A partial fix for suicidal AI generals

[Jambo]

Hi,

I think I may have come up with a partial fix for suicidal generals, particularly those involved in sieges where the gates are pre-opened (via spy) or there’s a breach already present in the wall. With the increase in spy numbers in 1.2 these battles have slowly become the bane of my Rome experience.

Normally under these situations, the AI force simply discards all their siege equipment and rushes the breach or pre-opened gate. Due to the fact that general units are cavalry they always reach the opening first and long before the infantry. Consequently they are the first to be slaughtered. Losing the general so pathetically and so early can ruin the fun factor of siege battles, as it makes defending them against the AI much easier than they already are.

So what have I done?

I’ve tinkered with the descr_formations_ai file. Within this file, all the AI formations are located and detailed – how they’re formed (in blocks), distance of the blocks relative to each other, etc – and this is the key factor. Essentially, what I’ve done is modded the starting position of the AI general block to move it further back behind the siege and handler groups (previously the siege and handler groups used to be the last blocks in the formation). In particular, I noticed there was a siege deployment formation and for this I’ve substantially pushed back the AI general (and also standard cavalry for that matter) relative to the rest of the army.

Where am I at now?

Well, I’ve run a few preliminary tests and I think I’ve reached a happy distance with the siege deployment one, which should mean that upon charging a pre-opened gate (opened by spy), the AI general and rest of the cavalry will arrive AFTER the infantry. This should prolong the battles somewhat and lessen the immediate suicide AI general, which we were seeing under vanilla.

What’s still to be done?

Plenty. I think there’s a wealth of opportunities and potential to optimize the behaviour of the AI by modding their basic formations. I’ve listed below some things which still need to be tested and some future ideas:

1. Standard non-siege AI formations. As I mentioned above, I also moved the AI general block back relative to the main force in standard non-siege formations. The impact of this modification in field battles still needs to be properly assessed, as I spent all my time testing the siege one.

2. Siege AI with pre-opened gates. There’s a possibility that widening the spacing between the infantry in the siege deployment formation could result in the AI force splitting and entering via more than just the single gate in front of the main force. You see, under these battle situations the AI effectively selects their entire force and simply charges the town plaza. On vanilla setup the closest entrance is the gate immediately in front and therefore they all try to squeeze through that single gate. Widening the formation may mean troops on the far edges will go through gates on the side of the town. It’s likely impossible, but that is the hope anyway…

So that’s where I’ve reached. I’ve included the text file on a thread over a twcenter for those interested in testing and helping. Please provide any feedback and results.

http://www.twcenter.net/forums/index...T&f=45&t=22511

Cheers

[therother]

Hmm, an interesting and ingenious idea. I will try to find time to do some testing and report my results.

[Jambo]

To test it should be fairly straight-forward, as this can be assessed alongside "normal" play. All you should look out for is the how the AI general operates in its new positions and compare it to what you're used to from the vanilla AI formations.