|
Post by Rycchus on Sept 28, 2007 14:00:05 GMT -5
*slips Delayed Antispell up sleeve* ___ Let's say almost never, but it's worth pointing out that Warlocks doesn't have 64 choices per move either: there's never a reason to select a null gesture. That brings it down to 49. There's never a reason to select a half clap (assuming you control your own spellflow, and have all the choices available) - that brings it down to 37. I could mention the relative uselessness of stabs, but I'm going to move straight on to disruptions - Amnesia gives you just 1 choice, Fear 4, Maladroit 7 (discounting the null again), Charm 7, Paralysis 7. Chess has no equivalent, and the combinatorial explosion is thus much more potent - chess generally has 20-30 options available each move through most of the early game. Charm and para actually have 49 choices each still as you still input options for both hands. The output for each of them is 14, but the input is 49 (neither of which = 7). Also choices are slightly increased with regards to targets. Say I have DP and you hit me with Amnesia. I have at least three reasonable options - shield myself, amnesia myself, or amnesia you, and that's without any monsters in play.
|
|
|
Post by Rycchus on Sept 28, 2007 14:01:12 GMT -5
If anyone ever reaches the "Grammatica" title at its new and lofty postcount, I'm going to give them the secondary power of being able to strike down gratuitous abusers of language: those who uses "2" as "to" in a single-thread conversation where even the "to" is redundant. ...which just goes to show, I'd like to see a bot out-trash talk me :-D *exalts Slarty for his Grammar-Naziness* ;D
|
|
|
Post by Slartucker on Sept 28, 2007 15:23:18 GMT -5
I'm afraid I'm not a Grammar fascist of any persuasion; far from it, I am a staunch descriptivist. Typos, misspellings, and syntactic perversions are my friends, and I've never laughed harder than when my syntax professor had cummings quoted at him ("since feeling is first...").
That said, I am not afraid to stand up to ugliness and inutility in any form, and I have some idea of how easy it is for a mild-mannered forum to develop into a net slum.
|
|
|
Post by Dubber on Sept 28, 2007 19:32:39 GMT -5
... those who uses "2" ... Yep, ur no grammarian 4 shur In the afternoon, no less. No excuses at all
|
|
Derfel
Ronin Warlock
Did I Do That?
Troublemaker
Posts: 283
|
Post by Derfel on Sept 28, 2007 19:35:27 GMT -5
... those who uses "2" ... Yep, ur no grammarian 4 shur In the afternoon, no less. No excuses at all Priceless!
|
|
|
Post by dni on Sept 30, 2007 11:35:30 GMT -5
2 Slartucker: Grammatica is great. Why you think that in my messages '2' compares to 'to'? It's only a curve form of list bullet, and no more... By the way, in russian no that association at all. '2' will be 'dva' or 'dvoika' and there's no nearly pronounced words. 2 Dubber: u -> Some form of 'you right', if it not so clear. 2 taliesin: I think so too. However, i don't mean with my words of 'varying their play' as changing gestures. What you describe is a tactics, not even strategy. I mean if player will change his strategy and form of tactics from one game to another to fake the auto-teaching methods. It will be too vulnerable to player 'cause he can't keep in mind all the forms of game in one time... Undoubtedly. There's no task to write AI than beats all players. There's a task write any of AI player which at least comparable with average player. Minimax is fully predictable for smart player. Why good players knows that beginners have a chance to beat them? 'Cause they can make a 'unstandard' (reads: fool) move which will be unpredictable but also will disrupt long combination. Never heard of that version of chess. Funny... 1. e2-e4. d7-d5. 2. e4:d5. d5-d4. Pawn hunting, huh? Some relation it is. What's about bets in poker? Cool. What if it happens? Program hangs? I'll be fake that AI with null gesture then. Ok, ok... You wanna be right? Sure you're right. You wanna to find the real problem? Try to count variants of spell targetting and monster attacking. Let it be. You feel it? Simultaneous form shorten all spells by one move on factor of unexpectance. One-move variants (P, >) can't predict at all. You may only believe that it will be. What happens if this game will be sequential? All stabs, shields will be predictable. All other spells will be longer by one move. There will no variants of change spell-targetting: opponent will see anything. And more, and more. Hm-m... For example, can you use your tactics of change-targetting way of playing which you described before, in sequential game?
