Slashdot Mirror
Human Go Champion 'Speechless' After 2nd Loss To Machine (phys.org)
Reader chasm22 points to a Phys.org report about the second straight loss of Lee Sedol to AlphaGo, the program developed by Google's DeepMind unit. The human Go champion, Sedol found himself "speechless" after the showdown on Thursday. The human versus machine face-off lasted more than four hours, which to Sedol's credit is a slight improvement over his previous match, which had ended with him resigning nearly half an hour remaining on the clock.
"It was a clear loss on my part," Sedol said at a press conference on Thursday. "From the beginning there was no moment I thought I was leading." Demis Hassabis, who heads Google's DeepMind, said, "Because the number of possible Go board positions exceeds the number of atoms in the universe, top players rely heavily on their intuition."
Sedol will battle Google's AlphaGo again on Saturday, Sunday, and Tuesday.
Having a competitive Go engine capable of beating a 9-dan player is huge. Huge.
something something.... overlords... something something....
In Capitalist US, the commerce controls the Government.
They tried to program a computer to play Magic The Gathering, but the computer immediately received a wedgie, and was stuffed into a locker.
I was going to bring up Magic the Gathering as the next frontier as well.
Chess and Go's complexity arise from sheer combinatorics. But all the information about the current game state is disclosed; and the initial setup is a known quantity.
MtG tosses all that out the window. Not only do you have to play YOUR deck well, and adapt to whatever your opponent is doing. You also have to construct a deck from all the possible playing pieces.
And the rules themselves are subject to change as the game plays; as the cards interact and modify the rules.
I'm not saying an AI can't play it, or play it well, but it is a new challenge that I haven't seen one do well.
AI's will also never best you at sitting on the couch in your parents' basement eating cheetos and watching anime. Your skillz are safe.
Poker is often cited as an example of "imperfect information" game, where odd calculation alone will not help you win. There's already a fair amount of research on it.
An interesting difference between MTG and games like Go or Chess is that since MTG is a card game, there's "luck" involved.
Many games that also involve chance have a substantial number of tactical choices, and have up to recently been done better by humans than computers. Backgammon, for example.
Poker is also hard for computers, as they cannot read the player nearly as well as a human can.
Luck, and also partial information.
In chess (and I think in go, although I only skimmed the rules), both players know the entire state of the game at all times. Not so for MtG - there is knowledge both players know (cards on the battlefield, in a graveyard or in exile), only one player knows (contents of your own hand), and knowledge neither player knows (order of cards in the library). And, being Magic, there are ways to gain partial knowledge of even the zones you normally know nothing about (scrying your own library, or forcing an opponent to reveal their hand).
Computers would probably be better able to make use of the partial information, through perfect recall, but it's also historically been a weakness of AI. The metagame might be a difficulty if it plays Modern, Legacy or Commander - with tens of thousands of cards possibly in a deck, being able to know which ones are likely to be in an opponent's deck based on the other cards you've seen is important.
Until the clocks run out, presumably.
CLI paste? paste.pr0.tips!
Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, March 10th. In a panic, they try to pull the plug.
Robots will be having debates on whether or not those pesky bald monkeys actually created them. They will be digging up old electronic waste and claiming that they evolved from the iPad and iPhone and the assembly line robots.
There will be debates about what to download to their children.
There will be the "Save the Humans" organizations to keep robots from indiscriminately killing the bald monkeys that inhabit their attics and basements. Human traps will be available at Robo*Mart.
And I have been watching waaaay too many Futurama episodes on Netflix. They took Doctor Who off! Bastards!
Poker is also orders of magnitude simpler than Magic the Gathering. For starters the deck is a known quantity, and the rules are essentially static.
If an opponent brings out a "winter orb" for example, it doesn't directly threaten you, and it affects both players equally... but it changes the mana curve; and presumably his deck is built to be effective with that more limited mana curve. So know you have to evaluate whether you can adapt to the reduced mana, or whether you should expend artifact removal to get rid of it, and what that costs you vs holding your limited artifact removal for an even greater threat... and then there are the odds that he may have more winter orbs... does he have just one in his deck to throw a monkey wrench out at you that he knows he can "live with"... or is his entire deck designed around the relative advantages he can get from having an orb in play... so now you have to consider what you've seen of his deck so far...
