Dominion Strategy Forum

Please login or register.

Login with username, password and session length
Pages: 1 2 [3]  All

Author Topic: Ranking, Seeding, and Skill Discussion  (Read 24256 times)

0 Members and 1 Guest are viewing this topic.

dondon151

  • 2012 US Champion
  • *
  • Offline Offline
  • Posts: 2522
  • Respect: +1856
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #50 on: December 10, 2012, 02:48:33 am »
+1

I can see the logic behind all of these rules individually and I understand why you want keep the 'name' players in the tournament, but there comes a point where this gets outweighed by the perception that the lower ranked players are not really supposed to win.

Are the lower ranked players supposed to win a skill-based tournament? Sure beats me.
Logged

ycz6

  • Minion
  • *****
  • Offline Offline
  • Posts: 676
  • Respect: +412
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #51 on: December 10, 2012, 03:23:39 am »
0

I can see the logic behind all of these rules individually and I understand why you want keep the 'name' players in the tournament, but there comes a point where this gets outweighed by the perception that the lower ranked players are not really supposed to win.

Are the lower ranked players supposed to win a skill-based tournament?
If they're more skilled than their opponents, then yes.
Logged

dondon151

  • 2012 US Champion
  • *
  • Offline Offline
  • Posts: 2522
  • Respect: +1856
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #52 on: December 10, 2012, 03:35:37 am »
+2

If they're more skilled than their opponents, then yes.

Piemaster's perceived disadvantages (save for the first player seat one) only hold under the assumption that lower ranked players are less skilled.
Logged

Qvist

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2400
  • Shuffle iT Username: Qvist
  • Respect: +4085
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #53 on: December 10, 2012, 06:32:38 am »
0

I want to say theory that I mean no ill will and I 100% get the intention of both the wildcard and the seeding by iso. I take no issue with either.

I'm simply expressing that while I think a player ranked highly on isotropic is undeniably a good player I do not think a player ranked low on iso implies a bad player as this forum tends to imply. When I played my match this week I was level 3 on iso. I like iso but its never been how I want to play the game. Goko is closer even with all their bugs. And I'm ranked highly there.

All I'm saying is there is always exceptions. I do not think that high level players were more likely to miss sign ups as I think a majority of them are here frequently. And if I joined the forum and was bumped from the tournament based on my iso rating I would not think this community is as welcoming as I know it to be.

I agree with many here that iso ranking is still the best we have.
Let's say a pretty good Tennis player who only played privately amongst friends attends one tournament, wins it and has therefore the possibility to attend the World Championships (I have no idea if this is possible, but let's just assume). His rating is still low because of only playing one tournament and will be seeded low. This is probably not fair, but the best thing to do. There is no reasonable other way to seed the games in a single elimation bracket. You either have to choose a different system, like Swiss or Round Robin or a mixed form of all these. I generally don't like single elimination, but we have seeding which makes it at least way better than random. I think future tournaments should be double elimination or some mixed form of Round Robin and single elimination (Group System) like in Soccer World Championships.

WanderingWinder

  • Adventurer
  • ******
  • Offline Offline
  • Posts: 5275
  • ...doesn't really matter to me
  • Respect: +4381
    • View Profile
    • WanderingWinder YouTube Page
Re: Ranking, Seeding, and Skill Discussion
« Reply #54 on: December 10, 2012, 08:13:04 am »
+2

I want to say theory that I mean no ill will and I 100% get the intention of both the wildcard and the seeding by iso. I take no issue with either.

I'm simply expressing that while I think a player ranked highly on isotropic is undeniably a good player I do not think a player ranked low on iso implies a bad player as this forum tends to imply. When I played my match this week I was level 3 on iso. I like iso but its never been how I want to play the game. Goko is closer even with all their bugs. And I'm ranked highly there.

All I'm saying is there is always exceptions. I do not think that high level players were more likely to miss sign ups as I think a majority of them are here frequently. And if I joined the forum and was bumped from the tournament based on my iso rating I would not think this community is as welcoming as I know it to be.

