of total points contributed by #1 scorers, #2 scorers, etc. and I used those percentages (with #1 scorers set to a multiplier of 1) to weigh the standard deviation values. In other words, because #4 scorers only account for about 37% of the scoring burden as their #1 scoring counterparts, I multiplied the standard deviation by .37 of #4 scorers (and made adjustments to #2s and #3s as well) to give a rough weight that accounts for the burden of being a #1 scorer.

Code: Select all

weighted	Player
1.40	Jonathan Magin
1.13	Seth Teitler
1.12	Selene Koo
1.04	Matt Bollinger
1.04	Jeff Hoppes
1.03	John Lawrence
0.86	Jerry Vinokurov
0.81	Auroni Gupta
0.80	Matt Weiner
0.70	Andrew Hart
0.69	Gautam Kandlikar
0.65	Ted Gioia
0.65	Brendan Byrne
0.61	Matt Lafer
0.58	Mike Sorice
0.50	Ike Jose
0.47	Matt Jackson
0.43	Dallas Simons
0.42	Kurtis Droge
0.29	Kevin Koai
0.28	Will Butler
0.27	Chris Ray
0.22	Guy Tabachnick
0.21	Rob Carson
0.07	Ahmad Ragab
0.01	Libo Zeng
0.01	Seth Kendall
-0.04	Trevor Davis
-0.04	Andrew Ullsperger
-0.06	Charlie Dees
-0.07	Mike Cheyne
-0.10	Mike Bentley
-0.12	Billy Beyer
-0.13	Richard Mason
-0.13	Bruce Arthur
-0.15	Marshall Steinbaum
-0.16	Michael Arnold
-0.21	Nick Polk
-0.23	Bryan Berend
-0.26	Paul Gauthier
-0.30	Michael Schreiber
-0.30	Margo Emont
-0.31	Michael Hausinger
-0.32	Surya Sabhapathy
-0.33	Mik Larsen
-0.60	David Seal
-0.65	Chris Borglum
-0.67	Emily Koenig
-0.69	Jimmy Ready
-0.69	Charles Martin
-0.82	Joe Hansen
-0.83	Dan Passner
-1.02	Adrian Blaszkiewicz
-1.19	Charles Hang
-1.22	Greg Peterson
-1.43	Carsten Gehring
-1.64	Rom Masrour
Andrew Hart
Minnesota alum
User avatar
Maxwell Sniffingwell
Auron
Posts: 2164
Joined: Sun Feb 12, 2006 3:22 pm
Location: Des Moines, IA

Re: Individual player stat v. 29192

Post by Maxwell Sniffingwell »

Cool stuff. According to this last stat (correct me if I'm wrong): Rob Carson/Mike Cheyne/Andrew Ullsperger/Seth Kendall would be the most average-for-CO team possible (barring category overlap).
Greg Peterson

Northwestern University '18
Lawrence University '11
Maine South HS '07

"a decent player" - Mike Cheyne
User avatar
Cheynem
Sin
Posts: 7220
Joined: Tue May 11, 2004 11:19 am
Location: Grand Rapids, Michigan

Re: Individual player stat v. 29192

Post by Cheynem »

These stats make sense.To what extent is this stat logical in evaluating teams with extremely balanced scorers? I've played on teams where all 4 people basically scored 20 PPG--how useful is it to compare each player to the average #1, #2, #3, and #4 scorers, in that sense (in this example, it seems like the player who by luck ended up #4 scorer would do very well in the statistic).
Mike Cheyne
Formerly U of Minnesota

"You killed HSAPQ"--Matt Bollinger
User avatar
Mechanical Beasts
Banned Cheater
Posts: 5673
Joined: Thu Jun 08, 2006 10:50 pm

Re: Individual player stat v. 29192

Post by Mechanical Beasts »

