Online-Rated and Bad Auto-Disconnects

Running Chess Tournaments
1

I had an interesting experience directing my first online rated tournament a week ago, and thought I’d share it for education and feedback. This was a G/60 d5, third round, played at one-round-per-week on chess.com. This is a club event, and only the two players playing had arrived in the third week to play rated. I was using WinTD to conduct pairings and having players manually challenge one another. All players who had played understood this was USCF rated.

The core of what happened is that an experienced online player, Black, after his move 11 stated that the interface told him that he had resigned, unexpectedly. He explicitly stated he had not hit the resign button. The download PGN states White, “won by resignation.”

It appears that chess.com does have an auto-resign on disconnect function from what I can gather, implemented sometime after 2011. Unlike ICC there apparently is not a way to turn that function off. The only conclusion that seems logical to me is that the player temporarily lost connection to chess.com and it auto-resigned him, but he regained connection after that occurred and it displayed the resignation to him.

Here’s what I did: First, I considered Black’s complaint valid (as in the player was being truthful). I did not look at the board or consider the position nor ratings involved at all during the ruling.

My first proposal was to split the remaining time and have both players begin a new game and play the first eleven moves identically. That would have advantaged White in that he had used the majority of time to that point (just under 25 minutes White used to just over 3 minutes Black used). However, White did not want to continue the game then stating he would have other matters to attend to. The round did start relatively late and the players did have problems with issuing the initial challenge; it took several minutes to get a handle on the situation and would have been many more before play would have resumed at the disconnect point. I then proposed an adjournment to sometime later in the week. White then asked about taking a half-point instead. Construing that as a draw request and therefore a draw offer, I confirmed with White that he was willing to offer the draw and asked Black if he would accept a draw. The draw was accepted and reported that way.

I think my rulings were consistent with the OTB sense of a result requiring a meeting of minds. I realize, though, I could have ruled consistent with, “the interface says it therefore it’s law.” Had it been cheating, or something larger than ‘routine’ club play, I’m not positive what I would have done. I could have just let the loss result stand, or I could have required play to continue and forfeited White if White wouldn’t continue. Also, I now recognize I’m not sure how I’d recognize if the system had applied a cheating algorithm other than the player not being able to log in again (?) - does chess.com append that something was forfeited? It apparently doesn’t mark an auto-resign as such, assuming that’s what happened.

It is a cautionary tale, however, that TD’s should clearly indicate that games may show result due to connection issues, and clearly spell out what the remedy for that will be. Perhaps a rule proposal should be in order akin to what to do with ‘mouse slips’ in the current 8th Edition Chapter 10. Or was there a more fitting rule that I somehow missed? Or is what I did the model of what we should do when confronted with situations that don’t quite fit the rules?

2

I don’t know if there’s any way to distinguish a network failure from some player-initiated action like severing his network connection.

Even frequent disconnects might be symptoms of a bad network rather than undesirable player behavior. (We have two different network connections in our house, I see disconnects on both of them frequently, remote sessions, like to US Chess servers, just fail.)

3

It sounds as though you felt your way through the situation in an excellent way. It’s rule 1A – sometimes you need to innovate!

One minor point:

Are you sure white was offering a (rated) draw, rather than suggesting a “forfeit draw”, i.e. not rated but 1/2 for each player? I’m guessing you got clarifications from both players as to what they were agreeing to.

I’m glad you didn’t do that! Common sense must always prevail.

Bill Smythe

4

Does chess.com support delay time controls now? It used to be only increment.

5

For that matter, does any online chess server support delay? Note that even CCA, which has never used increment OTB, is now doing so in their online events with ICC.

Bill Smythe

6

I have occasionally had something similar happen when playing on chess.com with a friend of mine. Each of us has on occasion had chess.com claim, when we were on move, that we disconnected and had therefore abandoned the game (even though we both had plenty of time remaining on our clocks), and counted it as a lost game for the player who was on move, which can be very frustrating for that player if he was winning at the time! But it made me think about what I’d do if I were directing an online tournament where this sort of thing happened.

I’d consider it to be the equivalent of an OTB tournament game in which a TD was walking by the playing table, lost his balance, and bumped into the table, scattering the pieces.

It would be inexcusable in that situation for the game to be counted as a loss for the player who was on move. If there were no way to reconstruct the position, you could count it as a forfeit. But that would be unfair to the player who was winning, so what you should really do is replay the game using the scoresheets, set the clocks to what they were when the accident occurred, and have them continue play. I wouldn’t consider this to be optional. If the players were in agreement that they didn’t want to continue play, they could always agree to a draw. But if one player wanted to continue and the other didn’t, the player who refused would be considered to have resigned.

Since chess.com keeps an error-free scoresheet, there should be no problem replaying the game. But I’m not aware of any way to adjust the players’ clocks in chess.com (which, if true, is an inherent problem with running official tournaments in chess.com, since there are situations in which clock adjustment is needed (e.g., to impose penalties)). If the clocks really cannot be adjusted, then you’d have to work out virtual time adjustments (e.g., White has to complete all his moves before his clock reaches 27 minutes, while Black has to complete all his moves before his clock reaches 39 minutes, and if either player sees that his opponent has exceeded his allotted time, he can claim a time forfeit).

Bob

7

Counterpoint: Players go online to play chess, not to put up with their opponents’ insufficient internet connections, which are an aggravation at best and an opportunity to cheat at worst.

I have zero problem with a loss upon disconnect convention under any circumstance.

8

Even a VERY GOOD Internet connection can have interruptions.

9

Acknowledged. Don’t care.

10

But it skews the rating reliability.

11

These are online ratings. They are separate precisely because standard conditions are lacking.

12

And what if the problem is at chess.com’s end? I’ve already seen one case where that was true.

The bottom line is that chess.com needs to fix this. As things currently stand, it is assuming that a player has abandoned his game when this is not true. If the session is reconnected and the player who is on move still has time on his clock, chess.com needs to allow play to continue, and if it doesn’t, then this is a bug in chess.com that they need to fix.

Bob

13

So are you saying that online ratings don’t need to be as accurate as regular over the board ratings? That maybe online ratings are meaningless?

14

What do you mean by accurate? A player who frequently loses due to disconnection might ought to have a lower online rating than would otherwise result. Seems like that would be a more accurate measure of “play” under these conditions (which are very different from over-the-board conditions).

(Note that this does not apply if it is the host server that is causing the disconnection.)

15

It’s not that they’re meaningless; it’s that they are what they are. My ICCF rating isn’t meaningless, but it has little to do with OTB or real time, engine-prohibited online play.

16

But if the online ratings are used for pairing purposes it provides a built in way of sandbagging for players that are far better than their online ratings predict. It even skews to unintentional results both ways up and down. If it happens occasionally then eventually the skewing will work its’ way out. But in the meantime you have a rating system that is untrustworthy.

17

I’m not convinced that sandbagging your rating gets you easier pairings. It depends on whether you wind up above or below the midpoint in your score group in each round. My guess is that especially in later rounds, a sandbagged rating is going to be below the midpoint more often than it is above it.

Sandbagging to get into a lower prize category is a separate matter, but that hasn’t been much of a factor in online events yet.