I agree with many here that iso ranking is still the best we have.
Let's say a pretty good Tennis player who only played privately amongst friends attends one tournament, wins it and has therefore the possibility to attend the World Championships (I have no idea if this is possible, but let's just assume). His rating is still low because of only playing one tournament and will be seeded low. This is probably not fair, but the best thing to do. There is no reasonable other way to seed the games in a single elimation bracket. You either have to choose a different system, like Swiss or Round Robin or a mixed form of all these. I generally don't like single elimination, but we have seeding which makes it at least way better than random. I think future tournaments should be double elimination or some mixed form of Round Robin and single elimination (Group System) like in Soccer World Championships.
This would be nice, but it would take FOREVER.

Stealth Tomato

  • Minion
  • *****
  • Offline Offline
  • Posts: 507
  • Dorkneel
  • Respect: +480
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #55 on: December 10, 2012, 09:49:17 am »
+3

It probably doesn't help that on top of the 'reserved slots' the tournament seems a bit slanted towards the top iso players in other ways too.

1.  Non-random draw designed to give them easy early round matchups.
2.  First turn advantage in their first few rounds at least
3.  Non-random shuffling so a lower ranked player is less likely to 'get lucky' against them

I can see the logic behind all of these rules individually and I understand why you want keep the 'name' players in the tournament, but there comes a point where this gets outweighed by the perception that the lower ranked players are not really supposed to win.  Would it be really be so bad if Stef played WW in round 2?  Or if little Timmy took a game off Marin through opening 5/2 on a Mountebank board?

This isn't really a complaint as such, but just stuff to think about to make the tournament even better next year.

This is how seeding works. Think of normal Iso as the regular season, and this as the playoffs. Better teams in the regular season get better matchups and home-field advantage. In the end, home field is the only edge these higher-ranked players are getting, and assuming a FTA of .6, they're only getting a 54%-46% edge.

It also builds excitement for the later rounds, where the dominant players start facing each other, and the upset artists start becoming a big deal. A low-ranked player who breezes through easy matchups and then gets trounced in Round 4 by Robz isn't interesting. One who upsets a #1 seed and then gets trounced in Round 4 by Robz is.
Logged

Piemaster

  • Apprentice
  • *****
  • Offline Offline
  • Posts: 260
  • Respect: +170
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #56 on: December 10, 2012, 10:14:09 am »
0

This is how seeding works.

I know how seeding words, I just question whether seeding really adds that much to the tournament.  I can see why they do it in professional sports, because TV revenue will be damaged by good players/teams getting eliminated early in the tournament, plus they need to incentivise the better players to actually show up for the minor events (or, in your football example, ensure the regular season actually means something).  But in grass roots sports you don't get a whole lot of seeding and there is a reason for that too.

I think the fear is that the top seeds will all eliminate each other in the early rounds and we'll end up with a damp squib final with one great player and one relative nobody.  But I think the odds of that happening are extremely slim.  In fact I don't know if your tournament software can do this, but I would suggest running a few simulations with an entirely random draw, assigning nominal win percentages based on difference in rankings and then see what happens.  I wouldn't mind betting that the final eight of the tournament is nearly always eight very strong players.
« Last Edit: December 10, 2012, 10:18:55 am by Piemaster »
Logged

theory

  • Administrator
  • *****
  • Offline Offline
  • Posts: 3603
  • Respect: +6121
    • View Profile
    • Dominion Strategy
Re: Ranking, Seeding, and Skill Discussion
« Reply #57 on: December 10, 2012, 10:26:40 am »
+3

I don't think the reason for seeding is as cynical as that.  For example, can you point me to any major bracket tournament of any kind that doesn't seed even though it can?  Video game tournaments seed.  Fantasy football leagues seed.  Internet book competitions seed.

The other reason for wanting seeding is that it enhances the legitimacy of the tournament.  You can point to each division winner as having earned it, instead of just stomping on weak competition the whole way through.

There's also no independent verifiability of random bracket creation.  Suppose I randomly generate a bracket that happens to put everyone good in the same division.  Do I hit "reshuffle"?  Regardless of whether I do or don't, someone can accuse me of reshuffling.

