Escaping by Timeout


#61

a significant portion of results

How do you know it’s a significant portion? As far as I can see there’s a dearth of evidence for that.

Edit: O I think I know what you meant. Not the “games intentionally timed out” are the significant portion, but “games that have to be annulled due to timeout” make up a significant portion of the ranked games, as would follow from the statement that “if all these timeouts counted, the rating system would be significantly disturbed”. Gotcha.

Well, the trouble is that bad data is bad. If a large portion of corr games (serially) time out, leaving those results in would reduce the quality of our data. And the robust system you’re talking about is the one currently in place. :slight_smile:

I do like the “if a user logs in, cancel (serial) timeout annulment” suggestion though.


#62

Yes, the inclusion of osyang was a mistake because since I do not have access to the properties of the games it is not easy to quickly find the users in question.

Obviously, I found these users by searching through the users who recently timed out against me OR who did not time out against me because I prevented this by starting holiday mode.

Do you have anything to say about the existence of unrated time-out correspondence in the other cases or do I have to continue to accept that I have lied about the fact that this happens?


#63

Going back to the very first sentence of this thread…

And those three players you refer to…

What exactly is the issue?

Are you just offended by the ‘rudeness’ of bulk time outs?

Do you feel these people are actively managing their rank to yours and everyone else’s disadvantage?

Do you feel that your rank suffers because of these timeouts?

I can’t see that those players game history really bothers me hugely.

If I was inclined then perhaps I would be annoyed at the ‘rudeness’…

…but I’m not


#64

I referred to these players simply to defend myself against the funny idea that I am inventing that this happens.

What exactly is the issue? That these are ranked games with certain time settings that are not counted as ranked. This is the issue. It is essentially a bug, except that the behaviour has been programmed on purpose.


#65

You conveniently “overlooked” that I did not play any games from January to mid- September. Congratulations on this excellent presentation of data. Also, what prompted this thread is that I prevented two time-outs of won games of mass time-outers by switching to holiday mode. But of course, that is probably a lie as well.


#66

Bad news: Garbage in -> Gabage out. If you give a system bad input, you cannot expect good output. That’s true for all but not only for rating systems. So you discredit all sites with a rating system?

Glicko doesn’t work that way.

There is no normal play if your rank dropped to the bottom. You can only play ranked games against player within 9 ranks. And you need as many wins as you lost by timeout.

For what? There is nothing special about ‘Provisional’ players but the label.


#67

So even a game that has timed out that has only a few moves and a game where perhaps one is actually losing, that should be counted as a win and a resultant positive affect on one’s rank…?

I don’t think I would be happy with that personally


#68

I hadn’t looked, so I couldn’t overlook anything. I have better things to do than to click on each game in your game history to see what time settings they have.


#69

Well, since the name got published, and I feel the inclusion relevant to our discussion I will let them in. Seems fair to try and let them know @Cold (rest never registered to forums it seems)

And let me immediatelly offer “defense” for those players.

  • Toranius has a history of huge chunks of timeouted games. BOTH winning and losing positions and then coming back into play. Yes, I too find that rude, but sometimes that is life and Go should NOT be a priority in life for any of you! :stuck_out_tongue: Per chance (lucky us) that is exactly the case for which I consider the rule very good. Had it not been here, his/her rank would have been completely destroyed and again while you might think it fair, it would certainly not be fair to all the users he would then have “equal” games with they would have no chance of winning.

  • cold has had up to the most recent case a beautifull record full of fair resignations apart from a few sporadic and not batched timeouts sometimes in (at least as far as I can guess) not decided positions. And the 2 of the 3 uncounted games were against stronger players and thus would not affect rank that much. I am pretty confident in saying that he is not doing it to cheat anyone. (I also know cold a bit from chat)

  • Wimanado again having basically spotless record up to last 3 uncounted games one of which was far from being lost in my opinion. As far as I am concerned not misusing the rule to cheat, just something got into way, or stopped playing, for which the rule is well not good, but designed?