I don't see a good point in comparing third scorers, for example. Unlike in baseball, where you really do need a second baseman, if Chicago C loses its third-most-valuable player it wouldn't pass on Matt Weiner to fill in. So, not to attack your inadvertent example, but the reason you're 1.91 standard deviations above a fourth-place scorer just means you were on a really good team: on most other teams, you wouldn't have been a fourth place scorer. Indeed, apply this model to, say, the history tournament that had mostly pairs and a couple teams of three: those third scorers look like gods by this stat for no good reason. Unless you can separate by actual "positional" roles, like categories specialties, "best player who is worse than two or more players on his team" won't mean much.
Andrew Watkins
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

Cheynem wrote:These stats make sense.To what extent is this stat logical in evaluating teams with extremely balanced scorers? I've played on teams where all 4 people basically scored 20 PPG--how useful is it to compare each player to the average #1, #2, #3, and #4 scorers, in that sense (in this example, it seems like the player who by luck ended up #4 scorer would do very well in the statistic).
EDIT: The answer to this is Rob Carson/Kevin Koai/Will Butler/Trevor Davis. They were average in team PPG and higher-than-average Bonus conversion, and they had the most clustered in player PPG (32.14/24.64/23.57/10 respectively). As a result, WIll and Kevin came out higher than Rob.
Also, what Andy said.
Last edited by Tower Monarch on Tue Aug 16, 2011 4:12 pm, edited 1 time in total.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Mechanical Beasts wrote:I don't see a good point in comparing third scorers, for example. Unlike in baseball, where you really do need a second baseman, if Chicago C loses its third-most-valuable player it wouldn't pass on Matt Weiner to fill in. So, not to attack your inadvertent example, but the reason you're 1.91 standard deviations above a fourth-place scorer just means you were on a really good team: on most other teams, you wouldn't have been a fourth place scorer. Indeed, apply this model to, say, the history tournament that had mostly pairs and a couple teams of three: those third scorers look like gods by this stat for no good reason. Unless you can separate by actual "positional" roles, like categories specialties, "best player who is worse than two or more players on his team" won't mean much.
I don't think that the "positions" have any inherent meaning (for example, you could make Seth Teitler the "fourth" scorer on our CO team and make me the "first scorer" and our standard deviations would still add up to approximately the same thing; for what it's worth, I did this and it equals out to 1.13 Seth, 1.91 me in the original and 2.80 Seth, .22 me if we're flipped). It's just an illustration of how much a player added over players who were in the same sequential order of scoring. The order is merely an arbitrary one, so I don't think it has much to tell us about individual skill so much as how much a player adds over the person who is arbitrary "lined up" across from him/her. I believe adding up the stdevs across a team would be a great measure of team strength, however.

edit: it's probably possible to eliminate the arbitrary "lining up" factor by measuring how close a player is to being "lined up" to the second scorer. For example, looking at Matt Weiner's team, he's the "first scorer" by virtue of only a couple of points, while Matt Bollinger is the third scorer by virtue of only a couple of points (and John Lawrence is in the middle, again by a very small difference). You can probably create a hybrid position value that would illustrate that these guys are all relatively close to one another, and would also account for teams in which there was a little more dispersion. The way you'd do this would likely create a sliding scale in which the "average #1 player" for someone who's relatively close to their team's #2 player would be adjusted to be somewhere between "average #1 player" and "average #2 player."
Andrew Hart
Minnesota alum
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

theMoMA wrote: I believe adding up the stdevs across a team would be a great measure of team strength, however.
This is true, but it ends up being almost equivalent to a stat that could be found much faster using team bonus conversion and team 15s/10s/5s instead of individuals.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Tower Monarch wrote:
theMoMA wrote: I believe adding up the stdevs across a team would be a great measure of team strength, however.
This is true, but it ends up being almost equivalent to a stat that could be found much faster using team bonus conversion and team 15s/10s/5s instead of individuals.
Right, the advantage here is that you can line up the players in any way you want to see where the surplus value is coming from. It also lets you see who's playing "out of position." For example, someone like Gautam, who contributes roughly an average number of points for a #1 player according to these stats, is particularly advantageous to have as a #4 player, whereas someone like Rom, who would stack up favorably against #4s, struggles as a leading scorer.
Andrew Hart
Minnesota alum
Susan
Forums Staff: Administrator
Posts: 1812
Joined: Fri Aug 15, 2003 12:43 am

