Dominion Strategy Forum

Please login or register.

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

Author Topic: Handling disconnects  (Read 10326 times)

0 Members and 1 Guest are viewing this topic.

Polk5440

  • Torturer
  • *****
  • Offline Offline
  • Posts: 1708
  • Respect: +1788
    • View Profile
Re: Handling disconnects
« Reply #25 on: April 20, 2015, 09:30:19 pm »
+1

Reasons not to award fractional points to the highest precision:

Awarding fractional points advantages players who are better at early/mid-game play over players who are good at late game play. It's not just shuffle luck that determines who wins at the end.

It doesn't. You obviously take into account your own and your opponent's skills when estimating your win chances.


In a perfect world, yes. But I laugh if you think players actually have perfect information about this. And are able to honestly evaluate their opponent in the heat of the moment. See WW's over estimation of himself; the thread that sparked all this.

Now two examples is not a lot, but it's infinitely more than I know of fractional points leading to cunning bargaining exchanges or loud yelling, so I would really appreciate it if people actually looked at reality instead of just generating some fact-free rhetoric.

In fact, the only controversy involving loud yelling that I remember was of a case of people sticking to this "loss or replay" dichotomy, it might even have sparked this very debate ;)

So what was that other thread that sparked this one about then if not a full blown argument about win percentages?

There was enough confusion going around about this to spark several non-particapates to comment.
Logged

AdamH

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2833
  • Shuffle iT Username: Adam Horton
  • You make your own shuffle luck
  • Respect: +3879
    • View Profile
    • My Dominion Videos
Re: Handling disconnects
« Reply #26 on: April 20, 2015, 09:57:10 pm »
0

...

So yeah we don't disagree on anything actually. How strange  :P

I understand your reluctance and I don't think you should feel obliged to take this upon you if you don't want to. In fact if no one else does either I'd be happy to volunteer for the job. It sounds sort of interesting actually, I don't mind making potentially unpopular decisions, and if someone is going to hate me because he thinks I shafted him on 10% equity, then the only way I'm going to lose sleep over that is because I'm laughing too hard!

OK maybe we do? I'm not sure. Assigning win percentages to games is totally subjective, do you agree with that? Like, I don't even think you could make an argument that a skilled player is more likely to have a more accurate result than a non-skilled player. I also don't think it's a useful thing because of this.

But in any case, the important thing is that it seems like there's consensus about what the rules should be, the only big disagreement has to do with the granularity of percentages used.
Logged
Visit my blog for links to a whole bunch of Dominion content I've made.

SCSN

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2227
  • Respect: +7140
    • View Profile
Re: Handling disconnects
« Reply #27 on: April 20, 2015, 10:45:56 pm »
+1

It doesn't. You obviously take into account your own and your opponent's skills when estimating your win chances.

In a perfect world, yes. But I laugh if you think players actually have perfect information about this. And are able to honestly evaluate their opponent in the heat of the moment. See WW's over estimation of himself; the thread that sparked all this.

The beauty of it all is that you don't actually need that perfect information, just like you don't need to know how a good would be valued in an infinitely liquid market with zero transaction costs in order to agree to a mutually satisfying price with an interested buyer. You just both make guesses as to what the game state/good is worth to you, and see if you can find some common ground.

The objective is not to have zero estimation error, but to improve upon the crude {win, loss, replay}, which sets a ridiculously low bar, like asking whether I can get to the other side of the room faster than this snail. Well, yes, I can!

OK maybe we do? I'm not sure. Assigning win percentages to games is totally subjective, do you agree with that?

Judging whether a game is {win, replay, loss} or {win, 0.25, replay, 0.75, loss} are both subjective assesments, with the only difference being that the latter allows for more accurate and thus fairer expression. Why force someone into saying the grass is either red, blue or yellow when he can just say it's green?

Quote
Like, I don't even think you could make an argument that a skilled player is more likely to have a more accurate result than a non-skilled player. I also don't think it's a useful thing because of this.