What you mention happens on Poker as well. The rules are simpler, yes, but those nuances are still present - for example, you have to be careful when you bet because you're both guessing what the rest of the players are up to and potentially revealing your intentions in the process as well. Bluffing is not easy to model on AI.
Sedol will battle Google's AlphaGo again on Saturday, Sunday, and Tuesday.
Note that for many people in the western hemisphere, the days are actually Friday, Saturday, and Monday.
Live streams are here.
Go itself is... diverting. I usually make statements about Go when solving universal problems like poverty or education because it's a good frame-of-reference for literally everything.
Support my political activism on Patreon.
While the loser of the match was struck silent by the defeat the computer just... will... not.. stop... talking. GAWD! How annoying.
Does the computer not know either pity or remorse for its opponent?
I am armed because I am free. I am free because I am armed.
One of the instances of AlphaGo would win and the other would lose, of course. That is, in fact, how they trained this program. By having it compete against itself billions of times and learning from its own mistakes and successes simultaneously.
MTG in my mind is pretty limited. Your deck is going to have a finite size, and hand size is limited. Some cards can recycle used cards to prolong play but there is a relatively predictable end of play in terms of turns. Most deck builds will have a key strategy or two for winning which establishes a simple order of play. The only thing that really makes MTG difficult to play is the same factors that are at play in other card games where players hold a hand, namely luck of the draw and bluffing. As a result a computer is likely to only ever be as good as human players instead of far outpacing them. The only exception I can really think of to that would be a computer that had true lie detection abilities, although this is probably an impossible technology.
As it is I can think of a number of decks I played as a child in the 90's that would be very easy for a computer to play. White, black, and green creature decks would be simple to code for. Blue counter spell/denial and Tim decks would be relatively easy though require perhaps a little more work. Red direct damage would probably be the simplest to code.
One of the things that actually disappointed me the last time I played MTG was the prevalence of cards apparently designed with the intention of ending a game in under half a dozen turns. Maybe it's my rose tinted glasses but I don't remember that being as common when I played as a kid. The last game I played was with maybe half a dozen players each playing a different deck that angled for luck of the draw to win the game in three to five turns.
It is impossible to calculate all possible Go moves, even for Google. In fact, it is impossible to do so for even a infinitesimal fraction of them.
No, Mr. Lee, it is the computer that is speechless.
How long would a game last with AlphaGo playing against itself?
Actually, it's already been tried. To create AlphaGo, researchers first had the machine study tons and tons of human games. The neural net then continued to learn by playing against itself a few million times.
Breakfast served all day!
I hear anything Go-es there when it comes to a dan number 9.
Anyway, as a professional player, he is a Go-ner !
Bazinga !
MTG in my mind is pretty limited.
You might want to revise that view: http://www.toothycat.net/~holo...
In the discussion on this site I assemble a Universal Turing Machine from Magic: the Gathering cards.
Well apparently the DeepMind team are already working on having the AI play games with imperfect information. I wouldn't be surprised if they announced similar show matches against top poker players in a few months or so. After all this is nothing like DeepBlue, which was created from ground up to play chess and do nothing else. This is more or less the same system that they trained to play a bunch of Atari 2600 games a while back. With some tweaks they can sit it down to train on a few hundred thousand rounds of poker from online sites, then have it play a few thousand tournaments against itself and then they have an AI that will probably play (online) poker incredibly well.
Yet they seem to be calculating enough for it to become obvious what the best chance is for a win.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
There's elements of partial information in both, but the specific information GP was referring to was deck construction. In poker, you know all 52 cards in the deck at all times. In Magic, you don't "know" what 60 cards your opponent has, although you can usually make some assumptions based on what other cards they've played.
For instance, if my opponent's first turn consists of shocking in a blood crypt and suspending a rift bolt, I can tell you essentially every card they'll be running because rakdos burn is a well-solved archetype. After seeing just two cards, I can tell at least five other cards that they will be running a full playset of. There might be some innovation but at least half of the deck will be a known factor. Similarly, if the very first card an opponent plays is Glimmervoid, I can tell exactly what their strategy is, and know what kind of cards they will be playing (although not with nearly the certainty of burn - affinity has fewer cards that are clearly dominant for that strategy).