Re: Individual player stat v. 29192

Post by Susan »

theMoMA wrote: It also lets you see who's playing "out of position." For example, someone like Gautam, who contributes roughly an average number of points for a #1 player according to these stats, is particularly advantageous to have as a #4 player, whereas someone like Rom, who would stack up favorably against #4s, struggles as a leading scorer.
Thank God that we now have a tool to help us discern these formerly impossible-to-reach conclusions!
Susan
UChicago alum (AB 2003, PhD 2009)
Member emerita, ACF
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

Susan wrote:
theMoMA wrote: It also lets you see who's playing "out of position." For example, someone like Gautam, who contributes roughly an average number of points for a #1 player according to these stats, is particularly advantageous to have as a #4 player, whereas someone like Rom, who would stack up favorably against #4s, struggles as a leading scorer.
Thank God that we now have a tool to help us discern these formerly impossible-to-reach conclusions!
This is essentially my reaction.
For some reason, I'm missing why a flat .37 across the board was a better multiplier for the #4s instead of, say, #4's PPG/#1's, or pretty much any other team-by-team approach that doesn't force a team to conform to the average distribution of player abilities. To me, it's those weights that make this last list less than helpful for practical purposes (whatever those may be).
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
Cheynem
Sin
Posts: 7220
Joined: Tue May 11, 2004 11:19 am
Location: Grand Rapids, Michigan

Re: Individual player stat v. 29192

Post by Cheynem »

Would this be any more insightful by being applied to regular season lineups? I feel like at opens most people are aware that the presence of superteams is going to cause people to play out of position, but I don't know if it's perhaps more interesting to note that "Hey, the third scorer at Dickety Doo U is actually way above most team's third scorers and could probably carry a team." Then again, I suppose this is pretty intuitive too (I'm sure most people were aware that Gautam Kandlikar, a fourth and third scorer the last two seasons at Minnesota, is a really good player), but perhaps not as much as one would think as third/fourth scorers, in my opinion, are generally underrated in the infallible player poll.
Mike Cheyne
Formerly U of Minnesota

"You killed HSAPQ"--Matt Bollinger
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

Cheynem wrote:Would this be any more insightful by being applied to regular season lineups? I feel like at opens most people are aware that the presence of superteams is going to cause people to play out of position, but I don't know if it's perhaps more interesting to note that "Hey, the third scorer at Dickety Doo U is actually way above most team's third scorers and could probably carry a team." Then again, I suppose this is pretty intuitive too (I'm sure most people were aware that Gautam Kandlikar, a fourth and third scorer the last two seasons at Minnesota, is a really good player), but perhaps not as much as one would think as third/fourth scorers, in my opinion, are generally underrated in the infallible player poll.
Yeah, if you made up a list of the top regular season players based on who put up high PPG at ACF Regionals and somehow missed someone like Gautam, then this stat (and presumably many others) could give the necessary insight to find the key 3rd-bests. The problem is that, along the lines of Matt Weiner's objections, the importance of subject matter expertise in team formations outweighs all of the subtle differences that this stat and all currently-implementable ones measure. Mehdi Razvi is going to be undervalued if he's been playing with Eric too much, since alone he can clean up on science questions that rarely make it to him.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
Mechanical Beasts
Banned Cheater
Posts: 5673
Joined: Thu Jun 08, 2006 10:50 pm

Re: Individual player stat v. 29192

Post by Mechanical Beasts »