Of course they can. Properly valueing one game state over another is most of what makes good players good. And again, the objective is not to have less than 2% estimation error or something ridiculous like that, it's just to do better than {0, 0.5, 1}, which is ridiculously easy to get right.

Quote
But in any case, the important thing is that it seems like there's consensus about what the rules should be, the only big disagreement has to do with the granularity of percentages used.

At least we agree on that :)
Logged

AdamH

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2833
  • Shuffle iT Username: Adam Horton
  • You make your own shuffle luck
  • Respect: +3879
    • View Profile
    • My Dominion Videos
Re: Handling disconnects
« Reply #28 on: April 20, 2015, 11:18:11 pm »
+1

The objective is not to have zero estimation error, but to improve upon the crude {win, loss, replay}, which sets a ridiculously low bar, like asking whether I can get to the other side of the room faster than this snail. Well, yes, I can!

OK maybe we do? I'm not sure. Assigning win percentages to games is totally subjective, do you agree with that?

Judging whether a game is {win, replay, loss} or {win, 0.25, replay, 0.75, loss} are both subjective assesments, with the only difference being that the latter allows for more accurate and thus fairer expression. Why force someone into saying the grass is either red, blue or yellow when he can just say it's green?

You draw comparisons to things that are falsifiable: "I am faster than the snail" "I can measure the color of the grass." The whole point of win percentages here being subjective was that they aren't falsifiable, so they aren't really useful for talking about anything. In my opinion, that includes the outcome of league matches. There's no way to show that one estimate is better than the other with any rigor because it's all subjective.

Quote
Like, I don't even think you could make an argument that a skilled player is more likely to have a more accurate result than a non-skilled player. I also don't think it's a useful thing because of this.

Of course they can. Properly valueing one game state over another is most of what makes good players good. And again, the objective is not to have less than 2% estimation error or something ridiculous like that, it's just to do better than {0, 0.5, 1}, which is ridiculously easy to get right.

I disagree with just about everything here. Properly valuing game states is not what makes good players good. The objective is not to do better than (0, 0.5, 1) because you can't even do that most of the time. Maybe there are game states that are advantageous to Player A over Player B, but if the game disconnects while Player A has a 60% chance of winning, if Player A is better than Player B, he might increase that margin as the game goes by. These are human people playing these games, who are imperfect. You can't possibly predict what they will do in the moment, how they will react to their draws, with any accuracy.
Logged
Visit my blog for links to a whole bunch of Dominion content I've made.

-Stef-

  • 2012 & 2016 DS Champion
  • *
  • Offline Offline
  • Posts: 1574
  • Respect: +4419
    • View Profile
Re: Handling disconnects
« Reply #29 on: April 21, 2015, 10:29:00 am »
+1

While I like this discussion, it feels like it's getting a bit too theoretical and leads us astray from the task at hand: come up with a rule on how to handle disconnects in the future.
In fact, I think I will get back into this discussion myself once we've completed that task.

For now I'll just do a new suggestion:


Handling incomplete games
If a game unexpectedly ends because either goko breaks down or one of the players accidentally disconnects (internet connection lost, computer crashes, ...) please use the following guidelines.
  • Discuss the game together with your opponent. Was the game completely open, or was either player ahead? And if there was a lead, was it small, significant or almost insurmountable?
  • Convert the size of the lead into a game result. If your felt like the game was completely open or the lead was small, simply replay the game. If the lead was significant, report this game as 0.75-0.25. If the lead was almost insurmountable then the player who was ahead just won the game.
  • Both suggesting an outcome and accepting your opponents proposal may be scary. You will never have a solid scientific argument why you picked the right one. Note that this rule doesn't expect you to pick the right one. All it tries to accomplish is make you feel less bad about the disconnect. If you really feel like you were ahead, and your opponent feels the same, then you are entitled to some equity and that's what you're getting here.
  • Please be willing to lose some equity for the mere sake of coming to an agreement. E.g. If you estimate there was a 5% chance to win the game, resigning it just because of a disconnect won't feel great - you just lost 5%. It's still the best option, because 5% is a lot closer to nothing then to 25%.
  • If you and your opponent cannot agree on an outcome simply because you continue to disagree about the game state, you have the option of leaving this decision to your group moderator. Your moderator won't enjoy making this decision for you either, so please do try to prevent it by coming up with a solution together, but if you really can't the option does exist. Both players should post in their groups thread what they think about the gamestate and what they consider fair before making a request to the group moderator to decide. There will not be any discussions about this - it's just the players stating their opinions first and then the group moderator making a decision. This decision could be a win/loss/tie, a 0.75-0.25 score or a replay.
« Last Edit: April 21, 2015, 10:49:19 am by -Stef- »
Logged
Join the Dominion League!

