Escaping by Timeout

How about a system-check: If the player is active playing in other games, the timeout’s he gets count’s as loss ? Is that possible?

and what’s wrong with that?

So the “victim” of a time out is to blame? for not blocking someone who break a rule? interesting idea.

2 Likes

Yes it’s possible. The proposal to have the ‘Recently Timed-out’ flag disappear as soon as someone logs back in would achieve this result. #3 in this post.

In my experience most of the games in question are from tournaments so blocking is irrelevant.

3 Likes

don’t you think it is a little bit unfair to criticize my effort as “at least approaching factual data, (but still unconvincing)”, while at the same time I have not seen a single piece of factual data (convincing or not) supporting the “rank inflation” argument.

3 Likes

I’ve been trying to write a form of words to explain myself but I can’t, so let me just say that the sentiment expressed here is very troubling for me

1 Like

Unfair? No, I honestly did not, my appologies if it came out as somehow offensive :slight_smile: You said you thought you proved “it would not mess up ranks”, and I merely pointed out what data I still felt lacking in the proof in the hopes that someone capable of doing such a simulation would repeat it including for those eventualities.

And since I have serious concerns that we are bound to repeat the same arguments over and over, let me try to summarize the whole thread in simple points:

The concerns are:

The rule can/is be misused for cheating:
While potentially it clearly could, I personally think there is no evidence of this being currently any major issue (also with much easier and avilable ways to cheat avilable), thus I personally do not find that to be in urgent need of any rushed fix.

It is unfair
Because in those cases where you are winning you do not get the points. That is obviously true. But in those cases where you were losing you do not lose any points, so it might be partially negated.

(are there any more not connected with these? let me know I will add them)

If the rule was canceled:

rank problems
In case of multiple innocent timeouts the rank of the timeouter gets misrepresented. Which is (in my opinion unnecessary bother for the player, since it does not make it fair for others anyway) but more importantly against the very purpose of ranking.

it is still unfair
While less annoying, games are still being scored wrongly in the other direction, and if the player comes back he/she will be unintentionally “sandbagging” which too is super annoying and unfair to play an “eqal” game against 5stones stronger player. And while the annullments simply “do not matter” these games WILL count towards your rank.

might mess with ranking further
Truth or not it has been raised as a concern and AFAI am concerned not disproved, thus it remains a potential worry.


My completely personal opinion is that canceling the rule hence brings no major solution to the concernes raised and with the added worry of other problems makes no sense (to me) to have it removed. I do not think players’ ranks suffer any substantial misrepresentation because of a few uncounted (not misscored) games. But it is of course up to discussion if you feel otherwise.


Sugested other solutions:

1. Allowing the winner of such games to determine how it is counted
maybe I am a sceptic but I just feel most players would take the win without really thinking about it (not out of spite, I don’t blame them, we click without reading all the time, or they would feel it deserved since the other player did not “fulfill” the contract).

2. Removing the ‘Recently Timed-out’ flag from accounts as soon as they log back in.
This seems to only solve the cheating problem which as far as I am concerned does not seem to be too big a deal.

3. Automatically triggering Vacation for people who start a string of TO’s.
This is probably my absolute favorite, making it hard to cheat, while being helpfull to people with unexcpected events that still want to play. Can anyone think of any potential problems with it?

Personally, I am happy to try and press the auto-vacation a little if you feel like it would help resolve your problems, but it will still on itself not resolve the case with people who stop playing altogether.

To adress some minor comments with the same idea:

  • it is their problem/ they deserve the rank drop, etc.
    I disagree. The ranking system is NOT means to punish people, it is to try and represent their strength, nothing else.

I urge you again, try to stay stricly on topic and away from personal potshots at this one. You are only runing the chances of some resolution otherwise. And without wanting to sound rude, I will from now on become much less active in this thread as it is taking too much of my time and I feel like we are still largely in the “emotions” rather than “ideas” territory. If you feel like you have something new/important to present, please take the time to tag me, I will try to have a look/answer when I have the time. Thank you and sorry for the long text.

4 Likes

Do I need to repeat that if I see a clearly lost game slight towards time-out I can simply resign? Or are you asking me to renege on my commitments to fair games just to “make it fair” again? I resent this aspect most of all: being pressured to simply act as inconsiderate and respectless.

1 Like

I want to point out, how a intentional cheater can exploit the suggested solutions. (In case there are any intentional cheaters.)