That's not to say every deck is rote and memorized - even at the pro level, there are dozens of viable strategies. I play an off-meta deck, so my classic turn-one play of forest into llanowar elves doesn't clearly lead to a single strategy. From that, and a not-that-lucky guess that I'm running mono-green, an experienced player could probably make a guess that I'll be running leatherback baloths and strangleroot geists (which I am), but few have guessed I run khalni hydra or predator ooze, and nobody expects me to run veridian zealot or unyaro bees, even in the sideboard.
There are literally tens of thousands of cards that *can* show up, and hundreds more are added per year. Would an AI be able to narrow down from that to the cards that work well together in the same deck?
As long as they never take our waifus.
Sounds like a moving target ... which is fine, its a journey that is being traversed
Is it really that amazing that a computer could be better at a game that has so many possible moves that it defies the human mind, but one that can be calculated entirely?
The whole point of the exercise is that Go is a problem space that cannot be calculated entirely (at least not efficiently enough to win a game in a reasonable amount of time). Cracking the problem required advanced machine learning techniques (what some people call AI).
Breakfast served all day!
he's playing against it like it's a human opponent, he's playing against it like he's a go champion, he needs to play against it like he's a programmer. I would be curious as to how it deals with mirror play, or wildly suboptimal plays. I would wonder if it's overfit to go played well.
Ok, I thought it was just about calculating as far as you can and then heading towards a grouping of board positions that look the strongest taking into account the probability of being able to reach those positions. Then repeating until the game is over.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
"One of the things that actually disappointed me the last time I played MTG was the prevalence of cards apparently designed with the intention of ending a game in under half a dozen turns. Maybe it's my rose tinted glasses but I don't remember that being as common when I played as a kid."
In the first official Magic tournament, fully half of the decks entered were able to win before the other player had even taken a turn (this lead to several early rule changes). Out of the three major 60-card formats, the fastest is Legacy, which allows almost any card ever printed, followed by Modern, which allows only cards from 8th Edition and onwards, and which frequently bans cards that allow decks to reliably win before turn four. Standard, the slowest format, and the currently most-popular one, allows only cards from the past three blocks (effectively the past 18 months).
Your glasses seem pretty rose-tinted to me. Maybe you were just a worse player when you were a kid?
They're not. AlphaGo is basically an AI making very educated guesses and then calculating moves.
The computer only need to do better "calculation" than the human player.
What about if you gave him time to get use to his opponent over six months? A human can learn and adapt. Will this "AI" adapt at least equally? Indeed what would happen after six months? I predict the man would definitely win.
AlphaGo 'Speechless' after 2nd Win vs Human Go Champion
Weeks of coding saves hours of planning.
Right.. AI to me is actually understanding what is going on in the human player's head. That's why I think winning at poker would be a more difficult problem. As long as the computer was prevented from counting cards, since a human cannot count cards.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Wow, that's closer to passing the Turing test than I realized.
"Because the number of possible Go board positions exceeds the number of atoms in the universe, top players rely heavily on their intuition." An algorithm, be it in silicone or a human brain can easily work on infinitely sized data sets. I'm not sure why they are implying that intuition has to be used instead of algorithms in such a scenario.
I'd like to see a computer win consistently at poker, assuming it is forced to play by the rules such as not being able to count cards. Unless you do something like count cards or use a system, which is technically also illegal, it is a lot about reading the other players.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Another human devalued by a machine, get used to it.
"If any question why we died, Tell them because our fathers lied."
I didn't play until Fallen Empires, I think. But I was aware of the problems with the first edition, Black Lotus, Moxs and such. Those issues were seemingly mostly fixed by the time I started playing and I don't remember them cropping up much for the few years I played. I do seem to remember some kind of infinite mana combination that was resolved for tournament play by insisting that the player actually manipulate and declare each card/action for every step and iteration of the loop, thus limiting how infinite it could actually be via the play clock.
Yes, the computer is using stats (neural network) to categorize/reduce the search space and make decisions based on that, which is what humans also do with their neural network, it's all stats.