Tower Monarch wrote:
Cheynem wrote:Would this be any more insightful by being applied to regular season lineups? I feel like at opens most people are aware that the presence of superteams is going to cause people to play out of position, but I don't know if it's perhaps more interesting to note that "Hey, the third scorer at Dickety Doo U is actually way above most team's third scorers and could probably carry a team." Then again, I suppose this is pretty intuitive too (I'm sure most people were aware that Gautam Kandlikar, a fourth and third scorer the last two seasons at Minnesota, is a really good player), but perhaps not as much as one would think as third/fourth scorers, in my opinion, are generally underrated in the infallible player poll.
Yeah, if you made up a list of the top regular season players based on who put up high PPG at ACF Regionals and somehow missed someone like Gautam, then this stat (and presumably many others) could give the necessary insight to find the key 3rd-bests. The problem is that, along the lines of Matt Weiner's objections, the importance of subject matter expertise in team formations outweighs all of the subtle differences that this stat and all currently-implementable ones measure. Mehdi Razvi is going to be undervalued if he's been playing with Eric too much, since alone he can clean up on science questions that rarely make it to him.
The other issue is that unless Dickety Doo U is playing a nationally competitive field, the third scorer is going to have his basic stat _and_ these stats inflated. It'll tell you he could probably carry Dickety Doo State, but nothing more than ordinarily about carrying Illinois.
Andrew Watkins
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Underrated in what sense? These stats are only meant to illuminate what happened, not what will happen in the future, so it may be true that on some other team, a Mehdi-like player playing with an Eric-like player might score more points, but that's not really the purpose of this measure. Also, for what it's worth, the forays into "position" stats are just something that I concocted based on the original idea; I don't think they're particularly valuable, just perhaps one interesting way to slice the data (that, as Susan points out, seems to tell us many things that we already know, for whatever that's worth). The real value comes from the first measure, at least in my mind.
Andrew Hart
Minnesota alum
Tower Monarch
Rikku
Posts: 360
Joined: Sun Oct 07, 2007 6:23 pm
Location: Richmond, VA

Re: Individual player stat v. 29192

Post by Tower Monarch »

theMoMA wrote:Underrated in what sense? These stats are only meant to illuminate what happened, not what will happen in the future, so it may be true that on some other team, a Mehdi-like player playing with an Eric-like player might score more points, but that's not really the purpose of this measure.
I meant underrated in the sense that you were using "out of position." I think that taking into account the science overlap, these later stats would make Mehdi seem like he belongs as a 2nd or 3rd, when in reality (taking into account he would be ready to buzz a line or two later if Eric hadn't) he probably is "out of position." Again, this is hardly a criticism since we literally have no data to account for subject effects at this time.
theMoMA wrote: Also, for what it's worth, the forays into "position" stats are just something that I concocted based on the original idea; I don't think they're particularly valuable, just perhaps one interesting way to slice the data (that, as Susan points out, seems to tell us many things that we already know, for whatever that's worth). The real value comes from the first measure, at least in my mind.
Yeah, I definitely still like the first one as a quick (in the sense of spreadsheets) way of measuring contribution with much better accuracy than eyeballing everyone's comparative pts/tossup next to bonus conversions.
Cameron Orth - Freelance Writer/Moderator, PACE member
College: JTCC 2011, Dartmouth College '09-'10, '11-'14
Mathematics, Computer Science and Film/Media Studies
High School: Home Schooled/Cosby High '08-'09, MLWGSGIS A-E '06-'08
User avatar
cvdwightw
Auron
Posts: 3291
Joined: Tue May 13, 2003 12:46 am
Location: Southern CA
Contact:

Re: Individual player stat v. 29192

Post by cvdwightw »

No amount of sophisticated statistics-keeping software is going to be able to determine the Chris Ray Scoring Champion (cf. "Diary of Chris Ray," the player who knows the most things immediately after a teammate's buzz). Sure it would be nice to have more sophisticated data so that things can be broken down, but look, OPS+ is basically canon at this point and it uses essentially the same level of summary statistics we're using here (plus its "algebraic wizardry" includes adding two things together that have no logical reason to be added, and possibly using a correction factor that doesn't really apply to the thing it's correcting).

