Flenser

I'm sure we can add a max_attack_value to cause instand attack. Should need some testing to get a good value. I already had something like this in my code 6 month ago. Was about 12000 army value as trigger to attack regardless of oppositions strength. Have to reevalute for WA.

<{POST_SNAPBACK}>

All the difficulty levels apart from Advanced (ie. Harder and Insane) have

max_army_size = xxxx

which forces the enemy to attack before it gets too powerful. Could be easily tweaked to ensure Harder and Insane attack when their armies are near full size.

Flenser

The AI might be weaker because it has just fought a big battle and come out badly, but if its economy is still better than the player's, then it is only a matter of time before it can build up again. In such cases it should not jeopardize its rebuilding effort by engaging in the field, even on a limited basis. If the enemy attacks the AI while it is rebuilding the AI obviously has no choice, but it will be nearer its production facilities and so more likely to win than if its army were out in the field taking down an LP.

Where the AI's economy is weaker than the player's AND its army is weaker than the players then you are right, the AI will be unlikely to attack ever, since the player should be building up quicker than the AI and it will never catch up. How
can the AI win in such cases? It needs to improve its economy.

Improvement is relative, so damaging the enemy economy == improving its own economy.

1. The AI could capture any free SPs.

2. We could force the AI to attack anyway, if it sees that (1) the player's army is getting bigger faster than the AI's army - ie. the longer it waits the harder it will be, (2) the enemy's economy is better than the AI's and (3) there are no free SPs. Might as well go out with a bang and, you never know, the player might fluff it and get beat

3. The AI could look at the location of the main enemy force, and pick the enemy SP/ LP which is nearest the AI's army but farthest from the main enemy force. Bonus points for several enemy SP/LPs in the one place. Then it goes and decaps it and retreats. Lather rinse repeat. This is akin to a skirmish, and is also very close to cheating, since the AI should not know where the player's army is.

4. The AI could try and decap all the enemy SP/ LPs at once, splitting its entire force up into small forces - each at least strong enough to take one LP down, hoping that it confuses the player and manages to decap some. It will lose a lot of its army, however.

5. The AI could invest in economy improving tech at the expense of other expenditure. Very risky.

What else could the AI do to recover a losing position?

Flenser

I suggest not using the skirmishplan as long as Flenser hasn't double checked it.

<{POST_SNAPBACK}>

Hmmm - I've played that bit to death and never had a CTD there. We will see if v.2 does it.

Flenser

I just tested it since I was curious.

Insane resource rate is  40% faster
Harder is 20% faster
Hard is even.
Standard is 20% slower
Easy is 70% slower.

To test, I disabled the AI, started a skirmish, sat for a while. Then checked out the resource score.

I guess Hard should use the Advanced AI for people who want an even game. Actually, why not let all of them use the Advanced AI? The resource limit on Easy should still make it relatively easy, right?

<{POST_SNAPBACK}>

That is quite a change there, Relic! Am I wrong or did it used to be like this?

We could make all difficulty levels play the same, but it is easier to tell everyone to pick the level that the Skirmish AI works best at - saves time and testing.

From these figures it looks as if Hard is the level we should be aiming for, since I don't like the idea of the AI cheating as regards resources. The Skirmish AI should be good enough on a level playing field.

Flenser

AFAIK only the Insane AI cheats with resources. Harder AI is the Insane AI minus the cheating.

Flenser