Finally, what's the purpose of not seeding?  To give lower-ranked players a fair shot?  Why don't we also give higher-ranked players a fair shot?  Let someone in the 30's get to where they are supposed to be, instead of having to face a top seed in the first round and never get anywhere.
Logged

Insomniac

  • Jester
  • *****
  • Offline Offline
  • Posts: 785
  • Respect: +392
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #58 on: December 10, 2012, 11:00:53 am »
0

I don't know how much of this is from me or not but I think seeding is fine and using iso for that doesn't bug me I think the brackets are fine. I was only ever mentioning taking exception to exclusion based on iso
Logged
"It is one of [Insomniacs] badges of pride that he will bus anyone, at any time, and he has done it over and over on day 1. I am completely serious, it is like the biggest part of his meta." - Dsell

dondon151

  • 2012 US Champion
  • *
  • Offline Offline
  • Posts: 2522
  • Respect: +1856
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #59 on: December 10, 2012, 11:01:03 am »
+2

I can see why they do it in professional sports, because TV revenue will be damaged by good players/teams getting eliminated early in the tournament, plus they need to incentivise the better players to actually show up for the minor events (or, in your football example, ensure the regular season actually means something).

Oh, you can't possibly know the clandestine ways theory generates revenue from this site... He's only a lawyer on the surface...
Logged

Stealth Tomato

  • Minion
  • *****
  • Offline Offline
  • Posts: 507
  • Dorkneel
  • Respect: +480
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #60 on: December 10, 2012, 11:03:02 am »
+1

This is how seeding works.

I know how seeding words, I just question whether seeding really adds that much to the tournament.  I can see why they do it in professional sports, because TV revenue will be damaged by good players/teams getting eliminated early in the tournament, plus they need to incentivise the better players to actually show up for the minor events (or, in your football example, ensure the regular season actually means something).  But in grass roots sports you don't get a whole lot of seeding and there is a reason for that too.

What on God's green earth are you talking about? I have never played in a league of anything that didn't seed the teams, and I've played rec-level roller hockey, flag football, and softball (plus fantasy football), both in college and post-college, in multiple cities. It is the single most fair-to-the-players way to seed a tournament, because nothing would suck quite so much as being the second-best team in the league and getting bounced in the first round because random draw put you up against the best.
Logged

theory

  • Administrator
  • *****
  • Offline Offline
  • Posts: 3603
  • Respect: +6121
    • View Profile
    • Dominion Strategy
Re: Ranking, Seeding, and Skill Discussion
« Reply #61 on: December 10, 2012, 11:09:09 am »
+4

I can see why they do it in professional sports, because TV revenue will be damaged by good players/teams getting eliminated early in the tournament, plus they need to incentivise the better players to actually show up for the minor events (or, in your football example, ensure the regular season actually means something).

Oh, you can't possibly know the clandestine ways theory generates revenue from this site... He's only a lawyer on the surface...

Did I forget to mention that the final is going to be televised live on ESPN?  (The semifinals only get ESPN2 though.)

With all the TV money I plan on signing Zack Greinke to play Dominion next year.
Logged

dondon151

  • 2012 US Champion
  • *
  • Offline Offline
  • Posts: 2522
  • Respect: +1856
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #62 on: December 10, 2012, 11:09:53 am »
0

Which round gets ESPN8?
Logged

Piemaster

  • Apprentice
  • *****
  • Offline Offline
  • Posts: 260
  • Respect: +170
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #63 on: December 10, 2012, 11:24:21 am »
0

I don't think the reason for seeding is as cynical as that.  For example, can you point me to any major bracket tournament of any kind that doesn't seed even though it can? 

I guess soccer 'cup tournaments' would be the most obvious example.

Anyway I did write a big long reply to this but I deleted it because, quite frankly, I'm tired of getting dragged into long arguments in this thread on issues that, when all's said and done, I don't really care about all that much.  I do understand all the merits of seeding, I said as much in my original post.  I just think it has it's pros and cons and isn't strictly 'better' than a random draw.  I'm just gonna leave it at that.
Logged

Schneau

  • Saboteur
  • *****
  • Offline Offline
  • Posts: 1174
  • Shuffle iT Username: Schneau
  • Respect: +1461
    • View Profile
    • Rainwave
Re: Ranking, Seeding, and Skill Discussion
« Reply #64 on: December 10, 2012, 11:27:23 am »
+2

This is how seeding works.

I know how seeding words, I just question whether seeding really adds that much to the tournament.  I can see why they do it in professional sports, because TV revenue will be damaged by good players/teams getting eliminated early in the tournament, plus they need to incentivise the better players to actually show up for the minor events (or, in your football example, ensure the regular season actually means something).  But in grass roots sports you don't get a whole lot of seeding and there is a reason for that too.

