Obvious Blacklist Abuse

[Sharajat]

So an obvious abuse of every blacklist system is to just blacklist top players you lose to to avoid them.  Should there be a cap on the number of players that can be blacklisted, or will the admins watch for accounts that have blacklisted high level players?

[Jacob Marley]

This would only be a problem insomuch as it allows people to game the leaderboard, by trying to get easier wins.  Otherwise, it is only limiting the pool of players they can play against.  The problem with gaming the leaderboard, is that it should be hard to climb the leaderboard if all you are playing are low rated players.  I don't know enough about Glicko 2 to tell how much of a problem there is.  At this point I don't think there need to be limits, but if time goes on and it seems that there are a lot of players high in the leaderboard who are only playing low ranked players, then maybe revisit the question.

[SkyHard]

And here I thought it would be easier to climb the leaderboard playing against strong players. Why? A single win has a great impact whereas a loss not that much.

On the other hand, loosing against a low ranked players will have a big impact (and with some bad luck it is always possible to lose - my opinion as a non pro)

[tufftaeh]

With "Equal Opponent" you are already guaranteed not to be matched with a player more than 12.5 levels above you. And if you create a table, you can limit the level difference even more (and even allow only much weaker opponents). So there is no need at all to use the blacklist for this. Anyway, as others have already pointed out: You don't get too many ranking points for wins against much weaker players...

[JW]

If the rankings were badly mis-estimating the win probabilities for players with substantially different rankings, there might be a way to game the rankings system as you describe. But if you never take the risk of playing against anyone better than you, this will stifle your ability to improve. 

[gitsticker8]

Of course the blacklist can be abused in this way and in others as well.  This is why most popular online multiplayer games don't allow it, but offer instead /report and /ignore features.  I understand people want a blacklist so they can avoid trolls, but the problem isn't the trolls.  The problem is the game design that allows trolls to be so effective with slow-rolling and no option for ignoring their chat. 

This is off topic, but if less time were allowed per turn (5 minutes +10 seconds per action taken for example) slow rolling would be less of an issue and the calls for a blacklist would diminish.

This of course is only relevant if the goal is to actually have a meaningful leader board which I assume is the case.

[JW]

Anyone who argues the blacklist should be removed because of the potential to abuse it to alter ratings should also provide a description of why the ratings system is doing a poor job of predicting the outcome of certain games. If that were to be fixed, the potential for abuse would go away. I'm much more interested in playing against opponents I enjoy playing with than with the precise accuracy of the leaderboard, and I assume many users are the same way.

[gitsticker8]

why the ratings system is doing a poor job of predicting the outcome of certain games.

Doesn't this have something to do with the game design itself?  It's quite likely that a lower rated player can win just by luck alone (go first, draw 5/2 with witch chapel on the board for example).

If that were to be fixed, the potential for abuse would go away.

I don't see how improving the rating system's performance would eliminate blacklist abuse unless I'm missing something.  Could you elaborate?

I'm much more interested in playing against opponents I enjoy playing with than with the precise accuracy of the leaderboard, and I assume many users are the same way.

I think you make a very valid point here. What's the point of playing?  To have fun.  But what is the point of a leaderboard?  To provide accurate information that allows players to see how they stack up against others.  If the information isn't accurate (ie open to abuse) then I don't see a point of having a leaderboard.  Inevitably there will be people who rise to the top by abusing the system which discourages people from trying to be competitive in the first place. 

I do think both concerns can be addressed however.  If the ability to troll people were mitigated to a greater extent, we wouldn't need the blacklist. 

[JW]

why the ratings system is doing a poor job of predicting the outcome of certain games.

Doesn't this have something to do with the game design itself?  It's quite likely that a lower rated player can win just by luck alone (go first, draw 5/2 with witch chapel on the board for example).

The rating system takes into account that the lower rated player is expected to win a certain percentage of games.

If that were to be fixed, the potential for abuse would go away.

I don't see how improving the rating system's performance would eliminate blacklist abuse unless I'm missing something.  Could you elaborate?

Whether you rise or fall in the ratings depends on how you do compared to the rating system's expectations.

Note that you don't rise in level merely for winning, you have to perform above the rating's system's expectation.

Using purely hypothetical examples, if the rating system thinks that a player 15 levels higher than another one wins, say, 95% of the time, when the higher rated player actually wins 85% of the time, then the higher rated player would have reasons to avoid that matchup. If the rating system thinks that a player 15 levels higher than another one wins 85% of the time, when the higher rated player actually wins 95% of the time, then the lower rated player would have reasons to avoid that matchup. If the system is basically accurate, then no one has reason to avoid the matchup.

[gitsticker8]

Thanks for that explanation, well said.  You're arguing that a perfect matching system with rewards perfectly proportional to the risk of losing is impervious to abuse.  And I agree completely based on your explanation.  I would also argue that the more control you give the player as to who they play against, the less meaningful the leaderboard becomes in general.  Blacklisting effectively limits the player pool.  If my player pool has 10 players and yours has 10 completely different players, should we both be rated on the same leaderboard?  I think that would be a good case for 2 separate leaderboards.  Granted this will likely never happen, but the more blacklisting that goes on, the closer we approach that scenario.