AdamH

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2833
  • Shuffle iT Username: Adam Horton
  • You make your own shuffle luck
  • Respect: +3879
    • View Profile
    • My Dominion Videos
Re: Handling disconnects
« Reply #30 on: April 21, 2015, 10:41:19 am »
0

Other than this typo

If the leas was almost insurmountable then the player who was ahead just won the game.

I like this very much :)
Logged
Visit my blog for links to a whole bunch of Dominion content I've made.

mpsprs

  • Explorer
  • *****
  • Offline Offline
  • Posts: 332
  • Respect: +169
    • View Profile
Re: Handling disconnects
« Reply #31 on: April 21, 2015, 01:02:59 pm »
+1

I like Stef's plan as well.  I think it'll cover most situations, and I think it'll work well, and I'm inclined to think that f.ds people will negotiate fairly and won't abuse the system.

One idea I did have (stolen from duplicate bridge (and no, I don't mean what you can now do with adventures)) is the idea of a recorder form.  The idea is to create some formal system that can track this to find patterns that might be abuse of the system.  It would be private (i.e. pm's to moderators), and only actionable with multiple reports from multiple people.  This could be a situation where one party felt a bit pressured towards one result, or maybe somebody consistently disconnects and potentially gains a slight advantage in the league, or perhaps if a third party (perhaps in the same bracket) feels some incorrect agreement that affects demotion in the group.  Things that are as one-off things stuff that will simply happen, and aren't a problem, but that could, if done repeatedly, give someone an advantage that is not seen for the pattern it is because no one person gets taken advantage of more than once.  These "recorder forms" would just sit and only be needed if somebody accumulates a lot.  Which is hopefully never. 

An implementation could be to simply have a dummy account setup where people after dealing with a disconnect could send a pm with a short summary, and which other people could also add to.  Maybe it's not necessary.  I certainly haven't ever had a disconnect that I felt was handled unfairly, but it could happen.

Also, it might be an interesting thing to discuss (in the chat perhaps of streams such as AdamH's) what class of disconnect a game would be at various points.  I'm not advocating always doing that, but having some better players occasionally talk about it might help lesser players such as myself have a better sense.

GeoLib

  • Jester
  • *****
  • Offline Offline
  • Posts: 965
  • Respect: +1265
    • View Profile
Re: Handling disconnects
« Reply #32 on: April 21, 2015, 10:15:45 pm »
+1

How do timeouts compare to disconnects? I've had timeouts before because I was chatting with my opponent and then had to make a difficult decision (can I win right now?). Hopefully timeouts will be handled better by Dominion 2.0, but in the meantime I think a clear ruling here would be good. This is one place where I think benefit of the doubt should not be given to the person who times out.
Logged
"All advice is awful"
 —Count Grishnakh

WanderingWinder

  • Adventurer
  • ******
  • Offline Offline
  • Posts: 5275
  • ...doesn't really matter to me
  • Respect: +4386
    • View Profile
    • WanderingWinder YouTube Page
Re: Handling disconnects
« Reply #33 on: April 22, 2015, 11:29:40 am »
+3

First of all, I perfectly understand if you guys want to entirely ignore all my opinions here, given that I have resigned the league and thus don’t really have any kind of vested interest as whatever comes out of this won’t directly affect me. Nevertheless, I have some thoughts on this. So read if you want, ignore if you’d prefer to do that.