|
|
taliesin
Ronin Warlock
Grand Master
Posts: 156
|
Post by taliesin on Sept 30, 2007 18:27:15 GMT -5
I think so too. However, i don't mean with my words of 'varying their play' as changing gestures. What you describe is a tactics, not even strategy. I mean if player will change his strategy and form of tactics from one game to another to fake the auto-teaching methods. It will be too vulnerable to player 'cause he can't keep in mind all the forms of game in one time... Not at all. Sometimes in the past I've faced players who've got into the habit of ducking counters because I often shadowcast aggressive spells. I adjust and start firing aggressive spells into those non-counters, and their attempt to snatch an advantage ends in defeat. You just need to get a feel for what it's currently doing to be able to take advantage of it. It's not going to be changing its play within a single game. There's no task to write AI than beats all players. There's a task write any of AI player which at least comparable with average player. It's quite possible to write an AI that would beat all players - not necessarily in every game, but in a majority of them. It's just a lot of work. Minimax is fully predictable for smart player. Why good players knows that beginners have a chance to beat them? 'Cause they can make a 'unstandard' (reads: fool) move which will be unpredictable but also will disrupt long combination. Sorry, dude, this doesn't happen. A beginner will pretty much never win against a top player. A merely strong player will seldom win against a top player. A true beginner (of perhaps less than ten games) won't even stand a chance against a strong player - an 1800, say - unless that beginner has a prodigious logical talent, analytical fervour and knack for pattern-matching. Minimax works: it's the main strategic thread that top-level players have in common. When people make non-standard moves against a minimaxing player, they will gain at most a small advantage, and if they make the wrong non-standard move, they will lose the game. This is especially true if you make the wrong move during a well-understood sequence with limited escape routes. If the non-standard move were actually a good move, it would have been accounted for by the other player's strategy. I hate to say it, dude, but I don't think you understand the game very well. To this end it's a bit pointless addressing your assertions on strategy. We can throw around analogies all day, but you're in disagreement with myself, Slarty and ExDeath (who wrote a nice layman's description of minimax in another thread), and given that we all are over 2000 ELO and you're not, it's hard to see your view as anything but a naive understanding of the game. Never heard of that version of chess. Funny... 1. e2-e4. d7-d5. 2. e4:d5. d5-d4. Pawn hunting, huh? You'd need to stick in a rule or two to resolve that kind of conflict, but yeah. That's what you'd have. Some relation it is. What's about bets in poker? Vastly less impact on the game. You'd need the previous hand to impact the cards you got dealt and the allowable variation in the stakes to make poker anything like Warlocks, and even then, Warlocks is a game where you choose what to submit, you don't get your options handed to you. It's like a vastly more complex version of Rock-Paper-Scissors, not poker. Without Confusion, there is no random element in what you submit. Cool. What if it happens? Program hangs? I'll be fake that AI with null gesture then. Nope, the program does what chess AIs do when people do stupid moves they didn't calculate for - start calculating all over again on the basis that your opponent has done something stupid and you can now beat them down mercilessly. Seriously, this is very basic AI stuff here. You narrow down your statespace with heuristics - if a move of your opponent is absolutely guaranteed to hurt him, there's no point in even looking at the continuations until he makes it. You calculate your moves on the basis your opponent is playing smartly, and if he fails to do so you know you can punish him. You feel it? Simultaneous form shorten all spells by one move on factor of unexpectance. One-move variants (P, >) can't predict at all. You may only believe that it will be. What happens if this game will be sequential? All stabs, shields will be predictable. All other spells will be longer by one move. There will no variants of change spell-targetting: opponent will see anything. And more, and more. This is not a problem. Again, you misunderstand how AIs work. In an AI state-space search, you evaluate positions. Now, you cannot evaluate the positions in one move's time accurately without looking at what they lead to, so you evaluate the positions leading off them, and the positions leading off them, until you're looking several moves deep. The positions that look best next move are the ones that have the fewest bad continuations for you. Because (by minimax) you're always looking to reduce the worst continuations for yourself, it actually makes sense to play as though your adversary can predict your moves - so looking at the game as though it were sequential rather than simultaneous is actually correct. Now, we could write an ideal AI that merely selected the least worst next move which would beat most of the active players in Warlocks. Why? Because most players do not have sufficient positional understanding to compete with it. The first really dominant warlock of my experience was Pig, who played very safe and solid moves which planned for his opponents' attacks, and ruled over most of the leaderboard simply through better understanding of the game. He was predictable only if you understood the game as well as he did, and hence for most players he was not predictable at all. However, you may eventually have your AI run into players who understand the relative values of positions as well as it does, and they will be able to calculate the safest move from its perspective for themselves and take advantage of the AI's response. For this case, you need to have some randomness involved so that it works also to minimise its predictability, choosing from a small selection of "best" moves. Obviously it would only do this where predictability matters - if you've disrupted your opponent with a double charm and his spellflow is full of nulls, it doesn't matter if he can predict it or not - he can't stop it. Hm-m... For example, can you use your tactics of change-targetting way of playing which you described before, in sequential game? You change the targeting according to your algorithm to reduce your predictability. You do not, however, count on getting results from it. You continue to calculate as though the worst possible outcome were going to take place - if your opponent guesses wrong and you manage to damage him or kill his monster, you then have an unlooked-for bonus. I'm not sure this is worth arguing with you further, dni. I'm sorry, but you don't appear to understand either the reasons some players win vastly more often than others, and why the methods they use lend themselves to a classical AI approach, or indeed how a standard game AI actually works.