Here's one thing I noticed about this statistic. The less even a team's scoring is, and the more tossups a team answers, the bigger the discrepancy between the sum of the individual values of the players on a team and the team's ppg. For instance, the Hart/Hoppes/Koo/Teitler team has 307.8 "value added" points (the sum of their individual values is almost 308 points more than their ppg as a team). You'd think this would somehow correlate with a shadow effect, but it's just as poor a measure of the shadow effect as anything else.
Dwight Wynne
socalquizbowl.org
UC Irvine 2008-2013; UCLA 2004-2007; Capistrano Valley High School 2000-2003

"It's a competition, but it's not a sport. On a scale, if football is a 10, then rowing would be a two. One would be Quiz Bowl." --Matt Birk on rowing, SI On Campus, 10/21/03

"If you were my teammate, I would have tossed your ass out the door so fast you'd be emitting Cerenkov radiation, but I'm not classy like Dwight." --Jerry
Magister Ludi
Tidus
Posts: 677
Joined: Thu Dec 01, 2005 1:57 am
Location: Washington DC

Re: Individual player stat v. 29192

Post by Magister Ludi »

I agree with Matt and Susan that these statistical mathematical concoctions don't really tell us much. For me, a new stat is only valuable if it actually changes my perception of a player because I'm certain that there are many players whose value I misconceive. One can construct a stat that shoots out random numerical rankings to prove any player is elite (as Andrew has in fact created a stat that shows he was the best player at CO).

The kind of stat that would actually change my perceptions about players would have to show how dominant players are in a category or how early they buzz. I know nothing about programming, but it seems like it could be possible to keep a crude stat that shows what percentage of questions within a category each player answered over a tournament. The editor would have to mark the category of each tossup, the scorekeepers would have to mark how many questions in each category a player answered at the bottom of the scoresheet, and the stats person could just keep a running tally of how many tossups in each category each player answered. It would be a lot of extra logistical work and would be a bitch for the person inputting stats, but it is the kind of extra information I would be incredibly interested in seeing. I'm sure someone who anything about programming could come up with a more efficient system than my crude tally mark plan for compiling this information.
Ted Gioia - Harvard '12
Editor ACF, PACE
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Magister Ludi wrote:(as Andrew has in fact created a stat that shows he was the best player at CO).
Actually, as I explicitly stated, I did not do this.
Andrew Hart
Minnesota alum
User avatar
Skepticism and Animal Feed
Auron
Posts: 3238
Joined: Sat Oct 30, 2004 11:47 pm
Location: Arlington, VA

Re: Individual player stat v. 29192

Post by Skepticism and Animal Feed »

aceste este Andrew Hart-ul de faina
Bruce
Harvard '10 / UChicago '07 / Roycemore School '04
ACF Member emeritus
My guide to using Wikipedia as a question source
User avatar
theMoMA
Forums Staff: Administrator
Posts: 5993
Joined: Mon Oct 23, 2006 2:00 am

Re: Individual player stat v. 29192

Post by theMoMA »

Perhaps I should explain this again just so there's no confusion. The stats that are shown in the unispaced code text are simply one way of breaking down the data: "lining up" players against opposing players for the purpose of comparison. As should be obvious, the result is going to be completely dependent on how you line the players up against each other. I couldn't think of any other uniform way to do it besides comparing all the #1 scorers to each other, all the #2 scorers to each other, and so on down the line, which is basically supposed to show how well a player stacked up with other players who carried a similar scoring burden (as an aside, I assumed I'd grade out well in this comparison, but I honestly couldn't think of another logical way to do it at the time; later, I thought of weighting the contributions by scoring burden to adjust for the fact that #1 players carry a heavier load than #4 scorers do, though as others have pointed out, there may be better ways to weight the data than my relatively crude measure).

Neither, however, is a pure measure of skill. As we can see, it's extremely dependent on team context; someone like Gautam or me, lucky enough to play with three players who push them to their team's fourth contributor, will do very well because the pool of #4 players is the weakest (I'd also note that Gautam and I were both above-average for #1 scorers, so you can see where that would be a competitive advantage; this is the sort of thing that these numbers can show you, though as Susan pointed out, most of the dramatic conclusions are pretty obvious to anyone who's looked at an SQBS screen before). Also, it's probably safe to say that it underrates someone like Gautam compared to someone like me, because Gautam certainly answered a lot of tossups and bonus parts that the rest of his humanities-centric teammates weren't likely to get, while it's a lot more likely that my points were going to get answered by Seth, Selene, or Jeff anyway. There's no real way to get around this except to have better input data.
Andrew Hart
Minnesota alum
Locked