New Developer Blog: AI and Battles

[econ21]

The odd corollary to this is that I find CIV4 combat results to be too unpredictable.

Off-topic - I agree, there's many a time when I've lost a precious veteran Civ 4 unit against the odds. I'm not sure of the maths, but I think it may be because there is not much averaging going on - in TW, you have scores of men hacking away for a fair time and so if they have a 3% kill chance or whatever, these random outcomes are being averaged over quite a large number of repetitions. With civ, it's more one-on-one and while there are combat rounds, there's presumably some kind of momentum effect whereby one bad roll disadvantages you in later rolls.

On the other hand, I think Civ4 combat "works", particularly in getting a decent trade off between quality and quantity. I have happy memories of trying to fight Russian tanks without having discovered oil. I felt like the Wehrmacht encountering the KV-1 for the first time.

[Bob the Insane]

I remember once many moons ago playing Warhammer 40K with my Imperial Guard company (I really did have around 200 of the little guys all painted up) and we were fighting my friend's Bad Moon Orks...

This was first edition rules and he had made an Ork Walker which resembled a giant pair of mechanical legs with platform on top for a sqaud of orks...

Well he charged it up the centre of the battlefield and I had everyone shooting at it...

Well you have never seen some many 1s and 2s rolled in your life, everyone missed, every single guardsman missed the giant mechanic thing...

And the Heavy Lasers and Missile launcher every squad had should have reduced it to a smoking ruin!!

How is that for an impondererable... If it had happened in Dawn of War I would have been tempted to launch the PC out the window for "cheating"!

[Myrddraal]

That's because unfortunately Dawn of War isn't really a 'wargame'. It's just cool to see the guys move.

TW is a wargame. I like the random factors of the game, they provide the challenge. Imagine a chess computer game, once you've mastered the difficulty level, the game becomes extremely repetitive and boring as you beat the ai the same way again and again. Now imagine that once in a while a pawn doesn't get taken buy destroys your bishop; you now have to adapt to the situation, adjust your tactics and overcome the problem. Fun!

[Polemists]

True but let's talk about limits. It is unrealistic to think that they are going to put e nough code in a game that is going to change every single time you encounter it. Coders who make games do not seem at that stage yet regardless.

What you can hope for is a certain set of variables that will at least offer some variety. I look forward to the AI, I think new units provide new tactics, and the world seems more immersive then RTW. Then again I would advise you all to wait until the demo. You've got two weeks come monday, depending who you believe maybe more.


Regardless this is all speculation. You are judging a entire game on ONE play testers, ONE battle, with a few paragraphs of info. I'm sure once everyone plays it and test it they'll all be going crazy. Point remains though buying is a choice, it is a way to support or not to support your game. Those who enjoy the combat will buy, those who don't won't. Modders are welcome to try and change what they wish. I'm just saying don't judge a entire AI system on one post(blog) without testing it.

[Bob the Insane]

That's because unfortunately Dawn of War isn't really a 'wargame'. It's just cool to see the guys move.

TW is a wargame. I like the random factors of the game, they provide the challenge. Imagine a chess computer game, once you've mastered the difficulty level, the game becomes extremely repetitive and boring as you beat the ai the same way again and again. Now imagine that once in a while a pawn doesn't get taken buy destroys your bishop; you now have to adapt to the situation, adjust your tactics and overcome the problem. Fun!

Sorry, but I totally agree with you and that was the point I was trying to make, TW is a wargame like WH40K (but not the RTS DoW) and as such sometime your unit of gothic solders will all roll a 1 and miss and sometime sthe peasants will all roll a 6 and do okay...

Mind you peasant stats are so rediculous that I do not wonder if most of the time they really not have a something like 1% chance of making a kill...

I mean I tried this the other day; one unit of Cataphracts verses a full stack of peasants, no upgrades (huge units). Yes the Cataphracts won easily causing ome rediculous amount of casulaties 2000+ before the peasant army's morale completely collapsed. But the Cataphracts did take some casulaties so even with a attack of one for the peasants and a defense of 23 for the Cataphracts there was some some chance of the peasant's attacks succeeding in there...

