Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[paddyodoors]

Obi Wan Bonogi, I'm 99% certain no one here holds any ill-will towards you... we have been using your example to spur on a larger discussion encompassing things that have nothing to do with you directly.  No insult or offense is intended.

Assuming you're still reading your intro thread, what are your thoughts on all of this?  I'd like to know...

[paddyodoors]

There are technical solutions to the problem.

For example, you, I, someone could run a ranking system where over-sampled games for a given player are corrected for by decreasing their weight.  Since Colony games count for 49% of Obi Wan's games, and they should only be 16%, then each Colony game updates his rating by only 16/49 of the ordinary update.

http://en.wikipedia.org/wiki/Importance_sampling

That suggestion is so... uncivilized.

[Galzria]

When I look for games personally, the only things I change are "ranked players only", and "point counter prohibited". I think veto mode is as much a cheat as selecting bias, or as declining games until you get a board you like.

The most unbiased model, I think, would be only to count random (no bias) games (not proposed between players, who could game the system) between two opponents whose name and rank are, along with the board, completely hidden until both sides accept. A player resigning would lose the game, while the remaining player earns nothing (yes, it allows win stealing, but takes away unearned win potential). The games only count for wins when ended on piles or province/colony.

[greatexpectations]

I'm with Geronimoo and rrenaud. I don't see a meaningful distinction between banning pre-determined kingdoms but allowing Colony bias as far as a ranking leaderboard goes. Both are very different games compared to random Dominion, just like various methods of outright cheating (Larry, Karumah, etc) mentioned above. I don't really agree with theory that there's a "slight" Prosperity bias either; Obi Wan plays 49% of his games with Colony compared to 16% when playing random cards iirc? That's a very different game indeed, imo, and it doesn't make for reasonable comparisons between different players, which is why it's weird to me to have it included on a leaderboard, much like the Larrys and the Karumahs etc.

when i ran the numbers a week ago it was 58.7% colony games vs a 19.8% average.  that is a hair shy of 3x the public average.  he maintains a 1.32 win rate on colony boards as compared to a 1.27 rate on all boards.

as jonts26 and Fabian pointed out though, the difference between OWB and karumah and larry and all those guys is that OWB is a really good player. part of me is actually proud/jealous of the guy for figuring out a clean way to give himself a bit of a competitive edge. 

i think we would all like a way of playing binding fully random games without seeing the set, but until then we will have these issues.  realistically, veto mode is no better (arguably worse) than deliberately choosing a strategic bias. smart players will know how to take advantage of veto mode to give themselves the best chance of winning.

[DStu]

I'm with Geronimoo and rrenaud. I don't see a meaningful distinction between banning pre-determined kingdoms but allowing Colony bias as far as a ranking leaderboard goes. Both are very different games compared to random Dominion, just like various methods of outright cheating (Larry, Karumah, etc) mentioned above. I don't really agree with theory that there's a "slight" Prosperity bias either; Obi Wan plays 49% of his games with Colony compared to 16% when playing random cards iirc? That's a very different game indeed, imo, and it doesn't make for reasonable comparisons between different players, which is why it's weird to me to have it included on a leaderboard, much like the Larrys and the Karumahs etc.

There are technical solutions to the problem.

For example, you, I, someone could run a ranking system where over-sampled games for a given player are corrected for by decreasing their weight.  Since Colony games count for 49% of Obi Wan's games, and they should only be 16%, then each Colony game updates his rating by only 16/49 of the ordinary update.

http://en.wikipedia.org/wiki/Importance_sampling

First, my opinion on the topic: While I would prefering playing "truely random", and a "truely random" leaderboard would sound like the right thing for the top of it, I think that for the lower regions of the leaderboard some kind of bias should still be possible for games that count for the leaderboard. You want to get matched to people that are "approximately your skill", therefore games have to be rated, but maybe you don't want to start with the full 150 or so cards, but bias to Base or Intrigue.
Also for the mid-regions, maybe you want to get some more games "you like", while getting ranked properly to find the "right" opponents.