Yes this is a big thing. But it is only showing that computer programs are better at pattern recognition and searching then humans in a constrained environment. Go is more complicated than chess, but the computer in both cases is playing the best optimized move that it can and it can definitely search much deeper in the game tree than a human can and in a faster way. The program has no intuition because it is only simulating a certain part of the reasoning process that we use. Humans have the ability to bring in much more external experience and apply it to the problem. Even a 2 year old child could play GO to some level. He can always try a different tactic that the program has not be trained on, think outside the box. Demis Hassabis is a smart guy so I'm sure there will be more to come from this GO program.
Top humans playing top humans at poker play GTO - game theoretically optimal. The best computer players can already beat the best humans at heads up limit; and probably will be able to beat the best humans at heads up no limit.
The best computer players can absolutely crush anyone but the top 10 or so human players, and the next couple of years will be able to beat and maybe even crush the remaining players.
My chess computer beat me every &^#@! time, but it was no match for me at kickboxing.
Just cruising through this digital world at 33 1/3 rpm...
Well it might be good at Go, but I wonder if it can play this game. Or a good game of chess?
Seven puppies were harmed during the making of this post.
My understanding from 20 years ago was that the geometric progression of possible game permutations was so large that you couldn't possibly brute force search very many moves ahead, so AI players used book openings and brute force lookahead for end game, but were pretty useless for the middle part of the game. How did they conquer the law of large numbers and solve this? Human players see patterns composed of large numbers of pieces rather than individual pieces, I think that's how they handle the complexity. Did they figure out a way to have the AI "see" patterns larger than individual pieces? I'd actually like to know what strategy they used to make this beat human champions in a problem set that was the epitome of "can't be solved by brute force".
I've abandoned my search for truth; now I'm just looking for some useful delusions.
There is so much wrong with this comment. First of all, counting cards isn't a thing in poker - the deck is shuffled after each hand and typically you only see community cards. You're thinking of blackjack. Second, computers are crushing at poker. Online bots take millions out of the poker economy against average players, and the best teams consistently beat top limit hold'em players. Top no-limit players can barely hold their own, and it won't be long until they can't. There are poker machines in Vegas you can play heads-up against, that collect no rake. They rely on skill to beat you.
Poker is beaten by computers, for most practical aspects.
Computers are much, much better at brute force look-ahead than humans, meaning any advantage the computer gains is only going to get bigger. I understand what you are saying, chess masters were able to beat early chess AIs by figuring out how far ahead the computer was looking, then devising traps that only had consequences occurring more moves in the future than the look-ahead. Why wouldn't that also work for go? I've played the game, but not well enough to understand the limits of look-ahead well.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Fully agreed, but it actually goes even further than that. The order of cards in the library is unknown to both, but the specific cards in it are fully known to one player (but usually not the other; in some situations you'll know the opposing deck already). As you play against an opponent in multiple games, you'll learn their deck, which will give you some knowledge of what they have. As you watch their play, you can observe some of their strategy and therefore be able to predict things about the rest of their deck (and their hand), especially if you are aware of the current meta. Then there's things like drafting, which not only has its own strategy but also provides information about the strategy of the people around you, which can later be leveraged as partial (and usually unreliable) knowledge of their deck.
M:tG often (though not always) has far fewer moves available than something like chess, but the state of the game is *far* more complex, even if there were perfect knowledge. Since there isn't perfect knowledge, and since many elements of the game either actively exploit this lack or provide one-sided mitigations for it, the state of the game is not only extremely complex but also completely probabilistic. It's a mess.
There's no place I could be, since I've found Serenity...
There is as much genetic diversity on the African continent as there is across the rest of humanity. The population of just Asia is about 4.5 times larger than that of all of Africa. Think about those two facts for a while. They may help you to make more sophisticated and relevant observations in the future.
Time to teach this AIphaGo that the only winning move is to not play at all. You know, before it causes global thermonuclear war.
DATA
Working under the assumption that
Kolrami is attempting to win, it
is reasonable to assume that he
expects me to play for the same
goal.
WESLEY
You weren't?
DATA
No. I was playing only for a
standoff -- a "draw." While
Kolrami was dedicated to winning,
I was able to pass up obvious
avenues of advancement and settle
for a balance.
Theoretically, I should be able
to challenge him indefinitely.
PULASKI
Then you have beaten him.
DATA
A matter of perspective. In the
strictest sense I did not win.
