opening outcome known? 4ir forum
25 replies. Last post: 2015-07-28Reply to this topic Return to forum
fhourplay at 2014-06-10
Does anybody know the outcome with perfect play of the game
4 4 4 4 4 4 4 4 (filling up a central column) ?
Play4Fun at 2014-06-15
Personnaly, i think it's draw. I've ever play 3 where the draw is easy to have and the win possible. Is it a quiz or for your knowledge? If it's a quiz, please let me know if i've wrong.
diego44 at 2014-06-15
Not that I have any ensured knowledge, but from my playing experience I can confirm it to be a draw.
As for Play4Funs questions, I'm interested in answers as well.
fhourplay at 2014-06-16
It's a quiz to validate the result of my exhaustive search, which indeed reports a draw…
diego44 at 2014-06-16
That's a pretty cool statement. May I ask if solving 8x8 is within reach now? Have you solved some positions with even less moves played than these 8?
fhourplay at 2014-06-16
Solving 8x8 might take a few years, unless someone gives me access to a supercomputer:-)
I have tried solving dozens of positions, from 8 ply upwards, which generally take many days,
while 12-ply positions vary from hours to days.
From these I have implied results for three 6-ply positions.
fhourplay at 2014-08-01
I have to revise my estimate downward, having made great progress in the past month and a half.
Solving 8x8 will only take a few months, rather than a few years.
diego44 at 2014-08-01
Nice. Are you referring to a strong solve or just a proof of second player's forced victory?
Will you publish your results once finished?
fhourplay at 2014-08-01
I'm referring to just the outcome. A second-player victory is not certain yet.
I will eventually publish my database of solved positions, but initially will just summarize
the top few plies, and perhaps some select variations of interest.
If you know of some particularly interesting (finished) game between top players,
where the optimality of moves is in doubt, then feel free to list it here and
I may subject it to analysis.
pedropajarito at 2014-08-01
Hadn't it been solved by victor allis in 1988?
ypercube at 2014-08-01
@pedropajarito, that was for the 6x7 size, not the 8x8.
Nelagend at 2014-10-22
Curious, have you found a1 or b1 (in normal chess notation, of course, not LG's odd stuff) to be significantly worse than c1 or d1 when you look at a principal branch? I suspect their early game trees may be narrower than those involving early center play.
fhourplay at 2014-10-24
No move can be worse than 1.d1 :-)
I'm currently (about halfway) analyzing c1, which looks just as bad.
Do you mean you expect more alpha beta pruning in the search of 1.a1 and 1.b1?
I won't be in a position to judge that. Those searches will benefit from already
solved positions through transposition, so it won't be an apples to apples comparison.
Nelagend at 2014-10-24
When I've played games out on paper starting from 1. A1 b1 or 1. A1 d1 especially, they seem to have a lot more early pitfalls for both players than those starting with, say, 1 D1 d2 2 E1 c1. I'd think your bot would eliminate more moves from ending in early tactical victories for one player or the other.
fhourplay at 2015-06-10
At the conclusion of the next Championship (#38), I plan to make available the fhourplay solver source code,a medium size opening library, and a list of all mistakes made in group 1 games.Barring the discovery of bugs in the program, 8x8 connect-4 can then be considered solved.
I believe a 9x9 size will remain impossible to solve for at least another decade, if not several,and offer more interesting endgames due to its odd height.
Recalcitrant at 2015-06-11
Very exciting! I can't wait for 9x9. Would it really take that long to solve though? All the better if it does.
mmKALLL at 2015-06-11
Wow, amazing! Will you publish a paper on the subject?
fhourplay at 2015-06-11
7x6 can be solved in a few minutes, while 8x8 takes on the order of a year of cpu time.That's about a factor 100,000 difference. The gap between 8x8 and 9x9 is only 17plies rather than 22, but with the odd height the game tree is expected to beless prunable, so solving 9x9 with the current approach is still expected totake over 10,000 cpu years…
I haven't decided yet if the 8x8 solution merits a new publication.Perhaps it depends on my ongoing experiments with new search algorithms…
MisterCat at 2015-06-12
Naively, I must opine -
I am under the impression that publishing results and solving games is done when it will bring the author some degree of fame, regard, notoriety, money, etc. 7x6 FiR is COMMERCIALLY available as 'connect-4', and is (was; has been) a world-renown game played (by kids) everywhere. Chess is famous throughout history and this is why computer results are regarded with such interest; same with Checkers, same with Go, same with Othello, and I could go on and on (like Poker). Hardly anybody plays Twixt, and I believe that it explains the lack of interest in putting the efforts into computer play and solutions; same with 8x8 FiR, and again, the same with many of the more obscure games of strategy; heck, even Dots and Boxes is not really in the public eye, despite Berlekamp's (obscure) book; same with Conway's Game of Life.
So what I'm saying is that I do not see anybody doing serious work on the little-known and little-played games we have here, including alternate sizes for FiR; if authors above DO, in fact, make some discoveries in this regard, I am certain that the players HERE (all few dozen of them) will be delighted to hear the results; but 'PUBLISH'?? Where, I would ask, and for whom? And, why??
I have a great time playing these FiR games, and so personally, I will be disappointed to see them solved to the point where the result is a foregone conclusion. This, for instance, happened to Qubic (3-D TicTacToe using a 4x4x4 board) - a game I loved to play, even as a schoolkid, and got reasonably good at; in fact, there used to be world championships. This game WAS well known, so the reasonably talented computer scientists and mathematicians took to cracking it, with success (forced win for the first player) and look now - nobody plays it anymore. How sad.
Boo-hoo, and meow.(hey, but maybe I'm wrong)
Christian K at 2015-06-12
Can 7x6 really be solved in a few minutes? Do you need any advanced tricks or just by brute force? If no advanced tricks are needed, it could be a good task for a programming class.
Loony at 2015-06-12
The version of Fhourstones that is available for public can solve 7x6 within minutes on my computer. It uses quite a few advanced tricks if you consider only brute force as basic.
ypercube at 2015-06-12
@MisterCat, I disagree. The same places where one can publish results about 7x6 Four in a row, will happily accept results about 8x8 or 9x9.
fhourplay at 2015-06-12
MisterCat may google for “solving connect-4 on medium board sizes” to find my 2008 publication in the ICGA Journal. The literature is full of articles solving less well known games.
BuilderQ at 2015-07-27
There was a paper a few years ago on solving infinite connect four, which as far as I know is much less widely played than 8x8. A question would be whether in solving 8x8 connect four you're using interesting new techniques, or just more-or-less the same ones that work for smaller board sizes.
What I would really like to have is some efficient algorithm to decide mxn connect four…
Odd heights won't necessarily make the endgame more interesting. 7x7 connect four, which I've played many games of, is very drawish.