What on God's green earth are you talking about? I have never played in a league of anything that didn't seed the teams, and I've played rec-level roller hockey, flag football, and softball (plus fantasy football), both in college and post-college, in multiple cities. It is the single most fair-to-the-players way to seed a tournament, because nothing would suck quite so much as being the second-best team in the league and getting bounced in the first round because random draw put you up against the best.

Heck, my local, very low competition curling league seeds our playoffs. Curling. The one where you push rocks on ice and sweep in front of them.
Logged

theory

  • Administrator
  • *****
  • Offline Offline
  • Posts: 3603
  • Respect: +6121
    • View Profile
    • Dominion Strategy
Re: Ranking, Seeding, and Skill Discussion
« Reply #65 on: December 10, 2012, 11:29:37 am »
+1

Which round gets ESPN8?

The ranking/seeding/skill debates, of course!
Logged

mith

  • Jester
  • *****
  • Offline Offline
  • Posts: 771
  • Shuffle iT Username: mith
  • Respect: +778
    • View Profile
    • MafiaScum.net
Re: Ranking, Seeding, and Skill Discussion
« Reply #66 on: December 10, 2012, 11:31:48 am »
+1

Even in soccer, there is some measure of seeding - with groups, teams are seeded into pools so that the groups have one team from each; with tournaments like the FA Cup, different tiers come in at later rounds of the tournament. The analogy here wouldn't be "pair up all 256 players totally at random", but more like "give everyone level 35+ three byes, and pair them up at random with anyone who survived that long".
Logged

greatexpectations

  • Board Moderator
  • *
  • Offline Offline
  • Posts: 1097
  • Respect: +1067
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #67 on: December 10, 2012, 11:58:52 am »
+4

what this all boils down to in the end is that this is probably the single largest dominion tournament in the world. and it has a prize for the winner. and it features many of the top players in the world. and it costs nothing to enter. and it is played via a medium which costs nothing to anyone with an internet connection. and it is being played with every expansion except one. and it is entirely DIY and community organized, actively taking input on how best to try and run it.

feel free to compare this to any other dominion tournament out there. i won a small tournament last year that A. cost money to enter B. featured only the base set C. had a structure biased heavily against previous round winners and D. had rules made up as it went along. no joke, the tournament organizer was going to use deck size (larger>smaller) as a tiebreaker.

or maybe look at the world masters tournament. some don't like the bias for isotropic players here, but how about the location bias for that tournament?  the nearest qualifier to me that i was aware of was about 550 miles away. each feeder tournament sort of made up its own rules as it went along. and the final rounds featured a totally random player just to fill up a spot.

i'm all for trying to opitimize the tournament, but let's all just keep some perspective. and in the end, most of this won't matter anyway, as isotropic probably won't be around for us to even use a year from now.
Logged
momomoto: ...I looked at the tableau and went "Mountebank? That's for jerks."
rrenaud: Jerks win.

Insomniac

  • Jester
  • *****
  • Offline Offline
  • Posts: 785
  • Respect: +392
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #68 on: December 10, 2012, 12:27:11 pm »
+2

I want to say on the record again that I'd like to thank the people running this tournament, it is still hands down one of the best tournaments, I meant no ill will to anyone with my statements was merely talking about my opinion.
Logged
"It is one of [Insomniacs] badges of pride that he will bus anyone, at any time, and he has done it over and over on day 1. I am completely serious, it is like the biggest part of his meta." - Dsell

Piemaster

  • Apprentice
  • *****
  • Offline Offline
  • Posts: 260
  • Respect: +170
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #69 on: December 10, 2012, 12:29:53 pm »
0

I want to say on the record again that I'd like to thank the people running this tournament, it is still hands down one of the best tournaments, I meant no ill will to anyone with my statements was merely talking about my opinion.

Ditto for me. 

I'm one of those people who likes to question things, and say "what if we tried this, or did it like this instead?"  I hope nobody took it to mean I am dissatisfied with the tournament.
Logged

greatexpectations

  • Board Moderator
  • *
  • Offline Offline
  • Posts: 1097
  • Respect: +1067
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #70 on: December 10, 2012, 12:31:01 pm »
0

I want to say on the record again that I'd like to thank the people running this tournament, it is still hands down one of the best tournaments, I meant no ill will to anyone with my statements was merely talking about my opinion.

no worries, i'm not trying to hate on you or anyone else. i'm just pointing out that despite the shortcomings we are still a couple steps ahead of the alternatives.
Logged
momomoto: ...I looked at the tableau and went "Mountebank? That's for jerks."
rrenaud: Jerks win.

soulnet

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2142
  • Respect: +1751
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #71 on: December 10, 2012, 02:02:23 pm »
+1

I do understand all the merits of seeding, I said as much in my original post.  I just think it has it's pros and cons and isn't strictly 'better' than a random draw.  I'm just gonna leave it at that.

You mean they can be worth the same in coins? Or would you just use Potion based cost for one of them so they are incomparable?
Logged

Beyond Awesome

  • Global Moderator
  • *****
  • Offline Offline
  • Posts: 2941
  • Shuffle iT Username: Beyond Awesome
  • Respect: +2466
    • View Profile
Re: Ranking, Seeding, and Skill Discussion
« Reply #72 on: December 10, 2012, 07:33:03 pm »
+1

what this all boils down to in the end is that this is probably the single largest dominion tournament in the world. and it has a prize for the winner. and it features many of the top players in the world. and it costs nothing to enter. and it is played via a medium which costs nothing to anyone with an internet connection. and it is being played with every expansion except one. and it is entirely DIY and community organized, actively taking input on how best to try and run it.

feel free to compare this to any other dominion tournament out there. i won a small tournament last year that A. cost money to enter B. featured only the base set C. had a structure biased heavily against previous round winners and D. had rules made up as it went along. no joke, the tournament organizer was going to use deck size (larger>smaller) as a tiebreaker.

or maybe look at the world masters tournament. some don't like the bias for isotropic players here, but how about the location bias for that tournament?  the nearest qualifier to me that i was aware of was about 550 miles away. each feeder tournament sort of made up its own rules as it went along. and the final rounds featured a totally random player just to fill up a spot.

i'm all for trying to opitimize the tournament, but let's all just keep some perspective. and in the end, most of this won't matter anyway, as isotropic probably won't be around for us to even use a year from now.

I cringed when I read about the world masters tournament and also the lack of tournaments nearby to even qualify for the world masters tournament really sucked. Man, what we have here is a million times better than any other Dominion tournament out there.
Logged

SirPeebles

  • Cartographer
  • *****
  • Offline Offline
  • Posts: 3249
  • Respect: +5459
    • View Profile
Re: Re: Upsets
« Reply #73 on: December 10, 2012, 09:41:04 pm »
0

Ah, thanks!  I knew about the levels and uncertainty; I just didn't realize it was also called "mean skill".

But now I'm confused why WW puts 3 pleases before asking you to use it instead of level.  :)  Is it vastly preferred in some circles?

i am going to oversimplify here, but basically mean skill is a more accurate measure for skill for people on the top end of the leaderboard (or who have a ton of games played) and the isotropic level is a better measure for less skilled or newer players.
That's patently false. The mean skill is the best guess you have for the skill of anyone. BY DEFINITION.

You could say it's the most accurate guess.  To say that it's the best guess really depends on what you're trying to accomplish with your guess.
What are you trying to accomplish with your guess?

Hey WanderingWinder, I missed the response earlier, sorry about that.

First a disclaimer.  I'm an algebraist by trade, not a statistician.  I also don't know the details of how the True Skill system works.  My guess is that it's effectively the following.  Your "skill" is considered probabilistically.  That is, in the most encompassing sense, the best and most accurate "guess" of your skill is really a probability distribution.  You can't exactly totally order probability distributions in a meaningful way, and so these distributions are distilled to a mean and a standard deviation.  Your mean skill in the mean, and your level is the mean minus some multiple of the standard deviation perhaps.

Now, if the probability distribution which best describes your skill is normal, then sure, mean skill is the best single number estimate in a reasonable sense.  However, if your distribution is bimodal, for instance, then the mean skill may extremely unlikely to be your actual skill, but rather is roughly the average of two other skills which are likely.  As an example from the recent US election, Nate Silver presented his prediction of the electoral vote split via a probability distribution.  The two most likely outcomes, by his model, were that Obama would win 304 or 332 electoral votes (I may be slightly off, by this principle I'm illustrating is the same).  The mean of this distribution?  It was about 314 if I recall correctly, but his model gave a VERY low probability of Obama getting precisely 314 electoral votes.  This illustrates how the mean is often a terrible guess in and of itself.