|
|
|
Post by ExDeath on Sept 30, 2007 22:14:54 GMT -5
When people make non-standard moves against a minimaxing player, they will gain at most a small advantage, and if they make the wrong non-standard move, they will lose the game. This is especially true if you make the wrong move during a well-understood sequence with limited escape routes. If the non-standard move were actually a good move, it would have been accounted for by the other player's strategy. I hate to say it, dude, but I don't think you understand the game very well. To this end it's a bit pointless addressing your assertions on strategy. We can throw around analogies all day, but you're in disagreement with myself, Slarty and ExDeath (who wrote a nice layman's description of minimax in another thread), and given that we all are over 2000 ELO and you're not, it's hard to see your view as anything but a naive understanding of the game. Hey, don't speak for me. From what I can tell, dni has a very good grasp of game theory. Great warlock player? Maybe not, but the foundation of your argument is rooted in opinion of general game theory and does not warrant personal "my Elo is higher than yours" attacks. I imagine there are many games dni would smoke you in. Anyway, it all comes down to play styles. Personally I'm much more afraid of certain low-rated players, far moreso than Toyotami, Tchichi, etc. They try things I have never prepared for or seen before. It generally costs them too much tempo to win the game, but they beat me more frequently than their expected rate in comparison to the stronger players. If I make a mistake, it's generally harder to recover, because they are playing to kill. My 4 or 5 losses within the past year have all come to unpredictable players who would not be considered to be in my league. The thing is, a weaker player tends to think "Ok, I'm going to play my way because it works for me." That's dangerous. When I face a stronger player, both of us think "Ok, I'm going to play the best way." And, of course, only one of us is right. I don't really know what's being argued here, but believe me when I say a poker AI is many times more difficult to create than a Warlocks AI. The reason poker AIs seem passable is that you have a large element of randomness, where the computer can play bad and still win, and seem "unpredictable" or "savvy". With a Warlocks AI of superior creation, the leaks would be more glaring, having no randomness to fill them up. Chess AIs, meanwhile, benefit from there being absolutely no psychological component to a perfect game of chess. Warlocks is not "solveable", and I would make sizeable ($50,000+) wagers that I could defeat any AI you may think has solved it. A master-level AI would be easy to create, but it would never beat a master.
|
|
|
Post by Slartucker on Sept 30, 2007 23:34:30 GMT -5
Warlocks is not "solveable", and I would make sizeable ($50,000+) wagers that I could defeat any AI you may think has solved it. A master-level AI would be easy to create, but it would never beat a master. That you would beat it the first time you played it, with no prior information about it? Or the first time you played it, with an exhaustive history of its games? Or that you would beat it eventually? The latter is obviously true, the former I doubt would be a sure thing for you, certainly not sure enough to risk $50k on. And I question the statement "A master-level AI would be easy to create" vehemently. As for my own opinion of the Dni-Taliesin thing: what I see when I read that conversation is Dni picking away with numerous small comments that are rational on the surface, but miss the point entirely, missing something that seems frustratingly obvious to those with a deeper understanding of the game. "Minimax is fully predictable," and so on. The Elo comment was clearly not meant to be a personal attack: it was basically saying: "Look, it doesn't seem like you understand the game well, and we're having a hard time communicating about it. I don't have the patience to see this through, so if you're going to contradict players with established sophisticated understandings of the game, you need a pretty argument for anyone to listen!" I do find it interesting how much of corrolary fields of knowledge influence our interpretations of the game. Taliesin with his chess and his judo, ExDeath with his poker and his success at being competitive, Yaron with his mathematical deductions, Zugzwang with his computational analysis, etc. What's my secret approach? Inappropriate analogy, of course...
|
|
|
Post by toyotami on Oct 1, 2007 1:45:41 GMT -5
Hey, you forgot "Toyotami with his fistful of gonads."
