Maytsh wrote:
No, they can't. They can only choose the game, not the opponents (unless the bad players are clustered together and play on specific-named games).



Sorry, I do not know what "ladder" means.


Indeed, but apart from "camping" and waiting for specific players (something that may favour but does not guarantee consistent wins), how else can a cheater choose opponents for ranking games (where kicking is prohibited even before the start of the game and even for the host)?

I mean how is that "choice" something on which the player has absolute rights on?


I stand corrected.


It means "you have to give A in order to get B", where A can be greater, lower or equal to B.


No, people who do not know how to play poker go broke.


We play online because we like poker (the game, the rules, the emotions).


Agreed. But in order for a bad player to be good, he has to play against the good ones and this employs a certain amount of risk taking capability.

Bad players who have risk aversion do not make good players even if they get paid to play.

I do not understand this explanation. A validity of a good player is shown by how many points he has (on a fair scale) that show how good he is objectively (in the long run). The amount of people that have less points than him tells us things about the competition level.

A ranking of:

{150,50,1,1,1,1)

is way more unreliable than a ranking of:

{150,140,139,124,50,1}


Clearly there are 5 players under the "150" ranking but in the second example the ranking is more robust.

(The above examples are simplified, I could not type an example of 30 rankings )



Agreed. We could (in theory) adapt both models as a test run and then compare the advantages and the weaknesses of them in order to choose the practically most fair.

> They can only choose the game, not the opponents

I suppose it will still be possible to leave the game when a good player joins?

The core of the problem, I think, is that once your rating is x, you have no reason whatsoever to play a game where your chances are below that. Your example used players that had chances of "1" (meaning winning everything), which - I think you agree - is very unrealistic.

So in practice, we will have a very definite point at which good players must stop playing random games in order to improve their score. As the good players will probably be very close together in terms of their chances in random play, this means that finding worse players has a much greater impact on their score than playing well.

Note that this is exactly the opposite of what we would want to see: You can't properly tell good and great players apart by having them play against poor players. I think everybody agrees that your handling of All-In-Newbies doesn't really tell much about how much you know about Poker.

I don't want to trample on your suggestion too much. I wouldn't mind if the devs decide to give it a test run, there's certanly worse options. But I hope you see why I think that it might be the wrong direction?

> Sorry, I do not know what "ladder" means.

Nothing fancy, just the result of the ranking system: Everybody in some kind of order.

> No, people who do not know how to play poker go broke.

Even good players have bad luck. I remember some interviews with pros where they all acknowleged that losing to much money is a serious danger, even for a professional. We seriously don't want that.

> The amount of people that have less points than him tells us things about the competition level.

Well, let's break it down: The fun of a ranking systems is the thought of being on top, with (1) a lot of people (2) wanting to be in your place.

Part 1 is what I addressed, and I feel it is - as trivial as it is - the most important, as it also implies the second part a bit. Just put a ladder of anything up somewhere, and some people will compete. We are just programmed like that.

But some people will start downplaying the importance of the top players. They will start questioning the validity of the whole process and whether the leader has actually "earned" it. And that's the point where a good ranking system comes in.

What do you have to prove at this point? You have to prove that the good players are still playing the same game as everyone, but are just better at it. That's where it starts to get tricky: These players are probably good optimizers, and will use every trick in the book to get to the top. Totally different motivation, but you need to make sure that they still arrive at the same result.

So how do "normal" people play? They would play against whoever they meet. A few games a day, but not a hundred. Seems easy enough? But now try to build a system that encourages exactly this kind of behaviour. Directly rewarding random behaviour and mediocre activity? That's highly nontrivial stuff.

At this point Elo at least gets as far as not actively encouraging the opposite behaviour: It it designed to stay roughly stable with whatever playing patters you use, and activity doesn't matter so much after a certain point.

So there are two important things - competition level and systematic validity. The latter is subject to lots of discussion (as evident) and will get attacked by players non-stop just by the nature of the competition. But in fact, there's little you can actually do here, because solving the underlying problem is mathematically hard.

Bottomline (finally): When you design a system and have to choose between friendliness to the average player and validity of the scores, go with the former. It's worth it.

Maytsh wrote:
That can be avoided by changing the method of arranging ranking games as:

1. Player clicks a button that tells server: "I want to play in rank game"
2. Server puts the player in a "waiting to play for rank" in a queue
3. No players can see who else is in the queue with them.
4. When the queue fills up, server starts a ranking game (i.e. "Ranking Game 001" and assigns the first dealer button to any person in the game.
5. When the game ends, players are presented with an option as to whether they want to play again in ranking game.
6. If player wants to play again, goto step 1, else goto lobby.


So, by eliminating the "camping"/lurking option in ranking as suggested in the above algorithm, would you agree that this kind of "preserving the score" attitude is diminished?


I get your point.



Since Elo ratings are cumbersome in multiplayer games and Glicko (en.wikipedia.org/wiki/Glicko_rating_system) appears to have similar problems, how about Microsoft's TrueSkill algorithm?

en.wikipedia.org/wiki/TrueSkill
research.microsoft.com/en-us/projects/trueskill/faq.aspx



I beg to differ as in "go with both" but that's another story...

I do not know if this is planned, but I think it would be better if not just first places are rewarded. I think every place should get a reward, maybe on a logarithmic basis.

The reason for this is that sometimes you get a bad beat heads-up all-in, maybe after a very long game. And then there is no reward. You may have learned something during the game, but without a reward for the second place it partially feels like wasted time. And even the third place might have been hard work that should have a positive effect on your online ranking in some way. Logarithmic calculations might really do the job.

U are right. It is planned. At the moment the DEV Team of PokerTH is working hard at this feature. ATM we just get first places from pokerth. This is something that will be changed.

> That can be avoided by changing the method of arranging ranking games as: ...

Well, yes, forcing randomization might work. But on the other hand, players might not like this solution because it doesn't allow them to play with their friends or a preferred configuration. On the other hand again, one can also see this as exactly the kind of thing that should be prevented.

For poker, it could actually be feasible. I feel like we're being a bit too theoretical here, so let's put forward a concrete proposal for this kind of ladder:
* min_games somewhere around 1/5th of the typical number of games in the season period
* games_average=0

(And dropping the requirement to have min_games for getting listed - with games_average = 0 that shouldn't be necessary anymore)

Oh yeah, and obviously replace "wins" by some measure where you get points for second place and beyond, like Bock suggested. Maybe like (1, 3/4, 1/2, 1/4, 0) or (1, 1/2, 1/4, 0, 0). The good thing about the bayesian rating is that we don't have to balance anything here, this can well be asymmetrical. Whatever gives the nicest rating distribution in the end.

> would you agree that this kind of "preserving the score" attitude is diminished?

There's still one minor "preserving problem" with this kind of approach: In case a player got lucky and hit a good run of bad players and/or good cards, he might be compelled to stop playing because he knows that he's got worse chances in the coming games. A player in the first place might stop playing. This is mostly a matter of setting min_games high enough, however.

> how about Microsoft's TrueSkill algorithm?

I know it, but I feel like it is too hard for players to "get", even compared to ELO. At some point players get the feeling that you're hiding behind a mountain of numbers in order to escape their criticism. I wouldn't want that. But that's just personal