Maybe it is a bonus of using huge units, the statistical oddities even out for huge units producing more predictable gameplay?

[SpencerH]

Sorry, but I totally agree with you and that was the point I was trying to make, TW is a wargame like WH40K (but not the RTS DoW) and as such sometime your unit of gothic solders will all roll a 1 and miss and sometime sthe peasants will all roll a 6 and do okay...

That was the problem with playing wargames with 1 six sided die. The smallest chance for an event to occur was 1/6 which is too high for what should be rarely occuring events.

We dont have that problem with PC games though. It would be easy to code for a morale or weapons upgrade (or something else) that is calculated during the battle setup with a probability of occurance of 1/1000 and that remains active during that battle.

[Myrddraal]

Regardless this is all speculation. You are judging a entire game on ONE play testers, ONE battle, with a few paragraphs of info.

But what else will we do to kill the time?

Seriously, everything in these forums and online cannot give a full picture of what the game's going to be like, but that's not going to stop us speculating.

[Puzz3D]

TW is a wargame. I like the random factors of the game, they provide the challenge. Imagine a chess computer game, once you've mastered the difficulty level, the game becomes extremely repetitive and boring as you beat the ai the same way again and again. Now imagine that once in a while a pawn doesn't get taken buy destroys your bishop; you now have to adapt to the situation, adjust your tactics and overcome the problem. Fun!

This demonstrates the conflict of interest between SP and MP. In MP, there is no AI which makes predicatable moves therefore you don't need 'imponderables' to make battles interesting. What you need is reasonably predictable results so that good moves aren't thrown back in your face as mistakes.

[Bob the Insane]

This demonstrates the conflict of interest between SP and MP. In MP, there is no AI which makes predicatable moves therefore you don't need 'imponderables' to make battles interesting. What you need is reasonably predictable results so that good moves aren't thrown back in your face as mistakes.

But surely the chance of something going awry is equal for both sides in the MP game?

Though I do see your point, no one wants to be accused of winning by luck...

But going back to my WH40K example, on the table top wargame, luck sometimes happens. This does not seem to have dulled the popularity of that franchise...

[Puzz3D]

But surely the chance of something going awry is equal for both sides in the MP game?

If it's a rare random event, it's not going to be evenly distributed.

[SpencerH]

I can see where it might be less useful in MP. Surely, it would be easy enough to code it as a selectable (or not) option.

[AussieGiant]

I'm not sure if I am even on the same planet as Puzz, Bob and Spencer, but....

IMO, the only difference that should be between SP and MP, is the use of AI or a human to control the opposition.

The mathematics of the game should be the same.

In the end you should have a series of statistics for all the units. Plus upgrades.

Then a series of modifiers for all the "variables".

ie Terrain, General's skill, high ground, cover, concealment, range, etc etc.

Then the mathematics should be calculated and a result given.

Using a 1-100 percentile range where the maximum is 2 or 99, you should still be able to have a 1% chance of critical failure or success...just to leave some uncertainty in :)

Fresh, Veteran, Gothic knights, charging downhill, in the middle of the day, on flat ground, against Green Peasants should result total in slaughter. But there is still a 1% chance that a few peasants could survive and after the initial charge there would still be a 1% chance of a successful hit by a peasant on a Gothic Knight.

For me war is uncertain...many Generals over the centuries have written that time and time again. In some instances these small chances can change the course of a battle or an entire conflict.

Therefore there should be some uncertainty.

If being a world class general was simply memorising unit stats and the variables that affect those statistics then we would all be experts if we could remember them or reference them fast enough.

When the odds are no where near 99% probable...and in reality they never are, especially if the opposition is even vaguely proficient....it should be possible, and especially so if the opposition has used the variables to the best of their ability, to have "unexpected results".

The trick of course is to get these "variables" balanced and accurate.

As an example, that is why table top military games, using a 100% range of probability can be very accurate simulations of WWII conflict. This can be applied to the Medieval period also.

I'm oversimplifying this but I thought it might be needed