Resigning on behalf of machine in human vs computer games

[Monsieur Plastique]

Hi,

A little while back I started a thread about playing computer vs computer games through to mate. Whilst I fully believe this needs to be done if seriously trying to evaluate programs against each other (and there are examples in that thread as to why the games need to go to mate), I believe the situation arguably needs to be different in human versus computer games.

SteveB brought up a very good point in that thread (which I completely agree with), that if a machine resigns itself, then that is a loss to the resigning machine. But of course, many dedicated machines do not have resign functions, so the human either has to play the game to mate or simply hit the "new game" button or power the machine off.

I played a game against Excalibur Alexandra yesterday which highlights the resignation issue. It was a 40 in 2 hour game, and by move 40, the queens were both off the board and Excalibur had both a worse position and was three pawns behind in material (piece material was the same, but I had 5 pawns to Alexandra's 2 and a better position). But of course, with no resignation function, I continued to play the game to mate. The last 40 moves of the game were - to be quite honest - a boring affair for me, since the win was inevitable and simply required time and avoiding an outright blunder. I felt like I was simply wasting 4 hours of my time. I would have preferred to go off and do something else or play another game.

When I had an engine evaluate the position after the game, it told me that my position was about 6.7 points ahead of Alexandra at move 40 (and the queens were off the board). For the remainder of the game (which went for around 80 moves), the evaluation steadily crept up consistently, meaning I was not making any blunders or inaccuracies for the rest of the game.

So I am thinking of adopting a rule similar to perhaps what some of the stronger dedicated machines use. I read somewhere here, for example, that some Mephisto machines will resign if Queens are off the board and the position evaluation is -4.99 or worse for three consecutive full moves.

And then of course there are resignation functions built into virtually all engines these days - therefore I could turn an engine on if I feel I have a won game and see if the engine resigns - if it doesn't, I would play on against the dedicated.

I am just wondering if other people here have adopted particular rules for resignation of a computer opponent if they have an obviously won game where any respectful human would resign in an OTB slow time control game? I would be interested toknow what rules you follow.

I realise that playing through to mate is sometimes required, as the position may be ambiguous or there may be drawing chances. And I would always play through to mate regardless in rapid-play or sudden death games (and obviously blitz). What I am talking about here, though, are slow tournament games (40 in 2). But if the position is clear cut, I am not sure that as a human I really want to spend another 4 hours merely doing what amounts to house-keeping and mopping up, especially as time is precious. I want to have fun playing games and I want to use that time to keep my mind sharp as I get older. Being bored is counter-productive to all of that.

Missing the Scisys Mark V Regards

[SirDave]

I totally agree Jon. Not only are games at that point boring, but then the more bored i get and more frustrated that I am wasting hours I'll never get back, the more my mind starts to wander & then it's more likely I'll make a careless tactical mistake and lose a game I knew I should have won. All this because the damn machine wouldn't admit defeat!

BTW, this is one reason why I like boards that give a realistic score based on pieces left & position. If I'm well ahead I can tell the the thing it's a loser with full confidence.

[Monsieur Plastique]

I hear you Dave. I never worried about this when I was a young person. 8 hours seemed nothing for a chess game back then. But now I am well and truly over the halfway point of my mortal existence, I get rather annoyed when I waste my time! I also admittedly do not have the mental stamina I had in years gone by, but by the same token I do not feel this is an "excuse" for a machine to simply play on in an attempt to exploit that in the hope I make a monumental blunder.

[Steve B]

So I am thinking of adopting a rule similar to perhaps what some of the stronger dedicated machines use. I read somewhere here, for example, that some Mephisto machines will resign if Queens are off the board and the position evaluation is -4.99 or worse for three consecutive full moves.

And then of course there are resignation functions built into virtually all engines these days - therefore I could turn an engine on if I feel I have a won game and see if the engine resigns - if it doesn't, I would play on against the dedicated.

I am just wondering if other people here have adopted particular rules for resignation of a computer opponent if they have an obviously won game where any respectful human would resign in an OTB slow time control game? I would be interested toknow what rules you follow.