Also please try to keep in mind that the rank does not exist to “punish” players, but to try and provide the best possible estimate of abilities. Therefore while those who timeout might “deserve” it if you will, it is not what we are going for. And on the off chance that it was something serious that prompted them to time-out, imagine the good feeling when they come back to still valuable account :wink:

Now, that still does not take away the fact that it was unfair to you. You are right about that, noone is denying that and it sucks, I know, and am sorry for that. Unfortunately just taking the rule away would not take the unfairness away. It just shifts it to those who would play a de-ranked (is that a word?) account, those who got undeserved wins (although I guess they would not be complaining :slight_smile: ) and might (yes, I am not sure about that, but AFAI understand it is a possibilty…) cause trouble with ranking.

That said, I am for changing it if possible. I too thought it was stupid when I first heard it, but if it is to be changed it should be done properly and well thought out. The simple alternative of taking it away in my opinion does NOT solve the problem, just shifts it elsewhere.

And as a warning to all those who are really passionate about it, I am sorry to repeat that, but as a fair warning, before you waste too much time making up arguments and such, as far as I know there are no plans to completely reverse the rule. Think of another way, possibly, but not take it away as such.


#70

What are you talking about? Didn’t you notice the question you quoted was directed at Pimenta, not you? Had it been directed at you, it would have had the standard reply icon on it. I asked the question honestly because it was not entirely clear that he understood that it would have to be part of a consecutive series. You on the other hand do know that, so the question obviously has no applicability to you.


#71

and who gets to decide what is ‘Garbage’. This leads us back to the earlier suggestion of allowing the winning player to decide how they would like the game to be counted.

I would like to see the modelling on how it would work without the rule. Otherwise I remain unconvinced. If we both have our ranks inflated then they will still be reliable relative to each other. We can still use them to arrange a good game with a fair handicap.

Does this apply to tournament games where the rank difference is greater than 9? Most of those who repeatedly mass-time-out simply join too many tournaments.

This is just wrong. The system places more significance on more recent games.

Replacing their rank with the ‘Provisional’ label removes false advertising of their rank. I believe you have posted previously about how high uncertainty changes the way auto-handi works so that too helps. Also applying the ‘Provisional’ label allows tournament directors to exclude them from tournaments if they wish to do so.


#72

I would thank @flovo for bringing some actual data in the discussion.
Knowing that a specific number of games (and which ones) were annuled is a comfort.

I’m sad seeing how @Wulfenia is taking it all so personally: I don’t think anybody here thinks you are a liar or want to be offensive to you. But the comfort of data is a fact.
I’m sorry you couldn’t play any game from january to september this year. It must have been a pain. It would have been for me, at least.
If you think that 2018 isn’t a good example, let’s see at 2017. I hope flovo can do this, if required. Anyway, knowing that those specific games were annuled is good news!

It can help to identify any possible serial escaper. It would be nice to point them all in each player’s history, as some were asking. Hopefully with a bright colour. We would see escaper’s profile lit brightly.

It’s also good to know that all other timeouts was taken into account for your rating. This can’t mitigate the rudeness still, but I would feel heartened by the fact that all that rudeness wasn’t done intentionally: for cheating or manipulating rank… or, if it was, it didn’t succeed, which can be even more satisfying.

Think to my proposal (asking the winner whether annul the game): if you were given this opportunity, you probably would have annulled more than just five games. Or maybe not. I don’t know. I think I would. Personally, I was far more concerned by undeserved wins (by timeout) and having my rating inflated, than by escapers.

Also, you didn’t want to blame on specific users, and that does you credit.
But knowing the names can help to better understand.
I’m trying to say that asking for proof is not intended to mark you (or anybody else) as a liar but is just to better understand the situation.

You’ve seen me starting skeptical and then trying to dig the mechanics of this phenomenon.
You’ve also heard @GreenAsJade talking about his experience when he tried to check impressions with facts.