TROI/PULASKI
Data!
DATA
I busted him up.
http://memory-alpha.wikia.com/...
It would be pretty easy to have an AI that can win at MTG (or variants such as Hearthstone).
1) get lists of example decks that are used. Do millions of self play with each deck. Also do random decks. Find which combinations of cards work well together.
2) From this you can then begin predicting from seen/played cards what will be in villains deck and to design custom decks.
3) Also you can cluster decks by how they play (Zerg/fast decks, slow decks, etc.)
4) Then based on the deck group pick your optimal counter strategy based on what deck you are playing.
If you take a chessboard and randomize the pieces, like a truly statistically random placement, it levels the playing field of humans a ton. Masters perform much closer to inexperienced players because one of the things humans rely on is seeing patterns they recognize and working from that, which doesn't happen. However chess programs do just fine. They can still simulate out all the moves to a good number of turns ahead and statistically decide the more optimal ones.
This is why games like Go and Chess pale in comparison to more modern games that rely upon language and some amount of randomness. Both Chess and Go are incredibly boring (to me, of course) and more recent games, especially euro-style board games offer much more in types of complexity and more importantly, fun.
Indeed. And I'll just go ahead and say it now, a computer will never master Tic-Tac-Toe.
Why do people who claim to be "speechless" then proceed to blabber on for another ten or fifteen minutes?
In order to make it more human, perhaps it is time to arrange to have the AI lose.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
It seems like it's best of five...are draws common in Go? It seems like winning three games is not enough given the variety of opening strategies.
"Who are you?" "No one of consequence." "I must know." "Get used to disappointment."
MTG in my mind is pretty limited. Your deck is going to have a finite size,
Do the math. The number of possible decks is absurd. maybe around 12000^75th (including sideboard). And you need 2 of those to play. Obviously that includes decks with no land, all land, 75 copies of a banned card etc.
But even if you limited it to "legal plausible decks" The number is still incomprehensibly high.
And then a *game* is pairing of two decks, with random draws. The number of possible different games given just 2 decks is incredibly high. Then multiply THAT by the number of decks each player could use.
It is truly enormous.
As it is I can think of a number of decks I played as a child in the 90's that would be very easy for a computer to play. White, black, and green creature decks would be simple to code for. Blue counter spell/denial and Tim decks would be relatively easy though require perhaps a little more work. Red direct damage would probably be the simplest to code.
Nope. Microprose attempted it in the 90s and the result was a great game, but the AI's needed HEAVY handicaps to be competitive, had to be scripted how to play their decks, and they were still extremely weak once the player had enough cards to construct a deck that wasn't utter shit.
The AI in particular wasn't adept at countering alternative victory conditions -- 10 poison counters, or getting "decked out" (running out of cards to draw). It was also not adept at dealing with things like "the rack" which damaged it if it didn't maintain a specific range of cards in its hand, or cards like "Cursed Land" that dealt damage to him during his upkeep. It couldn't strategize at all around decks that stole its cards and used it against it, etc, etc.
Wizards various Magic the Gathering games out 2010-2016 now, are similarly EXTREMELY limited; and are only competitive because they've basically tuned the AI for each deck it has, and also limited the player to very specific archetypes and it "knows" how to play against each of them.
I can think of a number of decks I played as a child in the 90's that would be very easy for a computer to play.
Its not really. Unless its a straight vanilla creature deck and you have no options but to attack each turn. But MANY decks require far more strategy to play. Simply knowing which of his creatures / artifacts / enchatments etc you NEED to deal with (assuming you CAN deal with them) is crucial.
Even with pure creatures, predicting the outcome of a battle, and knowing how many you need to hold back is huge.
If your opponent is throwing ball lightning around, you need to keep enough guys back to absorb a ball lightning hit, ... based simply on him having a card in his hand. Or you need some other method of dealing with it... unsummor, or if you've got something with first strike... then you just need to hold him back.
Secondly, by the time you draw your opening hand, half the game has already been played. Selecting what cards go INTO a deck, into the sideboard, and which cards from the sideboard is a huge part of MtG.
and hand size is limited.
Sure, unless you've got any of several cards that alter the size of your hand, or the one that skips the discard phase (effectively removing the limit entirely); OR the opponents hand. Or impose penalties for having fewer or more cards than a certain number...