Hi Jon
personally i dont like the idea of cutting games short for arbitrary reasons
but then again i never really play 40/2 games and i can imagine that playing another 2-3 hrs with one computer up a R+3P's can be very frustrating

you should ask the SSDF guys this question
Lars Sandin from the SSDF posts here all the time
i imagine these guys have grappled with this issue and have developed a good rule of thumb by now
their 40/2 ratings have been the basis for much information on the strength of many dedicated computers for several decades

SSDF Regards
Steve

[Monsieur Plastique]

Hi Steve,

Aren't all SSDF games computer versus computer though?

[Steve B]

Hi Steve,

Aren't all SSDF games computer versus computer though?

Woops ...
i see now you only meant premature resignations for human vs Comp games and not Comp vs Comp games
sorry..my bad
still..dont see how you can resign for a comp when you as a human can blunder a piece at any time ..the comp will never do that
it is precisely because of the fatigue factor or the fact that you got bored
that could cause a human to make a move that blunders a win into a draw or a draw into a loss or even a win into a loss
this happens all the time in OTB human vs human play
happened to me all of the time ...OTB ..when i played competitively
i cant tell you how many times opponents would make me mate them
instead of resigning in hopeless positions
or how many times i blundered in winning positions
turning Wins into draws and draws into losses
i used to resign for myself when i thought my chances were hopeless as i was not interested in playing on in the hopes my opponent would blunder
but thats just me..i never allowed myself to play on in hopless positions until i was mated and i was rarely ever mated (unless i didnt see the mate coming on the next move!)

your ideas just seem to favor the human too much

perhaps its just best to play until you no longer feel the game is interesting
these days thats what i do anyway
but again..i never play 40/2


Equal Rights For Humans and Computers Regards
Steve

[Theo]

The last 40 moves of the game were - to be quite honest - a boring affair for me, since the win was inevitable and simply required time and avoiding an outright blunder. I felt like I was simply wasting 4 hours of my time. I would have preferred to go off and do something else or play another game.

Hi Monsieur,

I have had the same problem. On the one hand, I wanted to have a valid result, not just a "I believe white will win this". On the other hand - the boring part of "finishing off" costs too much time.

This is why I play all my games with a fixed time per game, not per move.

There are also some games, which are thrilling till move 100 and beyond - they would cost you hours normally.


[Date "2013.06.25"]
[White "Revelation Super Expert C"]
[Black "Nigel Short"]
[Comment "Ns16, rsec17"]
[Result "1-0"]

1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Bg5 Be7 5.Nf3 0-0 6.e3 h6 7.Bh4 Ne4
8.Bxe7 Qxe7 9.cxd5 Nxc3 10.bxc3 exd5 11.Qb3 Rd8 12.c4 dxc4 13.Bxc4
Nc6 14.Be2 Rd6 15.0-0 Be6 16.Qb2 Bd5 17.Rfc1 Rb8 18.Rc3 Qe6 19.Rac1
b6 20.Bc4 Rbd8 21.Bxd5 Qxd5 22.Qb3 Qe4 23.Nd2 Qg6 24.Qb5 a6 25.Qxa6
Nb4 26.Qa3 c5 27.Rb3 Rd5 28.Nf3 R5d6 29.Ne5 Qe6 30.Nc4 R6d7 31.Qb2
f5 32.h3 f4 33.a3 Nd5 34.Nxb6 Rc7 35.Nxd5 Qxd5 36.Rb8 fxe3 37.fxe3
Rcc8 38.Rxc5 Qxc5 39.dxc5 Rxb8 40.Qc1 Rb7 41.Qc4+ Kh8 42.a4 Rc7
43.a5 Rdc8 44.a6 Rxc5 45.Qa4 Rc4 46.Qd7 R4c7 47.Qd3 Ra7 48.Kf2
Rca8 49.Qd5 Rxa6 50.Kf3 Ra5 51.Qe6 h5 52.Qf7 Kh7 53.e4 Rb8 54.g3
Rb6 55.Qe7 Kh6 56.h4 Rab5 57.Qc7 Rb3+ 58.Kg2 Rb2+ 59.Kh3 R2b5
60.Qf4+ Kg6 61.e5 Re6 62.Qg5+ Kf7 63.Qxh5+ Kg8 64.Kg2 Rbxe5 65.Qd1
Kf7 66.Qf3+ Kg6 67.g4 Re2+ 68.Kf1 Re1+ 69.Kf2 R6e5 70.Qc3 R1e2+
71.Kf3 Kf6 72.Qc6+ Re6 73.Qc5 R2e5 74.Qf8+ Kg6 75.h5+ Kh7 76.Qb4
Rf6+ 77.Kg3 Re3+ 78.Kh4 Rfe6 79.Qb1+ Kh8 80.Kg5 Re1 81.Qb7 R1e4
82.Qd7 Kh7 83.Qf7 R4e5+ 84.Kh4 Re3 85.Qd7 Re7 86.Qf5+ Kg8 87.Qc8+
Re8 88.Qc4+ Kh7 89.Qc6 R3e6 90.Qc7 R8e7 91.Qc2+ Re4 92.Qd2 Re8
93.Kg3 Re3+ 94.Kf2 R3e6 95.g5 Rf8+ 96.Kg3 Rb8 97.Qd5 Rbe8 98.Kg4
Re4+ 99.Kf5 Re1 100.Qd3 R8e3 101.Qc4 Rf3+ 102.Kg4 Rf8 103.Qd3+
Kh8 104.h6 Ref1 105.Qd4 Rg8 106.g6 Rf6 107.hxg7+ Kxg7 108.Kg5
Rf8 109.Qd7+ R6f7 110.gxf7 Rxf7 111.Qe8 Rf8 112.Qg6+ Kh8 113.Kh5
Ra8 114.Kh6 Rg8 115.Qh7# 1-0



