Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

More likely, those 2587 positions cover all possibilities.

There are other proofs like this - the 4 color theorem one, which was also reduced to a finite number of configurations which were manually colored.



But none of this explicit. This is at best an announcement of a potential proof. But it isn't a proof yet.

Also curious: in the end of the paper they talk about having "weakly solve" Othello... the paper overall reads really strangely.


"Weakly solving" a game is a technical term. If you have weakly solved a game, you can play perfectly (achieve the optimal result) when the game starts from its initial position. If you have strongly solved it, you can play perfectly starting from any position.


Sorry, I was unclear: I know what weakly solved means. What I find curious is that the title and abstract refer to "solved", and don't mention what they actually mean. To me "solved" would suggest "strongly solved". But perhaps equating "solved" with "weakly solved" is default in this area? Still, I would like expect an abstract to say something like that explicitly.

But given the overall state of that paper I think this is a side concern at best anyway.


> But perhaps equating "solved" with "weakly solved" is default in this area?

It is the default and all that matters.

This is one of the dangers of reading papers as a non-expert. You can dismiss or be wowed by something that is totally irrelevant.

They wrote the paper very much like the Checkers paper from Science 2007.


> To me "solved" would suggest "strongly solved". But perhaps equating "solved" with "weakly solved" is default in this area?

It's the default for all reasonable games - statespace is huge (i.e. tic-tac-toe is childsplay) and simple strategies don't exist (that'd make bad human game). You can't even iterate all positions - even less prove them all for one outcome.


They wrote about this in the introduction and in the context of algorithm 2.


"weakly solve" is not curious. Read from the start, they define this.


This looks like a George W "Mission Accomplished" moment.

- the game is obviously known to be a draw

- but we don't have computational power to enumerate that

- so test a bunch of game states

- and confirm that none of them are wins

- ???

- it's solved! Trust me!


The concept of "weakly solved" is not original to this effort. IIRC checkers was weakly solved for years but is now strongly solved.


And you can go further Heads Up (ie two player) Limit (ie you decide to raise or not, the size of the raise is fixed) Texas Hold 'Em (the style of Poker most played today) is essentially weakly solved.

The process used generates a statistical approximation and tells you how close it is to correct, in theory a perfect solution would beat this by that amount, in practice of course Poker is a game of chance, and so over any realistic game it wouldn't matter because the deviation from correctness they've computed is tiny. Could they make an even smaller deviation with more compute used? Sure, but why bother.

https://en.wikipedia.org/wiki/Cepheus_(poker_bot)

Cepheus is instructive also because some humans have played against this and believed they were outplaying it, which indicates there are real human poker players who misunderstand their own variance so much (and/or discount real variance from others so much) that they're completely unable to successfully rate their abilities.

If you lose 12Bb over 100 hands you are not, in fact, "winning except that it sometimes gets lucky". You're just losing, of course it sometimes gets lucky, that's how luck works, it's a game of chance.


Poker being one of my favorite hobbies (probably 300k hands played lifetime), it's wild how much variance matters. Like a 4BB/100 winrate (aka you win 4 big blinds every 100 hands) is very much an "I can be a professional" winrate.

You have a ~10% chance over 100k hands to be <0 dollars earned. Likewise, 10% of time time you'll make twice that. Poker is fascinating in that there are a ton of people who never actually hit the true law of big numbers hands and walk around thinking "I'll never be good enough to play at X level" or "I'm a poker god with big winnings" not knowing how good they really are.

Professional players do actually get in statistically significant sample sizes, but for amateur players, most don't get enough hands to really understand their skill level.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: