Saturday, July 19, 2008

Thoughts on "strategy" and "tactics"

Periodically the subject of "strategy and tactics" comes up on BGG. I roll these ideas around in my head (I hope the rattling doesn't bother anyone...much). They describe kinds of planning.

At the most basic level, everyone agrees that tactics are at a "smaller scale" than strategy. In the military "tactical" considerations refer to small numbers of troops, and "strategic" considerations refer to large numbers of troops. My rough rule of thumb for wargames is that tactical games include line-of-sight, and strategic games include supply lines. Many euros offer multiple "paths to victory," and these can be considered strategic choices; while implementing your strategy, you will make numerous smaller, "tactical" decisions. In Princes of Florence, you might pursue a building strategy, and you will have to decide on individual turns how much to pay for various items. You choice of strategy will inform your tactical decisions. To win with a building strategy, you probably need 3 builders, and you just pay whatever the price required.

Chess and go have a rich literature on planning. Chess books describe things like knight forks (where a single knight move attacks simultaneously attacks two pieces) as tactics, and things like connected passed pawns are strategic considerations. Chess books also call these features positional advantages, and at the higher chess levels the game is all about these. Books on chess "strategy" focus heavily on positional play, assuming the player has already mastered tactics. After all, controlling an empty file doesn't do you much good if you drop your queen to a knight fork.

The examples of chess and go show there is another idea floating around here. Sometimes a particular game state or position has a meaning that transcends move-by-move (aka tactical) analysis. It seems this "higher meaning" is related somehow to "strategy," but perhaps not.

Another example from chess. Kasparov was defeated Deeper Blue, and he later commented on one of his games in Time. He described the end game where Deeper Blue made a brilliant pawn sacrifice that shattered Kasparov's position. Kasparov knew he was in trouble, but he didn't see any forcing lines, so he accepted the sacrifice and took the pawn. Deeper Blue went on to win the game. In post-mortem, it turned out the game tree had shrunk to the point that DB was able to compute all possible moves to the end of the game. To DB, it wasn't a sacrifice at all; DB knew it was a game winning move.

It seems strategy and tactics are closely related to intelligence and the ways that humans and computers "think."

Interesting stuff, IMO.

Labels: , ,


At 2:12 PM, July 20, 2008, Blogger Carlos said...

One thing that has always frustrated me when playing computer games. The software seems to do poorly in large strategic planning. This is very true in the game "Hearts of Iron 2". If you are playing the Axis, you can rest assure that you will be safe from an Allied D-Day. Not because the software won't try it, but because it is bad at it.

As for the game of Go, I am told that because it is very strategic in nature that computers are not as good at Go as they are in Chess.

It is interesting that it took the mighty Deep Blue to beat Gary Kasparov. This is obviously not the game that one picks up at the local software store.

BTW, Hearts of Iron 2 is a great game. Much better that the first one. I have often thought that this would be an awesome multi-player game of Axis-Allies-Soviets with several people taking each side. For example one German player could take the East front, an other the Afika Corps, and another the Air force,... While a single player can do it all, it can sometimes be overwhelming.

At 2:18 PM, July 20, 2008, Blogger scott said...

any suggestion on something challenging yet not overly time consuming I could do with hearts of iron platinum?

At 7:46 PM, July 20, 2008, Blogger Ted Kostek said...

Deeper Blue included custom chips that were optimized for chess.

Back when I was "almost serious" about chess, I played ChessMaster 5000 (think they are up to 7000 by now).

Defeating Kasparov, arguably the best player ever, required dedicated hardware.

Defeating me required a $50 software package set on a very low setting with take-backs allowed...

On a completely different note, there's a back-of-the-envelope calculation that says the human brain runs at speed roughly equal to 2-5 pentaflops; a pentaflop (I've also seen it called a petaflop) is 10^15 floating point operations per second. Note this is different than clock speed, because it usually takes several clock cycles to do a floating point multiply.

The top end supercomputers have now achieved this speed.

Of course no one really understands how to program them except for highly specialized stuff like fluid dynamics.

Given time...who knows?


Post a Comment

<< Home