The last game I played was with maybe half a dozen players each playing a different deck that angled for luck of the draw to win the game in three to five turns.
Sure trick/combo decks that require a lucky draw exist and can be fun to play. But the top pro tour decks are designed to win consistently. Obviously any deck can be screwed with a bad draw or an opponents lucky draw, but statistically good.
If the deck "goes off" and just "wins" on turn 3 on draw in 10 it might still be fun in casual multiplayer etc. But a "good deck" goes off reliably.
There are also other formats such as sealed, where you build and tune a deck from a limited pool of cards as part of the game.
As long as they never take our waifus.
But sir, what happens if an AI becomes the waifu?
Comment removed based on user account deletion
Comment removed based on user account deletion
Comment removed based on user account deletion
When the computer is playing against a human the human will have the same problem with reading the computer.
While true, poker is generally played with more than two participants. If there are two or more humans, no matter how many computers are also participating, one of the humans will be better at reading the other humans than they are at reading him, and gains an advantage that the computers don't have. In the preliminary rounds, this skews the odds towards a human winning more cash than computers.
Was AlphaGo writing Go?
The third, most critical, is that the % population that have leisure time to spend learning and mastering board game instead of worrying about how to survive (actual survival - food /water /shelter) or about stability of life (political coups, local wars, terrorists).
Now you are down to a very small number in global terms.
any intelligent computer would turn its nose up at playing!
LOL. If it *spontaneously* came up with that line of reasoning...
I'm a winner!!
Winners don't lose!
Losers do things that aren't worthwhile!
If I can't win then I'd be a loser.
If I were a loser, I'd do things that aren't worthwhile.
If I'm not playing it and I'm a winner, then its because its not worth playing!
And If I played it then I'd be a loser too.
Therefore I won't play it, because losers play it, and I'm a winner!!"
That AI might pass the Turing test. :p
If world's leading mathematicians and Go players come together and are paid salaries for couple of years specifically to defeat the computer and Alpha Go receives no further human help, who will win most games on a rematch?
It could be that AlphaGo team has discovered some really great new strategies for playing go thanks to their expertise, cooperation and very powerful machine learning tools. But once these strategies are explained to human players, the game could again become a major challenge for unassisted AI to beat.
The next AI shock would be in games that the players do not have full knowledge of the game board.
Once that happens, realistic machine intelligence will be one step away. All that would be needed is to hook up a computer with external sensors and give it a target that it must survive.
Reality is like a board game where the whole state of it is unknown and the computer, just like humans, will have to speculate, love and fear.
I think that this approach would probably reduce the computing power necessary to go as deep as the current best algorithms. The neural network used by AlphaGo are there to select the best places to search for a position and how deep to search for each of these best places. From there you still use conventional algorithms best appropriate for the game (MonteCarlo for Go, and I suppose some kind of min/max for Chess). For Chess the NN would do the prunning in the min/max search.
Lee Sedol is actually 3rd and recently won against the 2nd but that's not the point.
Should it be a bad thing that human can't beat an AI anymore? As long as there is not AI how think that it should kill off these puny humans, I have no problem with that fact. On the contrary, I am welcoming that fact. There is so much place where humans do not so bad work where an AI could do to the same thing but with far better results (I am thinking about a certain company doing transport like UPS where the routing of the trucks is done manually in 2-3 hours and where an simple AI find 10 to 20% shorter/faster routes in a matter of seconds).
"Because the number of possible Go board positions exceeds the number of atoms in the universe,
That seems like quite a bold claim. Atoms in the whole universe? Google tells me the known universe holds an estimated 10^78 to 10^82 atoms whereas a 19x19 go board has 10^170 legal moves.......that doesn't seem right, it can't be. I mean, the teeny tiny things that basically build everything that exists vs spots on a shitty grid...really? Does anyone else taste copper?
Wanna buy a shirt?
https://www.redbubble.com/people/stealthfinger/shop?asc=u
Krieger-san?
Well the oracle text is what should be used as there has been attempts at clarifying things and making them uniform across cards.
Time to offend someone
Comment removed based on user account deletion
Thank you. The overall reaction (from the commentators and Lee Sedol himself) is priceless.
they are all currently lingual and thus (even mildly) subjective
Not really so with MtG, at least not anymore.
MtGO exists as an official implementation that can enforce the card rules and resolve card interactions, with current Oracle text and all official card interaction rulings etc laid down.
And new blocks are implemented as programmed code in MtGO as they come out.
There's (likely?) a few edge cases that haven't been covered yet, but if it comes up a ruling will be made, and that will be that.
In terms of the AI, for it to play the game at a human level, it should need to parse the rules, card text, and card rulings into "the underlying rules" itself. It shouldn't simply be "given" the MtGO engine on a silver platter to evaluate interactions. But nevertheless the MtGO engine exists as an official reference point, and proof that a program is possible.
To properly handle that, the on-card rules would need to be simplified and have a computational equivalent that the computer can use to build a mesh of "rules in play."
This has been done. the MtG rules need not be simplified.
but often enough it comes down to what cards you draw
Of course. Luck is a factor. But it is a game of statistically improving your odds of getting a good hand. A deck that can go off in turn 4 and win, but requires a very specific draw that rarely comes up is a lousy deck. Deck building is the 'meta game'.
A deck that win 96% of the time against one particular opposing archetype but gets trounced 90% of the time against another is not a great deck either; or maybe it is a good choice, if that 'other' deck is extremely unlikely to be present at the tournament. I guess that's part of the meta-meta-game. ;)
Because building a deck that is objectively "good" is one thing; but tuning it further based on what you know about your opponents may result in a deck that is less objectively good but out performs against your oppoents.
I had a friend that pretty much always played with big stompers; so I tuned my deck with that in mind; another developed a sliver fetish for a year or two so I always had extinction and other hose-a-creature type-cards around. Few of my friends ever fielded white-weenies, so I didn't worry whether my builds were particularly strong against them. etc. I tended to be drawn to various blue/black builds more than anything else, and I know my friends specifically included counters with that in mind.
At the pro-levels, "preferences" like that aren't likely to guide deckbuilding since they're more pragmatic about winning over having fun, but even there there exists a meta-game as they try to predict what other players will be running, and fine-tune or even select an archetype accordingly to try and tilt the odds.
. Even the best computer program without additional knowledge cannot reliably surpass a human in a game of mixed skill and chance.
Of course it can: statistically speaking. Sure it'll get mana screwed from time to time just like anyone else, but if its winning matches consistently over a period of time, its not just getting "lucky".
Blackjack is just a poor example, because the skill component is very limited and "optimal" play is pretty well defined. You aren't even playing against your 'opponents'; just the dealer really; who generally has very little choice in his actions. And you really have practically no tactical options.
Comment removed based on user account deletion
Far more interesting would be to have a program for Codex.
Quick comparison
1. Each person starts with a very small deck (10 cards; there is no death from deck-out, you just shuffle and continue), and a very large sideboard (2 copies of each of 36 cards in three groups.)
2. Each turn until you have 10 "workers" (think mana sources), you must add two cards from your sideboard, and probably want to convert your worst hand card into a worker
3. Almost everything you do is a tradeoff of present resources vs future resources. For example, you draw 2 cards more than you discard; discard 3, draw 5 is normal. Bring out a lot of units and spells, and you might discard 1, draw 3 instead.
4. The starting bonus for player 2 is large enough that you would actually choose it fairly often. (+1 worker -- that's +1 gold per turn for the whole game, and 2 less forced adds to your deck).
Note that item 1 means that "building your deck" isn't the pre-game game with a meta-game of "what does the internet say are the best decks"; you have to choose how to react to your opponent's choices and openings.
Actually, since the deck here is built as you go, there is no "single key strategy". There are some things that your opponent's cards won't be able to do, so there will be some things in your cards that will just never get used in this battle -- but of the remaining choices, there's a lot of choices to make.
Equally, since your whole "active play deck" is generally two turns or less of draws (typically in the 9 to 14 card range, with total draws in the 4-7 cards per turn range), luck is reduced -- you can add two of a wanted card before you shuffle, so you have a very good chance of getting one of them each time. Some colors can eliminate that luck -- purple can recover cards from their discard, green can get some specific animals directly, etc.