Importance sampling could correct the automatch bias, when you know how it works, but it can not correct for manually declining games, because you don't have the data on which games are declined, and even if you had them you would need some kind of model to predict the "probability that a game with cards c1,...,c10 will be declined, given historical decline data.".

[greatexpectations]

Importance sampling could correct the automatch bias, when you know how it works, but it can not correct for manually declining games, because you don't have the data on which games are declined, and even if you had them you would need some kind of model to predict the "probability that a game with cards c1,...,c10 will be declined, given historical decline data.".

if you had the data, you could model it off of the veto mode data dougz posted here. i would guess that games are likely to be declined or vetoed for similar reasons.

[Fabian]

greatexpectations, I'm of quite the opposite opinion. I think knowing how to take advantage of veto mode strategy is a competitive edge (though I'd much prefer if veto mode didn't exist, fwiw), whereas "figuring out" that playing 58% Colony games instead of 19% is good is not a competitive edge, it's playing a different game than everyone else.

[rrenaud]

When I use the built in features of the website to maximize my chances of winning, it's okay.

When you do it, it's not.

[Fabian]

Was that directed at me? Seems a bit harsh? As far as I know, everyone plays with veto mode, and consequently everyone makes choices on which cards to veto. Not everyone plays 58% Colony games though, I'm pretty sure.

[DStu]

if you had the data, you could model it off of the veto mode data dougz posted here. i would guess that games are likely to be declined or vetoed for similar reasons.

OK, with large enough data you might even come around that data and or estimate it given the games actually played. But I don't know if the samle size is large enough, for example if I decline combination (X,Y) against lower level opponents only, I don't think you can tell out of the data out of even 10k games if (X,Y) happend by chance only half as often as the should, or if it is because I declined.

[WanderingWinder]

Was that directed at me? Seems a bit harsh? As far as I know, everyone plays with veto mode, and consequently everyone makes choices on which cards to veto. Not everyone plays 58% Colony games though, I'm pretty sure.

Certainly not everyone plays with veto mode. I understand where rrenaud is coming from, and I also get where you're coming from. I think you're both making valid points.

Not that this doesn't relate to OWB, but at this point, could we get this forked over to isotropic discussion? I'd like to make some more points on the topic, but it does feel a little weird to post them here.

[theory]

Not that this doesn't relate to OWB, but at this point, could we get this forked over to isotropic discussion? I'd like to make some more points on the topic, but it does feel a little weird to post them here.

Done.

[rrenaud]

I did direct it at you.

I personally play with the default, opp can choose veto mode if he wants to.  In veto games I just turbo pick random.

I certainly agree that vetoing cards that cause too much randomness or first player advantage is a lot more subtle and skill-intensive than just knowing that colonies help the better players.  Also, there has been discussion from the high level players about what veto strategies they use.

But they still look the same in the abstract, it's using the features of the site to bias the kingdoms to where you think you have an advantage.

[timchen]

This discussion pops up from time to time... I think I have transited to the phase "unless I can get paid by my level on the leaderboard, I don't care."

But a side note on veto mode:
I have changed my thoughts. The marginal benefit as discussed before is pretty small. The real effect is that veto mode benefits BM-ish player. Killing a card from 11 to kill a potential engine is much easier than killing a money strategy.

[olneyce]

I get where people are coming from in the desire for total randomness.  I just don't fully agree.  Playing with veto mode makes the game more enjoyable for some people.  I used to set it as required for the first month or so it was out, but stopped because I didn't like excluding myself from playing a lot of people.  But when it does show up for games now, I do still enjoy pondering the strategy a bit.

Similarly, I can totally see why people would find Colony games to simply be more enjoyable than non-Colony games. 

Given that, I don't really see much of a problem with people employing those things.  I'd change my tune if someone could demonstrate that the upper-bound on comparative advantage someone could gain from 'exploiting' these loopholes is actually quite large. But given that I can only imagine it bumps players up a couple levels, I just can't get too worked up about.

But then, I also think it would be fine to allow a player to set a couple restraints on cards they simply want to avoid.  I understand that it's not 'true' Dominion to avoid playing with Familiar and Possession.  But those cards are just so incredibly frustrating that I'd rather corrupt the game a tiny bit than have to deal with them.

[greatexpectations]

the desire for total randomness is related to competitive play.  i don't think anyone here has said that biasing or veto mode make the game less fun.

[jonts26]

I think the way the ranking is currently structured is great for 99.9% of players. If the purpose is to give players a general sense of how strong they are and match them up with similarly skilled players, it does a great job. If the purpose is to give a player a sense of improvement as he moves up in rankings, it does a good job. The only place where it becomes problematic is at the very top, where we want this ideal of who's the best. But really, the way the ranking system is structured is too broad anyway for separating players of very close skill. We use the (Mean - 3 STD) formula for level which is kind of silly when you have some people with standard deviations points lower than others just because they play more. I'm not saying we should abide cheaters like Karumah or Paralyzed (and it was probably right to restructure the rankings after Paralyzed), but whether or not one uses biases or veto mode, well what's a few levels among friends?

[Captain_Frisk]

Crazy thought: since it seems that we all agree that colony games vs. province games are different beasts... what if we just just had 2 separate leader boards?

[paddyodoors]

I get where people are coming from in the desire for total randomness.  I just don't fully agree.  Playing with veto mode makes the game more enjoyable for some people.  ...

Similarly, I can totally see why people would find Colony games to simply be more enjoyable than non-Colony games. ...

With respect, olneyce, I don't think this is really addressing the question at hand.  I don't think people are making points regarding whether veto mode or biases should be present and available on isotropic, rather whether games employing these things (which are extraneous to the game itself) should be part of the objective ranking system of the leaderboard.

What is the point of a leaderboard or a ranking system if it isn't our best approximation of skill?  For the purposes of this discussion, the key word here is "best."  How do we, as a community, best approximate each player's skill in Dominion generally.  It is not a question of preferences.  Or at least it shouldn't be if we're talking about something objective (i.e. not relative) like the current leaderboard seems to be attempting.

If OWB (or Paralyzed or Karumah or Fabian or WW or whomever) happens to be the leader at a given time, I think it should mean something.  If a player can make decisions outside of the scope of the game itself (i.e. Dominion by Donald X.) and influence his/her ranking, then it damages the integrity of the honor of being the top player.  It strips the leaderboard of meaning.  Certainly not ALL meaning, but, just as certainly, still strips it of some.

========

I am not a top player.  With all honesty, I will probably never be -- and that's okay.  All of these questions/comments/discussions about "30+", "40+", "how to be at the top", etc. will most likely never apply to me.  For my part, of course I play for fun.  If someone is not here (or on iso) to have fun, I hereby respectfully declare that person, IMHO, an idiot.

And I, personally, don't need the leaderboard to have fun, but it still adds an additional flavor for those of us with a competitive streak (which is cool).  And so, IF the leaderboard is going to be there, I want it to stand for something.  And I'm just a lowly peon -- I can only imagine what others whose names are close to the top think (except when they post here  ).

If someday I ever do make it to the top (don't hold your breath), I want it to be because I was weighed and measured objectively, and found to be the best.

========

The best criticism of my position is this: "who is to say that full-on random is the most objective way to play Dominion?  Nowhere in any of the boxes published by RGG does it define pure Dominion vs. impure Dominion."  And the assertion in this criticism is 100% factual.

... to this I would reply that, by definition, any game which allows the introduction of any sort of player preference outside of the scope of the game itself isn't truly objective. (i.e. preferences and objectivity are mutually exclusive.)

(edit: sorry for the text-wall.  I doubt I make friends when I do this... )

[paddyodoors]

Crazy thought: since it seems that we all agree that colony games vs. province games are different beasts... what if we just just had 2 separate leader boards?

That is, of course, another route that could be taken.  However, I suspect that the distinction between Province games and Colony games is like a difference of species but not of genus.

