Perfect play doesn't exist.

It's about predicting your opponent, and no one will perfectly predict every action their opponent makes.

No.

Is it a draw when you ram two identical toy cars together?
No. You repeat until on of them flips over and exposes their soft underbelly. Gotta slip up some time.

Ah, but what if both cars were made to not flip over? + indestructibility? Then it would be a draw, yes? The analogy (and the assumption that goes with it) would be flawed.

@GoodKid, google Chinook (the program that solved checkers) [which IS a draw] {so, HAH}

It is not conceivable for two people to achieve 'perfect play' against one another. As explained before, it is all about anticipating the opponent's moves and acting accordingly in order to defeat the enemy as swiftly as possible. They cannot both do this. If there is no possibility of victory at the end, how can it be considered 'perfect play'?

I wish people like you wouldn't bother posting tbh.

To indulge in your troll:

Yes, in a game where both sides can see the board and have a set number of moves, there is a draw. In Warlight, you cannot see the entire board, and the number of moves (while still set) are far beyond checkers. Also, the board changes every game, so the picks and moves change per game. Each player gets 3 picks on the board, but since both players pick they don't always get 1 2 and 3, so "perfect" play could end up in a win or a loss due to one having superior picks in the same picking scheme. In fact this is more likely to happen than not. You'd have to have quite a special board for it to be a draw. From a purely AI standpoint, you'd have to have a scenario in which both parties get the same income, and are fighting on the same amount of borders at any given time, and are predicting each others orders each time with the same luck. Try to find that board if you would. I don't think it exists.

In top level play, you even see that games rarely drag on. One player picks better, or one player surprises the other with an interesting play.

Very tough question. I think no, its too assymmetrical, for example one person will get first pick and on certain maps that would be a game winning advantage.

Perfect play does exist, atleast in the sense that there is a nash equilibrium. Maybe not perfect play but certainly a perfect playing strategy. On some occasions that strategy may not make the most of errors made by the opponent, so on those turns i guess youd call it imperfect play? If one assumes perfect play to mean "even if i saw the opponents orders for that turn before i commited i still wouldnt want to change my orders because my orders are soooooo perfect" then in a rock paper scissors situation both players could not make perfect play, so the question is invalid for this definiton of perfect. I will assume OP is not asking an invalid question and therefore his definition of perfect play is a nash equilibrium strategy.

Also, obviously there is no such thing as a draw. Youre asking would there be a situation where a game continued indefinitely? Yes, in a situation where it was in neither players best interest to attack anywhere on the board. Can a situation like that happen? I imagine yes, its possible but not likely, it would be a perfect storm of equalish income, luck, positioning and very few points of contact, where you only meet once large stacks have already been established and no double borders.

Edited 5/13/2014 06:17:41

Strategic 1v1 is a long run draw if that's what you're asking, however you'll find it hard to implement perfect play every time.

As far as single gamesare a concern, a draw is possible:

http://warlight.net/MultiPlayer?GameID=5627905

I thought when you posted youd make some useful insight. That game wasnt a strat 1v1, with 16% luck eventually one side would get the upper hand and either win or sac the bonus and counterplay somewhere else. why do you think its a long run draw?

By long run draw I mean best players shouldn't exceed 50% win rate against each other given enough games, nothing more.

He is talking about in an individual game.

wow Sze impressive endurance to win that match =D. Theoretically I suppose a draw is possible on 1v1 strat but not likely. However, there is 1 board that ends in a draw quite often the scrabble board. If each player manages to get a corner or 2 players get two corners and meet in the middle it more often than not ends in a draw and players frequently vote to end.

Theoretically it would most likely become a draw and the game will last untill the warlight server dies.

Edited 5/13/2014 22:00:11

Here is the simplest answer: 16% Luck is involved in game play. If there is luck there can't be a tie.

And there is no possible exactly equal position. And there is no two players with exactly the same skill set. And no two players think exactly the same.

Run, what you say is nog true and dont make sence. He dit say [u]if both play perfect[/u], its a theoretic question. And it both players are both making evry turn perfect moves then it will prabebly be a draw.

I don`t see what so difficult about answering this question. It´s clearly a no.

In 1on1 strat settings you sometimes come into a position to have a 50% of winning by getting first move, while you still can win when you dont get it.

So, the first move luck not the 16% luck setting mentioned by TheCro is`n't critically in explaining why there can`t be a draw. (which doesnt mean its theoretically impossible)

If it`s not an individual game and a series of games, sze is of course right.

Edited 5/16/2014 08:18:59

A lot of games do work out to be roughly a draw in the early turns. When the two players meet, each has equal income and stacks, and neither can force an advantage. The reason it doesn't remain a draw, in my opinion, is the vast array of possible moves available by that point in the game. If the board were symmetrical, then a draw would be more likely to continue, but still unlikely. With a varied board (strat 1v1), there are clear pairings where some bonuses and board positions beat out others.

Then you add in the human variation of putting two extra troops here verses there, or attacking several areas verses one large stack, evaluating the potential to expand verses attacking, etc. Because the game is dynamic and your opponent's moves are not set, you can't know which moves are going to actually be better. You work with predictive models and probabilities, so there cannot be (without cheating) enough knowledge to actually make a "perfect move." Even if you select the 95% probable countermove, you aren't making a perfect move, just the best move given the board situation.

Lastly, 16% luck and the individual variation will eventually erode or build up one side in comparison with the other, and the game will no longer exist in a state of draw.

16% Luck is involved in game play. If there is luck there can't be a tie.
Ties (for individual hands) are definitely possible in Poker.

Even if we assume super-optimistic rational players, there is a point where, with equal armies, an attack, no matter how lucky, will lose more than the defense. This isn't at 0%, so a tie might be possible with luck. (16%, granted, is past that point.) Scrabble is a decent map for this. Assuming we tweak the bonus values a bit, we can construct a possible play that makes both players end up in the middle with equal income and armies. No player will attack because their attack will NOT beat defense. Tie.

Of course this requires a lot of armies.

Edited 5/16/2014 21:36:23