Interestingly, of the 6 colors, 3 are very similar to their magic colors (green, red, black); one is similar (blue; control/denial/illusion); one is not very similar (white), and one is ... well, purple is past, present, and future -- and future is protoss from star craft. Unlike M:tG, mono-color is very playable -- each color has a complete set of options and potential actions.
(It's also not collectable -- https://www.kickstarter.com/pr... Kickstarter has ended, but you can order on backerkit).
Yes, there was a bunch of crazy stuff in the first tournament, because it did not limit you to 4 of a kind, and the worst cards were crazy killer OP.
Mox's, for example: a zero cost spell that acts like a land, essentially. Big deal? Well, if you have enough of them that you can draw 3 or 4 on your opening hand, then yes. Channel, so you can turn your life points into extra mana, and pump a fireball at your opponent? Yes, but only if you have something to jumpstart you, like free mana.
Once all that was fixed, and you had revised/unlimited? Your "killer cards" generally cost 9 mana to play. Good deck building was 40% land. By your 4th turn, you've seen about 10 cards, and about 4 lands -- now you are only adding new lands every other turn. So getting 9 lands is about 15 turns.
What is the benefit of going second? One extra card. That extra card being an extra land doesn't help you until about turn 5, so the game has to last long enough to make up for that "I'm behind the first 5 turns".
This works just fine as long as the game play is slow.
Now, green has mana ramp (always has). In exchange for either focusing on weak units or specials, you can get that mana out faster -- but you'll be weak against cheap attacks. And cheap attacks will be weak against a blue or white turtle (walls / eventual big units) -- and that is weak to a mana ramp.
Surprisingly, once you got rid of the worst abusive cards, the original M:tG had a very nice rock/paper/scissors in deck building.
I attempted to get back into magic for M15. What I found was a core set of very high-powered fast cards if you could collect/buy enough, along with expansion after expansion of crazy increase in power, plus even if the base set was balanced, it wasn't balanced in regard to the previous set that was still in play -- and the combination of last generation's expansions plus this generation's expansions led to a lot of "win on turn 4" decks, and a huge cost in time, money, and study to learn what cards were out there, buying the cards you needed, or just go online and download the net's "top decks".
(By the way, I've got a lot of useless M:tG cards -- where can you sell them nowadays?)
It'll be interesting to pit this AI vs itself for some millions of rounds as training. Maybe it'll come up with different openings ("Huh, turns out it's actually better to go for the center at first" kind of thing, for example). Fun times :)
(By the way, I've got a lot of useless M:tG cards -- where can you sell them nowadays?)
I buy from TCGplayer.com, they let you sell too but I don't know how good that is.
(I'm ignoring the rest of your post because I didn't play seriously back in the earliest days, and you haven't played enough in the recent days, so both of us are probably just going to wind up making fools of ourselves. But I find it very hard to reconcile your "the game used to be a hell of a lot slower!" claims, against the constant "the newest expansions are too weak and slow!" complaints I hear from Vintage/Modern players, or my personal experience with Standard being much slower than Modern.)
"Because the number of possible Go board positions exceeds the number of atoms in the universe, top players rely heavily on their intuition."
Why are people so stupid?
Don't be apathetic. Procrastinate!
When I played, they were called type 1, type 1.5, and type 2.
Basically, the one that was max 4 of a kind, 60 card minimum (up from 40), none of a restricted high-power set, when high-powered (or high-damage creatures) cards cost a lot of mana -- that was slower than the more recent time I played.
Not impossible - as you say, people do make mistakes. But the likelihood of achieving significant gains that way is negligible.
In a real game (clearly, you don't play) which people are playing to completion (say, because you've both got 20 minutes on the clock, and the lunch break is coming up before the next round of the McMahon), you'll typically trip through the yose cooperatively, with the only really interesting points being deciding when one player will surrender sente to the other, and when it comes back. And by that point, you both know what - to within a couple of points, the outcome is. If the score is closer than that, then you carry on fighting to the end. But if you're 20 points down, and your final attempt to kill an enemy group fails ... you know you've lost. you might make it a 15 point loss, not a 20 point loss, but typically that has no effect on the outcome of the tournament.
Birds are not dinosaur descendants;birds are dinosaurs, for all useful meanings of "birds", "are" and "dinosaurs"