What about the case of a person who blacklists everyone who takes longer than 1 minute for their first turn?  What if it becomes common practice to share blacklists thus alienating players and reducing their opponent pool unbeknownst to them?

One more point about the rematch option since I think it is relevant to this conversation.  I believe rematch games should NOT be rated.  It enables win-trading (intentionally losing to your friend over and over to boost their rating) which is just another way to game the leaderboard.  The only games that should be rated are those that the matching system sets up.

[JW]

One more point about the rematch option since I think it is relevant to this conversation.  I believe rematch games should NOT be rated.  It enables win-trading (intentionally losing to your friend over and over to boost their rating) which is just another way to game the leaderboard.  The only games that should be rated are those that the matching system sets up.

The logic here implies that games also should not be rated if you set them up directly with friends. I would expect that the percentage of games set up directly with friends or rematch games that are legitimate (versus intentionally thrown to game the rating system) is greater than 99.9%.

ShuffleIT seems to have hit on the correct solution to this potential issue: "While it's possible to beat up dummy accounts in order to game the leaderboard, this is not a good idea if you're interested in the longevity of your account." http://forum.shuffleit.nl/index.php?topic=1679.msg5888#msg5888

I would also argue that the more control you give the player as to who they play against, the less meaningful the leaderboard becomes in general.  Blacklisting effectively limits the player pool.  If my player pool has 10 players and yours has 10 completely different players, should we both be rated on the same leaderboard?  I think that would be a good case for 2 separate leaderboards.  Granted this will likely never happen, but the more blacklisting that goes on, the closer we approach that scenario.

Players are already highly segregated based on what times they play (mainly due to different time zones). I expect this far outweighs any effects of the blacklist. And I don't think that people playing Dominion all over the world is a concern for the leaderboard, so this doesn't seem like a concern either.

[gitsticker8]

The logic here implies that games also should not be rated if you set them up directly with friends. I would expect that the percentage of games set up directly with friends or rematch games that are legitimate (versus intentionally thrown to game the rating system) is greater than 99.9%.

Yes, I believe custom games should not be rated either.  Only games that are set up by the matching system.  While I agree that .1% is not a big number of people abusing the system, it doesn't take many rotten apples to spoil the barrel because those are the names you will always see at the top of the board.  Again, the more power the players have to define their opponent pool, the less meaningful the boards become.

ShuffleIT seems to have hit on the correct solution to this potential issue: "While it's possible to beat up dummy accounts in order to game the leaderboard, this is not a good idea if you're interested in the longevity of your account." http://forum.shuffleit.nl/index.php?topic=1679.msg5888#msg5888

This is indeed the correct response, I agree.  I just wonder if it is all teeth and no bite? There's not even an option to report, how can we expect account bans when there is no /report functionality? Have there been any account bans yet?

Players are already highly segregated based on what times they play (mainly due to different time zones). I expect this far outweighs any effects of the blacklist. And I don't think that people playing Dominion all over the world is a concern for the leaderboard, so this doesn't seem like a concern either.

Good point, and something I forgot to bring up.  Separate leader boards for different regions/timezones would be ideal. I'm just not sure if the playerbase is large enough to support it.  So yes, people playing all over the world is a concern (of mine) for the leaderboard.

[tufftaeh]

Games for the Dominion League (http://forum.dominionstrategy.com/index.php?board=60.0) are very certainly some of the most legitimate around, and those include a lot of rematches. It would be completely stupid to exclude them from the rankings.
Currently, there seems to be no bad apple in the top 20, so your theory of inevitability has not yet manifested at least. Or such accounts have actually been removed...? :)

[gitsticker8]

On the contrary.  Games in a league or part of a tournament setting should not be rated for the same reasons I already stated.  I disagree that this a completely stupid idea.  Take a look at any other widely popular game with a rating system.  Starcraft for example, has a rating system that matches you with opponents of near equal rating.  Any custom games (including tournaments) are not rated.  I don't think this makes Blizzard completely stupid, quite the opposite.

[AdamH]

I'm just gonna weigh in here and say that the leaderboard would be far more legitimate if only automatch games counted towards it -- and any kind of blacklist wasn't taken into account. I agree pretty strongly with this statement:

I would also argue that the more control you give the player as to who they play against, the less meaningful the leaderboard becomes in general.

On the other hand, I don't think that means the best solution is to make the pro leaderboard this way. If there was money on the line for leaderboard position then maybe I'd say it was worth it, but I don't think anyone takes the leaderboard seriously enough that sacrificing this much fun is worth it.

I've never really cared about the leaderboard. I think that implementing all of the cards correctly, without bugs, and displaying all of the relevant information to the user, and all of the other stuff I constantly complain about, would have a more significant impact on how legitimate the leaderboard is than what's being discussed here.