Kind Regards,
Theo

[Monsieur Plastique]

i used to resign for myself when i thought my chances were hopeless as i was not interested in playing on in the hopes my opponent would blunder
but thats just me..i never allowed myself to play on in hopless positions until i was mated and i was rarely ever mated (unless i didnt see the mate coming on the next move!)

Hi Steve,

Actually what you describe there is typically what I have found to be a normal practice in adult tournaments with slow tournament games at mid club level and above. The only times I have seen hopeless games go onto mate at slow time controls are in adult novice tournaments, casual games and child / school tournaments, etc.

But for the most part I see that club level adult players will usually resign if the position is decidedly lost regardless of whether there is a theoretical chance of a draw or even a win if the other side makes a monumental blunder.

I understand the possibility of blunder but then again blunders happen to everyone right up to World Championship calibre players and can happen in the opening all the way through to the end of the game. But you never see such players play on in lost positions unless there are time control issues where there is, for example, a chance the other player will lose on time, or a chance they will make a monumental blunder because they are in severe time trouble.

In the end though, whether to play on or not comes down to whether a resign function is programmed into the software or dedicated computer one is playing against. If every dedicated computer was programmed like the original Scisys Mark V was, there wouldn't be a problem at all obviously. But in the positions I am thinking about, all engines would absolutely 100% resign and all dedicated computers with resign functions would resign as well.

So the practical outcome is that simply because the dedicated computer one is playing was not programmed with a resign function, the game goes on for twice as long as it otherwise might. And the reason these programs did not have resign functions wasn't because such functions couldn't be programmed into the units, but primarily because space was so severly limited in the early days. It would take a competant programmer very little time in relative terms to add a resignation function and it is one that is easily tested. So really, it either comes down to space available or slackness on the part of the programmer.

If the fatigue / chance of blunder thing was such a big deal and a deciding factor, there would be no such thing as resignation in chess. All games would have to go to mate regardless. And we would probably find that alot of the appeal of the game - especially for the players themselves - would disappear along with that.

[Monsieur Plastique]

This is why I play all my games with a fixed time per game, not per move.

That was something I had considered, but unfortunately many earlier dedicated machines did not come with sudden death time controls. They only really became ubiquitous in later generation machines. And sadly even the later generation machines do not necessarily have a wide array of game length choices. I tend to see typically game in 1 hour as the maximum levels for this type of game. That is a much shorter game than I like to play. It is the 8 hour games I don't like unless the game remains a tight contest for the duration (in which case I would adjourn it and come back the next day).

I do change levels at move 61 if possible to game in 30, since a proper tournament time control is for the remainder of the game to be finished in 30 minutes for each player after the second time control (machines like Alexandra simply begin a new 40 in 2 time control every 40 moves, which is not correct, but not a significant problem either because it is easy to change to a sudden death level at any time).