I believe nobody here is against you.
Let’s just try, all of us, to really understand how many escapers are around, how many games are annuled for that reason and, consequently, how much actually matters this disputed rule on multiple timeouts.


#73

Didn’t OGS used to have a feature where if a player timed-out, his “rating confidence” decreased, and thus players’ games against that player counted for less of a rating difference until he played more games?


#74

I’m not personally too offended by the rudeness of the timeouts. Neither am I too concerned about it having a large impact on my own ranking (since I mostly play ladder challenges where timing out seems to be less common since players get kicked off the ladder for timing out). I don’t personally feel cheated, since it really has not happened much to me.

However, I am concerned since several have expressed their frustration with this issue, and since the current situation makes it possible to abuse the timeout system to manipulate the ranking system. Since this vulnerability exists, griefers are actually incentivized to timeout in more games. The system rewards those abusing it at the expense of others.


#75

I’ve been playing almost exclusively correspondence games since starting on this server a couple years ago. Although I don’t feel strongly about this issue to the point of frothing at the mouth over it, I do agree with the argument that having such a rule in place does incentivize people to deliberately abuse the way the rule works, so I’m positive that there are at least some users who do so. I recognize that in many cases it’s difficult to prove any malicious intent, but ultimately I would very much like to see this rule removed altogether. My personal opinion is that having a longer time control should not exempt someone from taking a loss or series of losses on their record by timing out. If you are playing a live/blitz game, you inherently accept the fact that there are consequences to timing out, and I don’t think that correspondence games should be any different in that regard.


#76

Firstly, I don’t think the “inflating ranks” argument is valid, especially along side the “sandbag” argument.
Any ranking points he lost during the time outs (ie, “underserved wins” he opponents got), he got them back on his way back up sandbagging weaker opponents. So all the players‘ rank when counted as a whole, is not inflated.

Second, in the previous posts I have done a simulation. If I remembered correctly if someone times out many games of course his own ranking drops. But the wave dies out quickly. His immediate opponent’s rank would not rise by more than a kyu, and secondary impact (affect on opponent’s other opponent’s rank) is virtually nil.


#77

What about those getting sandbagged who he/she plays who are stronger than the timeout deflated rank?


#78

I’d like to see “concerns” and “possibilities” and “frustrations” replaced by facts.
Sometimes we are concerned by something that could be real but simply doesn’t happen or happens in a negligible way.

We are talking here about keeping or removing an existing rule that addresses a specific problem (some players simply stop playing for a long period and lose all of their games by timeout) but creates an even more specific vulnerability (cheaters could exploit it to automatically annul games that they are losing).

I’d like to know how many players stop playing and how many games are therefore lost by timeout in bunches.
I’d like to know also how many escapers are exploiting the rule and how many games are involved.
Is that possible?

If doing it on a full server scale is an unbearable waste of time for devs, could we do a portion of it on ourselves?

@flovo, please, how could we know for sure which games in our history were annuled because of that rule?
How did you compile the above list of Wulfenia’s games?
If I recall well, the termination-api skips those games, is it right?
So, should we compare two lists of games (profile history and termination-api history) or is there a better way?
Could it be done automatically for many players?

I would really know how big this issue is.

And talking about rating distortion: how much a bunch of losses could actually “destroy” one’s rating? And how much cheating the rule could inflate one’s rating?
And I would add this one: when honest and polite users know that they have to stop playing for a long period, they resign their games instead of letting them rudely run out of time. So, all of those games are regularly ranked as losses, whatever was the situation in game (ahead or behind). Would that destroy rating too? Would that have consequences on opponent’s rating?
Ironically, would it be better for rating system (with the present rule about timeout’s strikes) that they abandon their games instead of politely resign? :innocent:


#79

I would add that I’m not trolling or blaming anyone: I actually was a concerned player and my concerns were actually soothed by facts.
If you’re interested you can read it here:


#80

The mere fact that this question re-emerges every few months should be conclusive evidence that the rule is very unpopular amongst players.