1.   The most important thing by far is the mutual agreement of the players. If the players both think a game was too far gone, go with that. If they both think they should replay, go with that. Overriding a mutual decision of the players involved is not something that should be done.
2.   I strongly feel like which player disconnects shouldn’t matter. Adam basically already covered this, but I will add that, yes, this is potentially abusable, theoretically, yet I don’t feel that anyone would really do so. And further, it’s not like there are any prizes or anything – if there were, maybe more would need to be done. But if there were, it would really be necessary to have a more stable system which would allow reconnects.
3.   I believe that the default position should be to replay the game. Adam has made some points here, but the biggest thing to me is that with replays, the result is actually decided by a game of Dominion between the players, which is not something you can say for the other options. Now, I do think that there is a point at which you just award the game to the leading player, but it’s very difficult to say exactly where that should be (one of many reasons I want to underline point number 1).
4.   Having the players assign percentages is reasonable, but I am against it for a number of practical reasons. The biggest thing is that this is very difficult to do with any kind of accuracy. The Mic Q – SCSN situation has already been referenced, and if you were watching that, you know that Mic Q didn’t really want to assign a percentage. While you can stress here that it’s not that he had a problem so much with the concept of assigning percentages as it was that he had no idea how to come up with a  percentage, I think this is an endemic problem with the whole method. And even when people can come up with numbers some times, they’re very often going to be wildly inaccurate. I’ve seen tons of times where people think they are just winning or just losing when the game is completely unclear. Moreover, this leads to the nasty situation where you’re actually encouraged to disagree and haggle for extra percentage points. If a placement ends up getting decided by someone being able to haggle better, or being more persistent than the other player, this is going to be a really feel-bad moment. Yet having ‘assign percentages’ be the way things are done really does encourage this, to the point where you would almost have to push for this, and anyone who doesn’t overpress (which is I think a better way to act, in general) is going to be actively punished for this.
5.   Having third parties decide or adjudicate really seems the worst option of all. You can try to not take into account who the players are, but this is really inaccurate to how the game would play out. Do you assume perfect play? That seems very strange to me, since all players are really fairly significantly far off of that ideal. I’ve won LOTS of games I should be less than 5% to win if my opponent played perfectly. On the other hand, you can try to take that into account, but this leads into even more problems: suddenly the opinion of how strong the player is in the mind of the adjudicator is an enormous factor in determining how a ruling would go. Not only is a good evaluation thereof incredibly difficult to achieve, it’s also basically inviting the feeling that the decision is a personal one, even if the person arbitrating is trying to act as fairly as possible. Of course, the biggest elephant in the room is that the judgment of whoever is making the adjudication is very liable to be flatly wrong, on top of which all the time and effort those people would put in, as well as this leading to potential scheduling issues.
6.   How far into the game it is should be a real factor. While a player could easily be pretty far behind on turn 4 or even turn 2, I think that in these situations, you should be even more apt than usual to re-play such a game. The more of a game that’s been played, the more you want to just count that as ‘the official game’, the less that’s been played, the more of a preference there should be for playing a new one. You can’t really use turn numbers here, because some games will be over a lot more quickly than others.
7.   As a corollary to the last couple of points, there are going to be lots of situations where one player is, in actuality, way way ahead of the other player, because they’ve gone in for a much better plan, but the fruits of that aren’t necessarily borne out yet. I think these games are really poster-children for not adjudicating and indeed for going for a replay. Essentially, by awarding the game to one player or the other, it is just a nod of agreement with one plan or the other. And I feel very strongly that games should not be thusly decided, even if one plan really is much better. Of course, after game-state has played out enough to make it very clear whether the one side is going to ‘get there’ or not, it might be a different story. But it is still quite a tricky thing.
8.   Clear leads don’t necessarily mean the game is over by any stretch. There are loads of situations where deck A is clearly just better than deck B, say in a mirror, but shuffles can do a lot. Very simple example is having a 5/2 and getting just ambassador against a 4/3 player’s amb/amb. Amb/Amb is clearly ahead, but that game can really go either way. Obviously that’s a really simple example, but there are loads of things which are similar.