Now, if you have a reason to believe that the underlying distribution is normal, then the mean skill would be a fair choice for best guess.  But it certainly isn't by definition, as illustrated above.  Now, I imagine that True Skill probably assigns a new user a default distribution, and then updates the distribution in a Bayesian manner each night based on the results of the day.  Maybe True Skill only assigns normal distributions?  In this case, I suppose by a quirk of the system, yes the mean skill would always be the "best" guess.
Logged
Well you *do* need a signature...

WanderingWinder

  • Adventurer
  • ******
  • Offline Offline
  • Posts: 5275
  • ...doesn't really matter to me
  • Respect: +4381
    • View Profile
    • WanderingWinder YouTube Page
Re: Ranking, Seeding, and Skill Discussion
« Reply #74 on: December 10, 2012, 10:05:49 pm »
0

Ah, thanks!  I knew about the levels and uncertainty; I just didn't realize it was also called "mean skill".

But now I'm confused why WW puts 3 pleases before asking you to use it instead of level.  :)  Is it vastly preferred in some circles?

i am going to oversimplify here, but basically mean skill is a more accurate measure for skill for people on the top end of the leaderboard (or who have a ton of games played) and the isotropic level is a better measure for less skilled or newer players.
That's patently false. The mean skill is the best guess you have for the skill of anyone. BY DEFINITION.

You could say it's the most accurate guess.  To say that it's the best guess really depends on what you're trying to accomplish with your guess.
What are you trying to accomplish with your guess?

Hey WanderingWinder, I missed the response earlier, sorry about that.

First a disclaimer.  I'm an algebraist by trade, not a statistician.  I also don't know the details of how the True Skill system works.  My guess is that it's effectively the following.  Your "skill" is considered probabilistically.  That is, in the most encompassing sense, the best and most accurate "guess" of your skill is really a probability distribution.  You can't exactly totally order probability distributions in a meaningful way, and so these distributions are distilled to a mean and a standard deviation.  Your mean skill in the mean, and your level is the mean minus some multiple of the standard deviation perhaps.

Now, if the probability distribution which best describes your skill is normal, then sure, mean skill is the best single number estimate in a reasonable sense.  However, if your distribution is bimodal, for instance, then the mean skill may extremely unlikely to be your actual skill, but rather is roughly the average of two other skills which are likely.  As an example from the recent US election, Nate Silver presented his prediction of the electoral vote split via a probability distribution.  The two most likely outcomes, by his model, were that Obama would win 304 or 332 electoral votes (I may be slightly off, by this principle I'm illustrating is the same).  The mean of this distribution?  It was about 314 if I recall correctly, but his model gave a VERY low probability of Obama getting precisely 314 electoral votes.  This illustrates how the mean is often a terrible guess in and of itself.

Now, if you have a reason to believe that the underlying distribution is normal, then the mean skill would be a fair choice for best guess.  But it certainly isn't by definition, as illustrated above.  Now, I imagine that True Skill probably assigns a new user a default distribution, and then updates the distribution in a Bayesian manner each night based on the results of the day.  Maybe True Skill only assigns normal distributions?  In this case, I suppose by a quirk of the system, yes the mean skill would always be the "best" guess.
What I mean by the definition is that it is the 'expected value', which is of course not the most probable value. Further, TrueSkill (quite wrongly, I believe) assigns a normal distribution to everyone's skill. But you don't need to have a normal distribution to get mean = mode (=median) = any unimodal symmetric distribution will do. And I think this is quite reasonable prima facie for a distribution of skill, which is quite different than something like projecting electoral votes (though you probably don't want to open the Nate Silver can of worms here). But the thing there is that there are big discrete units that fly around with different states, whereas skill is something which is essentially continuous, insofar as it's measurable. Which makes an enormous difference. Anyway, the mode=mean is not at all a quirk of the system, but a design FEATURE of the system, and one that I believe is essentially correct.
Pages: 1 2 [3]  All
 

Page created in 0.064 seconds with 21 queries.