  1. Allowing the winner of such games to determine how it is counted
    It will not so effective, but as long as someone chose to annul the game, a cheater can at least reduce the number of counted losses

  2. Removing the ‘Recently Timed-out’ flag from accounts as soon as they log back in
    The obvious workaround is to let the clock tick down in all lost games while keep playing in the other games. Before the first games time out log out and take one or two days off. Then come back before your good games time out and continue playing.

  3. Automatically triggering Vacation for people who start a string of TO’s
    A cheater has only to keep the vacation time low while not intending to time out. When they want to timeout the lost games there is not much vacation time left. They now just activate vacation and stop playing in their lost games. After all vacation time is spend, the lost games still timeout on mass.


To address the not cheater:

  1. Allowing the winner of such games to determine how it is counted
    :man_shrugging: nothing special to say here

  2. Removing the ‘Recently Timed-out’ flag from accounts as soon as they log back in
    :man_shrugging: they not even know now that mass timeout is a thing. Now they are back and continue playing anyway in every of their games. (If they time out semi regular (like Koba) this could be a minor change)

  3. Automatically triggering Vacation for people who start a string of TO’s
    This could be quiet helpful for player which forgot to activate the vacation or suddenly got serious live events (illness, …) preventing them from playing.
    For player which always have problems to keep playing one move in 24h (like Koba) this will only be a minor help. Like Koba wrote, they are always low on vacation time anyways and would therefor timeout anyways a few hours later. But if vacation activates automatically, they wouldn’t have to activate vacation ahead of time, so they could have more vacation time left if it is really needed.


Additional notes to

  1. Allowing the winner of such games to determine how it is counted
    This will definitively disturb the ratings, since some player will always count timeouts as win and others will annul almost all timeouts (me for example). This would push the ratings of the first group upward compared to the ratings of the second group. Even if the difference will stay within a few ranks it will render ‘Handicap: Auto’ useless, since the ratings aren’t accurate enough anymore.

To run a simulation how big the impact of timeout rule on/off is, I would need the results of all ranked games (especially the games affected by that rule, the other games are fast and easy to collect). With this data one could recalculate ratings for both rule choices and check how it would impact the accuracy of the rating system for the outcome of future games (as needed for handicap calculation and different matchmaking options in tournaments).
Interesting cases would be

  • how long it takes mass timeouter to get back to their expected rank (if at all)
  • by how much the rank of correspondence players is inflated/reduced compared to live player due to the timeout rule options (especially for the players won by timeout). This would include the current rule affects the ratings.
  • with a database of all timeouted correspondence games, one could check at which move games usually end by timeout. This could provide evidence for the amount of intentional timeouts.
  • with a database and much computation power one could run a strong AI over the games to see how many games are clearly lost/won or undecided.

I have neither the data, time nor a reliable enough implementation of the OGS rating system to do this at the moment.


same here

1 Like

Not directed at you personally in any way. I am simply suggesting to consideration whether the fact that “the sure wins” are not counted is not partially negated (for the purposes of ranking) by the fact that the potential losses are also not counted.

Tbh I do not expect anyone to go through all games (where there is not your turn) and checking whether the opponent is not close to timeout and if so resigning those games when one feels behind. If you do that, you are crazy :smiley: but my deepest compliments for that.

Here is an idea for consideration. Since the potential effect of the rule, or absence thereof, on the whole ratings pool is a concern, how about doing this:

Going forward, suppose everyone who wants to play a correspondence game would have to do so in a separate account just for correspondence games (I imagine these would be hard coded for just that purpose, so there would be no mixing of games in either account). Games already in progress would be completed on the current account. The overall rating would be abolished; there would be just a correspondence rating and a live rating. In this circumstance, the correspondence timeout rule could be abolished (for new games), with everyone understanding the consequence of timing out a string of correspondence games, and it would not pollute the live game rating pool.

The obvious downside is that this would be a huge and complicated change to the whole way of doing things on this site, and it would likely be too much work for the devs to implement it.

Mark this: I am NOT proposing this, I am NOT advocating this, I am NOT at all sure this is even a good idea—I am just laying it on the table for discussion because I think it may be a new idea.

2 Likes

Good instincts! :laughing: but I do see something usable in the idea. Forget multiple accounts. Much of what you suggest could be achieved by having the system make active use of the separate ratings categories that already exist. Specifically ‘Overall Correspondence Rank’ and combine Overall Live/Blitz for the other one.