[Jfrisch]

I think people are forgetting that the primary purpose of the ranking system is to give a rough approximation of skill in order for similarly skilled players to play each other. The ranking system, as is, has larger imperfections when used as a metric to judge true ability. (note that for people who are active and have played 1000+ games, the current subtraction of 3 times the standard deviation is not useful as a measure of skill).

Beyond this, my impression is that people don't (or, perhaps more accurately ,shouldn't) judge the skill of top players solely by ranking but rather by their skill, judgement, and general ability to win. WW is a better player than me because he's more likely to beat me.  (we have an 8/11 record which granted isn't that bad, but still I feel like he outplays me more than the record implies). Rankings have enough inherent flaws that other means (for example, tournaments) are a far better way to determine skill at top levels.

[blueblimp]

I'd like to challenge the implicit assumption that choosing among all cards uniformly at random is the most legitimate way to pick a Dominion kingdom. The rules say "players can choose the 10 Kingdom cards using any method they agree on." Choosing uniformly at random makes some cards really awful, such as Scout and Transmute, which are both better in games dominated by their expansion. It puts just as much weight towards weak cards as strong cards, which favours money strategies.

I'd argue that Colony games are generally agreed to be underrepresented by isotropic's default kingdom generation. In the Dominionstrategy.com tournament, 5 of the 9 pre-designed kingdoms selected for the final had Colonies. Surely this means that the kingdom designers and judges feel that playing about 50% Colony games is actually a good way to determine player skill.

[Thisisnotasmile]

I'd like to challenge the implicit assumption that choosing among all cards uniformly at random is the most legitimate way to pick a Dominion kingdom. The rules say "players can choose the 10 Kingdom cards using any method they agree on." Choosing uniformly at random makes some cards really awful, such as Scout and Transmute, which are both better in games dominated by their expansion. It puts just as much weight towards weak cards as strong cards, which favours money strategies.

I didn't really want to have any input into this discussion because I thought the way it started was pretty distasteful, but I just want to say that this is 100% correct, and the key words are "any method they agree on". If I agree to play a game chosen at random from one copy of every card in the game and you put 3 more copies of a handfull of cards in the the randomiser without telling me, the set has not been picked legitimately because I did not agree to those cards being added.

[Robz888]

I don't fault Obi Wan for figuring out a way to play more of the type of games that he enjoys, that he is best at, while still having them count for the leaderboard. The problem is, now that we all know he is doing that, we sort of have the incentive to do the same thing for whichever sets we are best at (Council Room has this information, if you don't know.) For instance, I guarantee that I could bias Intrigue and Hinterlands and give myself a competitive advantage. And other players can do the same for Alchemy or Cornucopia or whatever cards they are awesome with.

So if he's doing it, we should all do it, and if we all do it, ugh. So I would support not counting biased games for the leaderboard in the future.

[olneyce]

Rankings have enough inherent flaws that other means (for example, tournaments) are a far better way to determine skill at top levels.

Hmm. Your ideas are intriguing to me and I wish to subscribe to your newsletter.

But seriously, I understand that people are concerned about the desire for rankings to reflect objective skill.  I'm just saying that these things (veto, biasing) insert a very small amount of noise into an intrinsically very noisy metric.  And since they increase enjoyment for a fair number of people, it doesn't seem worth worrying about very much.

The alternative, where the games stop counting for rankings creates a forced choice.  Players who enjoy biasing or vetoing but ALSO enjoy having a metric that assesses their rank are now obligated to put those values into conflict.

If this were something that really MATTERED, I would choose fairness over enjoyment for sure.  But it's just a game, and it's supposed to be fun.  Making it less fun for a lot of players so that the rankings are marginally more accurate just doesn't seem like a worthwhile trade to me.  It may very well seem worthwhile to others, and that's perfectly reasonable.  I just don't think the majority (or anything close to it) would share that belief.

That said, if the change were implemented, I don't think it would be catastrophic. 

And I tend to agree that if you ARE playing random, you shouldn't be able to see the kingdom.  Random ought to actually mean random.  But even there I'm not overly worried about the damage it does to the objectivity of rankings so much as I am annoyed at having people take forever to ponder kingdoms while deciding whether to accept on auto-match.