In general, I don’t believe that any solution whatsoever is actually fair. But (and tl;dr) the number one principle is that mutual agreement of the players involved trumps all else, and I strongly prefer deciding things with actual games of Dominion than by adjudication as  a default

SCSN

  • Mountebank
  • *****
  • Offline Offline
  • Posts: 2227
  • Respect: +7140
    • View Profile
Re: Handling disconnects
« Reply #34 on: April 22, 2015, 12:17:00 pm »
+2

Moreover, this leads to the nasty situation where you’re actually encouraged to disagree and haggle for extra percentage points.

Logged

drsteelhammer

  • Torturer
  • *****
  • Offline Offline
  • Posts: 1527
  • Shuffle iT Username: drsteelhammer
  • Respect: +1471
    • View Profile
Re: Handling disconnects
« Reply #35 on: April 22, 2015, 01:09:27 pm »
0

Firstly, to Stef's proposal: Personally, I like the general idea, but I feel the third person who decides should not be able to declare a game a tie. Such close games should be replayed. (except in that edge case where the game crashes right before p2 buys the last province).

To WW: I don't know why you think you are being ignored here, but most of your points are covered in Stef's proposal: Your first and second point are literally the same as in the  provisional rules.

Your third point is valid, but your reasoning is kind of misleading, I feel. The league consists of 6matches, why should the default position be that I play another to overrule my already started game? I might feel frustrated when I have board where I had a decent strategy on and then being forced to play another which I might not play very well. Of course one could make a case for replaying, but I don't think your reasoning is good here.

4: This is a valid concern, but our previous ruling had  much more problems with such cases. Look at the SCSN-Miq game: Mic would probably feel bad about losing  and SCSN would feel bad about replaying that game. This solution adds a middleway for such cases. (I like this being restricted to 0,25-0,75,win/lose or replay, by the way). Talking about persistence: The same problem existed before, people felt that you were being too persistent about replaying that one game. This ruling gives both people more negotiating power, because you are only talking abut 0;1/4;1/2;3/4 and 1 instead of only having 2 options.

5: Coming from a sport where there is no referee either in casual games: Sometimes it is nice having a third party as both the players are biased (esp in non-mirrors, I assume) Things here could be said about favoring the one who doesn't disconnect to diminish frustration, but that is another point.

6: I think both players will want to replay if you disconnect very early. I think this is covered in Stef's first point.

7: I agree that is why having a third person look over it might be good. (Our moderators are skilled, I doubt they overlook someone building a megagturn when the other one is  raising duchies and simply gift the game to the point leader)

Logged
Join the Dominion League!

There is no bad shuffle that can not be surmounted by scorn.

-Stef-

  • 2012 & 2016 DS Champion
  • *
  • Offline Offline
  • Posts: 1574
  • Respect: +4419
    • View Profile
Re: Handling disconnects
« Reply #36 on: April 22, 2015, 01:21:56 pm »
0

Firstly, to Stef's proposal: Personally, I like the general idea, but I feel the third person who decides should not be able to declare a game a tie. Such close games should be replayed. (except in that edge case where the game crashes right before p2 buys the last province).

Oh in essence I agree with you here, it's just a practical thing. That is why if the players themselves agree that it's close the game will always be a replay and never a tie. When the moderator has to decide a couple of days later a replay is still the preferred solution for close games, but a tie is an option if (and only if) the players don't manage to schedule another game.
Logged
Join the Dominion League!
Pages: 1 [2]  All
 

Page created in 1.972 seconds with 21 queries.