If the system was programmed to focus on these two ranks as two completely separate pools of data, then there would be a much better case for not annulling TO’s. The ability to finish a game can reasonably be considered to form part of your ‘Correspondence Rank’.

3 Likes

Why not just a separate server just for correspondence Go?

You could call it… Correspondence Server : Go, or CS:GO for short. Millions of subscriptions await. :sunglasses:

3 Likes

Well, since so many people are displaying palpable anger while digging themselves deeper into a heated debate which essentially boils down to “yay” and “nay”, I thought I’d offer some comedic relief.

I admit it wasn’t the most creative interjection, I’ll work on my delivery.

2 Likes

Why will this change anything?
If my rating depends on how many of my opponents timed out, it isn’t useful at all.
Rank difference = handicap stones, but if all timeouts count, rank difference = what?

In the end we all benefit from acting in good faith. If we suspect ill intent everywhere…

  • Someone didn’t greet you at the start of the game? Probably hasn’t registered an email yet. Clearly rude.
  • Someone only makes a move 5 minutes before his time runs out? Must have a busy schedule. Clearly wants to annoy you.
  • Someone continues the game despite a lost position? Ah, doesn’t know any better. Clearly griefing.

DGS offers the option to be benevolent.

Now I don’t know if this would help much, but maybe we could implement that, with a little twist. Users get a “balance” of wins they can give themselves by accepting the timeout. This balance might increase with every loss people voluntarily give themselves, or annulment or whatever.

The ostensible loser of said timeout would be notified that their game is on hold and they have some time, say 24 hours, to either accept the result (whatever that may be) or come back to the game (in which case the x hours it took them to react will be deducted from their vacation).

2 Likes

Thankyou for further illustrating my point. ie “total consideration is being given to the mathematics and no consideration is being given to the psychology”.

We already have a ranking system that 99%(not a statistic) of users could not possibly understand. It is therefor equally important to consider how the system is Perceived as well as how good a ‘best fit’ it is. Is it perceived as fair? Is it perceived as having legitimacy or is it perceived as being an esoteric system for the interest of mathematicians and programmers?

When I hear directly from Anoek saying that the SOLE purpose of the ranking system is to optimise handicap matches and that he is completely unconcerned with how the system is perceived by users, then I will stop exploring ideas that exist in-between ‘Yay and nay’ that might improve the system.

My attempt at comedic relief (click here)

Why do Go players insist on trying to make issues ‘Black & White’?

1 Like

I’m certain that the old OGS had something along these lines but how exactly it operated I can’t recall

What is the difference in functionality between separate accounts and separate rankings for live and correspondence, just as we had not so long ago?

1 Like

I find the separation idea to be a red-herring. For a start, I only play correspondence :slight_smile: And secondly, it seems to be arguing that only a live rating matters, and it’s OK for correspondence ratings to be subject to arbitrary impacts like “oh, you didn’t get that game because your opponent ran away”.

This whole debate is only about correspondence games, the rule only applies to them. So the idea of farming the problem off into an account that “doesn’t matter” seems odd.

There is actually a premise underpinning this whole thing: that we care about our correspondence rank.

Much of the line of argument that smurph is presenting can be encapsulated as “you simply shouldn’t care so much”.

The suggestion of “let’s just be able to arbitrarily give up some rank to receive later” etc comes from this approach. It builds on the idea that “your rank really doesn’t matter, just relax about it”.

I totally acknowledge the benefits of this philosophy.

But that being said, “you should see the world the way I do” is another way of saying “your experience of this problem is invalid”.

This way of solving someone’s problem is often less than helpful.

Actually people do care a lot about their rank, and they work hard to grow it. Those people do experience the problem being debated here acutely.

While we can give them philosophical advice to rectify that, I think it’s equally helpful to explore ways to actually fix the problem that they are experiencing.

Separating the problem into a separate account, or introducing new means to undermine the value of someone’s rank don’t seem do that.

2 Likes

I mentioned that because separating the game types by account seemed the easiest way to get rid of an “overall rank” that includes correspondence. I’m not saying I want that, but since the effect of the correspondence timeout rule on the overall rating appears to be the main objection to the rule, this would be a way to address that objection. An added advantage would be that it would simplify keeping track of the two types of games for people who play both (some people who play both may already do this).

1 Like