But seriously folks, i'm a creative genius. Give the AI a spark of life and I'll help her write shakespeare...given enough monkeys...wait...i have no right being on this thread...I apologize. Xade, say something stupid that has to do with computers.
|
|
taliesin
Ronin Warlock
Grand Master
Posts: 156
|
Post by taliesin on Oct 1, 2007 4:02:23 GMT -5
I wasn't. I was citing an earlier couple of descriptions of yours that are very much at odds with dni's argument: "And yes, my old style used to be high risk/high reward (leading with S/S does that), but as you get better you realize you don't want to put yourself in those situations because even if you're right 90% of the time, you still lose the other 10%. I'd rather win 100%. So, nowadays I plan out lines for hours a day on a huge whiteboard and that works for me." "Yes, it does come down to 50/50s, but not just guessing correctly so much as analyzing the risks involved with each scenario. The games would never be split because I think the better player always has at least a 66% edge, even if he is only very very slightly better." I think you're misreading what dni is advocating here: he said that minimax, i.e. the strategy you, Slarty and I employ, which is to minimise the worst outcomes achievable against our spellflow "is fully predictable against smart players". As I've pointed out, this is true solely if your positional understanding is equal to the person employing the strategy, and even then being able to predict their spellflow does not necessarily entail being able to beat it. From what I can tell, dni has a very good grasp of game theory. Great warlock player? Maybe not, but the foundation of your argument is rooted in opinion of general game theory and does not warrant personal "my Elo is higher than yours" attacks. I imagine there are many games dni would smoke you in. When it comes to assertions about the strategies that function well in Warlocks, neither dni nor myself is going to prove anything save by example. There's very little general game theory being discussed here. There are assertions being made that Warlocks does not fit the standard models that game AIs are designed around, and the roots of these assertions lie in the disclaiming of the value of the very same strategies that I employ, and that you claimed to employ in the text above. I don't really know what's being argued here, but believe me when I say a poker AI is many times more difficult to create than a Warlocks AI. Yes, yes and yes. And dni's argument was that because Warlocks involves simultaneous play, it is in effect a "random" game of bluff and psychology, similar to card games such as poker (which, ironically enough, given it's being held up as an example of a bluff-based game, is as much about understanding the odds as it is about bluff - wouldn't you agree?). Warlocks is not "solveable", and I would make sizeable ($50,000+) wagers that I could defeat any AI you may think has solved it. A master-level AI would be easy to create, but it would never beat a master. Solveable? No. However, you could in theory, and only in theory, create an AI which was optimal. By this I mean that it would evaluate every position perfectly and play for better position with sufficient randomness to eliminate predictability. Much as we flip a coin so that the guess we make on a 50-50 cannot be predicted, you could similarly randomise the gestures you make with weightings according to how many gesture pairs and targetings of your opponents they performed well against. Such an AI would be a superior player. It would still lose on occasion against players who had a similar level of positional understanding, but it would win more than it lost.
|
|
|
Post by xade on Oct 1, 2007 5:21:44 GMT -5
Hey, you forgot "Toyotami with his fistful of gonads." But seriously folks, i'm a creative genius. Give the AI a spark of life and I'll help her write shakespeare...given enough monkeys...wait...i have no right being on this thread...I apologize. Xade, say something stupid that has to do with computers. My computer tries to tell me *bing* each time I make a move against it in warlocks... Let's see it beat my 2 by 4... I believe the gesture is: |
|
|
|
Post by overmage on Oct 3, 2007 7:44:18 GMT -5
i lol'ed at that ^^
|
|
|
Post by dni on Oct 7, 2007 8:37:45 GMT -5
2 taliesin: So, it happens and you get me quite surprised. I'm sorry for misunderstanding your nature and manners. This shows me how the feeling of situation is important, not only in warlocks, but, much more, in life. I never expected that you, a person with high level of intelligence, may extend black-white perceptions from artificial game world to reality. My hopes of flexible discussion, not straight compete or argue, must broke after second 'False', but i don't matter of this and there's my mistake. So, taliesin, if you need to be a winner, rightful, toplist man, or any other, sure you can be him. But it's not my way. I don't want to move up by other's heads, i try to know and feel all i see, not compete and win it. I'm sorry for all i said in this thread that was painful for you. You're right, it's not worth arguing of me. 'Cause it's not worth to arguing at all. Not for me, not for you. The truth is in co-existance, not in competitions. Be heard of yourself. Be full of feelings. Sincerely yourth, dni.
|
|
|
Post by Slartucker on Oct 7, 2007 10:08:04 GMT -5
Dni, I think you misunderstand again. Frankly I thought Taliesin a better man than myself for being willing to have the conversation with you and take the time to explain things, something I never would have done. He comes across as vehement because he is vehement -- but vehement only in defense of truth.
My profile used to read: "This is a game where you can always learn something new." I attribute my own success in large part to my willingness to learn from others. That includes a variety of warlocks ranging from Rasteroid and dWb to Yaron and, yes, Taliesin. I always try to listen to others and see what they might have found out that I've missed. But sometimes, a fact is just a fact.
I would suggest that the language issue is, perhaps, the main reason this discussion went sour. And there definitely is a language issue. If your intake is as confused as your output is, it's not surprising you guys can't agree on an idea and are frustrated.
|
|