Can the best human players still defeat the best computer programs?

well..i dont know which 1 is the best program, but the best program that i know of, i can beat it easily…cause those programs have limit to think…so at beginning of game, they play randomly …. when they start to play more seriously at end of game, then its 2 late for it to win since i start off with really good moves at beginning…alot of people that i know of start to play seriously around the last half of the game…but i always start my best moves (even thinking for 5 days) at 1st half to lead the game and hopefully winning the game :D ….

also…not to forget, these programs that u mention where created by a real person, so it has limited thinking based on the creator's logic …

so Yea, BEst human players can Still defeat best computer games easily :)

All AI is made by humans, and there are several games where computers are virtual unbeatable. So that argument makes no sense.

I think D&B has about the same complexity as Go. Sure some AI might be able to calculate the final x moves. But this just means the bigger the size of the board, the harder it is for a computer to win from a good human competitor.

-michael

Hi all from a lurker. I think both answers given here so far are inaccurate. Firstly we have to distinguish between “dots and boxes in general” (i.e. perhaps on a 50x50 board) and dots and boxes as played on the “standard” 5x5 board (by which I mean 25 squares, not 25 dots). I have very little experience in the general case, but on a 5x5 board I think I basically know the situation.

Firstly I think michael's comments are not right *when applied to the 5x5 case*—he might be right that 19x19 dots and boxes is similar to 19x19 go in complexity—I have no idea—but I think 5x5 D&B is much simpler than go (even 9x9 go), and I have good reason to believe so: I have seen a database by one of Berlekamp's students where a position with only four moves played has been strongly solved—in the sense that any position that can be reached from that four-moves-in-position has been evaluated. For me this means that in some sense we are only “one order of magnitude away” from having completely computer-analysed 5x5 dots and boxes.

Having said that, I also believe that some people can play 5x5 dots and boxes highly accurately and perhaps near-perfectly. I am almost certain of the following statement: once computers have completely strongly analysed the 5x5 case, they will prove that it is a 13-12 win to player 1, and if then some ch1 tournament games from this site are analysed by computer, the computer will be able to check that P1 played perfectly in many cases. Certainly, when I was active in this game, I would sometimes computer-analyse ch1 games from about move 14 (i.e. seven moves by each player) and would then note that P1 often played perfectly from this point onwards. In particular, humans are often capable of playing perfectly from move 14 onwards, even if (at least this was true in my case) they were not capable of _proving_ over the board that their moves were optimal. My experience is that, when one player is 13-12 up, it is often the case that either the position is “generic”, in which case P1 often seemed to have many moves that kept the score at 13-12, or the position is “tense”, where the tension is occurring in one critical area, and in these situations P1 would often have only one or two moves which kept the score at 13-12 ***but a good player would be able to sense the tension and easily find the winning move anyway***.

The answer by “bored of DB” seems to me to be incomprehensible and perhaps based on incorrect assumptions, as michael points out.

Perhaps I wasn't too clear on that, but I meant Go and D&B on the same boardsize. I'm pretty sure that if 5x5 Go hasn't been strongly solved, it wouldn't take much longer as 5x5 D&B. And there isn't any D&B AI that can beat an above average player on a 19x19 board.

Nice to see you back wc!

-michael

@michael: on the same board size you might well be right. I'm not “back” really, I just couldn't resist commenting on this thread. I've been lurking since 2009. Nice to see you back in ch1!

well…..on bigger sized board surely its easy to beat a computer in dB…and yea…5x5, 6x6 wud be hard…but its mangable in 6x6 cause a computer player has limited moves to think ahead…

so lets say the computer can only take into account the possible 10 moves that can be made ahead, and if the game has 30 moves overall, then the game is in my control for 1st 20 moves …..

and if there exists such program of 6x6 that can think for all 38 moves (i think), it wud need alot of pc memory and loads of timeeeeeeeee….so yes i dont think such program is possible yet and therefore, taking its limitations into account, i wud say for 6x6-8x8 a really good player can still win and for 9x9 and larger, it wud be extremely easy…the only way that computer cant be beaten is in 5x5, but there is always ties in that…so that does not matter…

and even tho my rating may show low here wccnanard, that does not mean i am “incomprehensible” and making incorrect assumptions…..been playing this game for 5 years and against best players, so these assumptions are based on experience which i believe are better than theortical thinking in this case

still, most of your words seems nonsense.

well gregorlo i cant blame ur lil brain not to be able to comprehend what i said..

obviously u have little experience of DB and would not even understand this whole forum anyways…so idk whats the point ur trying to make by showing me how much of logic u lack

maybe ur better off in streetsoccer forum, cause opinions coming from noobs like u is not even worth considering for

a preposition is a bad word to end a sentence with

lol idiot…u just finished ur sentence with a preposition also :P

:)

Gregorlo, it's not just the words that are nonsense, it's the order in which they're placed that is mostly meaningless too. I'm pretty sure I could program RoRoRo to make similar posts, in case anyone wants to see more like them.

no thanks, we already have some trolls that use a bot to make syntactically impressive sentences that on first glance seem to have a point. But if you read several in sequence you get only dizzy, as the semantic content is random.

(yes I know you were kidding) it's a typical sign of limited understanding to not understand a joke (or of limited joke quality)

Aganju, you should register the trademark JQ - if someone else hasn't already!

aganju get a life, stop posting in everyy threadd loll…did u even leave any thread untouched? ahaha :P

no matter how good the computers are, they can only be as good as the chickenlord

“if then some ch1 tournament games from this site are analysed by computer, the computer will be able to check that P1 played perfectly in many cases”

How did this statement stood the test of time? Are there games where the p1 played perfectly?

Does anyone have a candidate game (or small list) that I should check?

For best results, it'd be good if at least one of “my moves” gets played early….

I wonder if this summer is going to be the time when I start completing my analysis.

If William Fraser could just buy me another hard drive and spare one of his computers for a year or so….

Here is 2 games between very strong players won by player 1 where I'm not sure if they were played correctly:

986572 gej vs Feen 918455 Scot vs flipster

I think Feen had good chances to aim for 1 chain with an opening using this many edge moves and I think he may have won had he sacrificed in the bottom right corner. The game Scot vs flipster is just very weird if not the weirdest game between top players I've ever seen.

gej used to start with 1.b3 a lot and he was a very strong player so he might have played many good games you could check.

I looked over gej vs Feen and discovered the Feen could have won at one point.  (After gej played e4 on turn 15).

I looked over Scot vs flipster, but was unable to determine who was ahead for much of the game (Not until flipster played f11 on turn 20).

You humans must still be better than I am at some positions.

I also looked over Geoff Cameron vs The_Shark_c (1579639), but I see that player 2 was an older version of The Shark which played a less useful move on turn 4.  As a result, I can only verify that Geoff Cameron's moves were correct from move 7 on.

Interesting your analysis failing until 20th move!

I am curious about this game: http://www.littlegolem.net/jsp/game/game.jsp?gid=1619343 because I think 7ics played quite well and I wonder if p1 had the advantage all the time.

Sorry, your move is only played on turn 18.

I looked over observer vs 7ics and discovered that 7ics could have won at one point.  (After observer played c6 on turn 15.)  I was unable to determine who was ahead before that point and observer was ahead for the rest of the game after 7ics failed to capitalize.