And of course the other consideration is that one can have a fantastic slow game that remains interesting for hours on end, but may be compromised by a time control that sets a strict limit for the total length of the game. I have played many of these types of games as well.

I think the best solution is to simply have the computer opponent resign - it is just a pity that it is such a relatively exclusive feature in the domain of dedicated machines. And in the end, there really isn't any excuse for a 32K program (or larger) not to resign. Even programmers from those 16K and 32K days often admitted that there was spare space in a 32K ROM even after having added the main code and the openings book. So it is a pity given that even, for example, all these "modern" Excalibur machines do not resign.

The more I think about all of this, the more admiration I have for the original Scisys Mark V - a machine I love and have always admired and coveted - probably the machine more than any other that I would love to still see in production today. You could program your own time controls into that and it resigned too of course. It was way ahead of it's time - in some respects even by the standards of today.

[Steve B]

Well my playing experience has been different then your i guess
i played for my high school and University teams in rated games
once i graduated university that ended my playing career
all i can say was the competition was fierce and if the opposing team needed a win there were very few early resignations to be found
in fact..it was common practice to find oneself up in material and being offered draws every 2-3 moves
just to annoy you and throw you off balance
sure you can call the ref over..but then it became a.. he said ...he said debate..again throwing your concentration out the window

club play is totally different
here you rarely ever see anyone play on even a piece down..and most times if only a P or 2P's down they resign
this is because folks dont want to get a bad reputation as a sore loser from the other club players
also club play is generally not rated play

My Perspective Regards
Steve

[Monsieur Plastique]

Sounds like your opponents were like my Fritz for Nintendo DS. I am sure I have related previously how I was testing out it's draw offer and resignation functions....

I was a rook, queen and two pawns up and it still would not resign. Then two moves from a forced mate, not only would it not resign, but it declined my draw offer.

Did your opponents act as consultants for the Fritz Nintendo project?

[Monsieur Plastique]

Actually, your comments made me remember an episode from my school days. It was 1977 or thereabouts and I was second board on the school chess team. We were playing a nearby school team - 4 boards in total.

I have to admit that my opponent was outplaying me and my position was not terribly good, but not resignable. At one point in the game, my opponent said to me very loudly and clearly so the entire room could hear: "Would you like to resign now?". I said "no, I would like to play on please". He replied: "It is pointless, you are going to lose anyway".

I was so taken back that I was speechless, as were the (adult teacher) team captains of both schools. I then seem to recall about 15 minutes later the chap did that golden age classical grandmaster versus amateur thingy of announcing a mate in upteen moves and then counted them off one by one.

I often wonder where that chap ended up and what he is doing now...the mind boggles.

[Steve B]

I often wonder where that chap ended up and what he is doing now...the mind boggles.

Hopefully he is presently incarcerated serving the first few years of a 15-20 year sentence

Sweet Justice Regards
Steve

[Steve B]

Sounds like your opponents were like my Fritz for Nintendo DS. I am sure I have related previously how I was testing out it's draw offer and resignation functions....

I was a rook, queen and two pawns up and it still would not resign. Then two moves from a forced mate, not only would it not resign, but it declined my draw offer.

Did your opponents act as consultants for the Fritz Nintendo project?

it gets worse
after i graduated i had no time for serious OTB chess so i entered Several rated USCF correspondence tournaments
now mind you..this was waaaay before the advent of PC engines
in fact there were only a few dedicated computers available and they played so weakly that you would only use them to see if they agreed with your move
if they did ..you went back to the drawing board and chose another move
anyway...the poor sportsmanship i experienced in correspondence play was beyond belief
in winning positions you would get accused of cheating because the handwriting on your recent post cards looked different then your earlier post cards
opponents would prolong playing in hopeless positions extending the play needlessly for weeks and months at a time..all the while hinting they would agree to end the game now...as a draw
here again ..you could petition to get an arbiter involved but who needed that hassle
for an intellectual game ..the opponents i faced would have been better suited taking up playing pool in smoke infested beer halls

i could go on and on but i guess we are far from the topic of your thread

I Digress Regards
Steve