RoflCopter4
Posts: 36
Joined: 4/19/2016
Status: offline
|
quote:
ORIGINAL: sillyflower
Brian G seems to regard R losses as irrelevant - though I suppose his AAR may say different. Also, it is the strength of the opposing forces you are better off considering rather than losses. Capturing just 1 manpower centre in the summer of '41 and holding it until the end of the game will cost R about 6000 men during the game, and, I think, give G about 500 HIWIs. In my game vs Brian, I'm currently getting just over 3100 HIWIs per turn from captured baby factories, and the number increases every turn.
In any event,it is not simply a numbers game. The key factors are:
1 force ratios. 2.5 soviet to 1 german (ignore allies) means G is in for serious hard time but 2:1 is enough to starting the serious hurt process
2 unit quality ie morale and experience is more important still. That means building up guards and grinding G morale down by winning combats when and where you can ie don't wait for the force ratios to be where you want them.
3 distance to Berlin over time: does G have running room?
4 quality of your generalship is also huge! The AI is easier to sucker into traps than a human - or should be anyway  . I haven't played vs AI for years so I can't usefully comment on it, but the simpler a game is, the easier it for the AI to win (chess is classic example) but wite is at the high end of complexity.
I'm not entirely sure this is the correct way of looking at it. First, what makes something complex vs simple is really not at all obvious. And two, despite any level of complexity, what makes a game easy for a computer to do is how easily it can create a simple summation of a point in time during the game, and how confident it can be in what the next likely such point will look like.
Take, for example, Go. If you didn't know, it's a truly ancient game that is to this day extremely popular in China. It has, basically, two rules, with another couple of optional rules, and that's it. A game couldn't be simpler in theory. Put stones on the board and never have a series of stones end up without a single open square by at least one or you lose all of those stones (and points). And yet, despite this, Go is so monumentally complex in practice that computers could scarcely beat amateurs until only very recently. This is largely because there's just no obvious way to evaluate positions, and worsened by the fact that there are so many possibilities at any time. When you reduce Go to a tiny board, a computer can do it very easily, but Go is not played on a tiny board.
Although despite this, just last year Google invented a whopping supercomputer that beat the world champion. The software that runs it is so complicated that even it's writers don't fully understand it, because much of the code wrote itself. It's a great achievement but certainly a very difficult one. If a game's face value complexity is all we measure then this should have been easy to crack.
Chess on the other hand is in fact extremely complex by any measure. It has a fairly large ruleset (compared with Go), the number of possible games is also astronomical (although much less than Go), and the sheer volume of ink used in Chess literature is immense. It just so happens that it is susceptible to simple heuristics to give a valuation of any position, and that because there is usually only one or two good moves in any position (after the opening) it is relatively easy to just check them all until the one with the highest score is found. This is why chess engines are so relatively simple, having only a few hundred kilobytes of code.
War in the East is certainly complex too. There are a great number of rules, a whole hell of a lot going on at once, incalculable possibilities at any time, and to top it all off an unavoidable element of chance introduced by pseudorandom dice rolls. However, the things that are difficult to have a computer cope with are not the many rules, as with chess that's actually very easy to do. And it's not necessarily the element of chance, since computers are probably better equipped to deal with that than humans are. It's just math. And it's not the number of things that have to be done. Tediously moving hundreds of units is a momentary task for a computer. What is actually difficult is making plans, predicting the enemy's plans, setting overall goals, and coming up with creative ways to fulfill them. The long term conceptual stuff.
This was also a big problem with chess engines for a long time. In the past one could play as conservatively as possible, taking no risks whatsoever and never allowing things to get complicated during a game, slowly improving one's position while the computer flailed about not knowing what to do. They would see mostly the same score from many different possible moves and would have to pick one almost randomly. They couldn't make plans, they just crunched numbers. Garry Kasparov (world champion, either the best or the second best player who ever lived) famously used this technique during his first match with Deep Blue (IBM's supercomputer). He lost the first game spectacularly by getting into a complex position and loosing a battle of calculation. After that, he played as boringly as possible and did not lose another game. He just waited patiently for the computer to make a mistake and when it did he would exploit it. He won the match. A year later for the rematch IBM doubled the speed of the computer and spent ages trying to come up with algorithms to better calculate a score for the subtle things that one must look at when things are dull and quiet in chess. Deep Blue actually won that rematch, although mostly because Kasparov played terribly for some reason.
Anyway there's probably no actual point I'm making here beyond just saying that you're thinking of what AIs can and can't do wrongly. You can see that the AI in War in the East actually does a moderately good job at defending, and can take advantage of tactical opportunities on the offensive, but completely lacks the ability to do anything bigger in scope. It defends strongly everywhere, and fails to account that its human opponent will just concentrate en masse, blow through at a point of his choosing, and surround everything. It has no ability whatsoever to do that on the offensive either. Big picture stuff is what's especially hard to program.
|
Printable Version
How many losses prevent a Russian recovery - 
New Messages
No New Messages
Hot Topic w/ New Messages
Hot Topic w/o New Messages
Locked w/ New Messages
Locked w/o New Messages
Post New Thread