AI-AI Chit Combat

[pzgndr]

So. I think I've got this working, after several weeks of code research making notes and then several days of code work. See screenshot. This is a playtest with me as Spain watching what the AI players are doing. France and Prussia finally got into a shooting war, with France sieging Hanover. A Prussian corps tried to relieve the siege and got smacked by a Napoleon counter-attack. Nothing left except to imagine the battle itself...

I still need to verify a few things and clean up the code, but this appears to be a resolved issue. It was a challenge.

[Soosh]

That's great news. Thanks so much for keeping this game alive!

[pzgndr]

Some news to share. I think I have this new AI-AI combat working as expected. If you check out the log messages you'll note several debug messages included that I've been using to help understand what the AI is doing.

From the battle summary we see the at-start corps and factors involved, France with 28 attacking Prussia with 18. I'm using this strength information to calculate battle odds for the AI sides, which now helps them make less reckless chit selection; e.g., France isn't picking Esc Assault and Prussia isn't picking Withdraw at 1:1. The calculated losses for each round are being correctly reported in the summary. Prussia broke after Round 2 with morale -1.77, and France pursued. Prussia took 1 cav loss, again correctly reported in the summary.

Getting the combat screen battle messages for human-AI and AI-human battles correct as well as the summary messages took a bit of effort over the past couple of weeks. Long story short, the way the combat screen handles combat and the way AI-AI combat is handled off-screen is slightly different, and finding and fixing the discrepancies took a while.

I need to playtest more and see some AI-AI naval battles to verify all else is well, but so far so good. [:)]

[pzgndr]

I would appreciate some feedback regarding AI chit selection logics. As I mentioned, the program can calculate the battle odds and the AI can use this FOW information to its advantage, for both AI-AI combat and against human players who do not know the exact battle odds going into combat. Here's what I have coded so far to exclude certain AI chit selections:

Attack Chit Exclusions:
Outflank & Less Than 3 Corps
Outflank & Less Than 2:1
Esc Assault & Less Than 2:1
Probe & Greater Than 2:1

Defense Chit Exclusions:
Outflank & Less Than 3 Corps
Outflank & Greater Than 2:1
Esc Counter Attack & Greater Than 2:1
Withdraw & Less Than 2:1

The idea is to help weed out a few of the more reckless choices without being too restrictive. There are already AI withdraw exclusions from island areas and national capitals. Do these look reasonable enough or has anyone got suggestions?

[Daniel Amieiro]

If turkish player, never forego a escalated assault or assault (and their counterattack partnerts). It's a ok choice allways for turkish even with bad odds. (specialty if you have cavalry)

[pzgndr]

ORIGINAL: Daniel Amieiro
If turkish player, never forego a escalated assault or assault (and their counterattack partnerts).

I can work that in for a future update. I was hoping to get some more comments. But maybe players need to see some of these AI-AI combats first, and Matrix now has approval to release v1.22.02 as an official update.

[Daniel Amieiro]

One thing:
Is there in mind to have other combat sistem or the EIH combat system? (with envelope and things like that)
Same for fleet combat?

[pzgndr]

I have no plans for alternate combat systems. The complexity of messing with game options and .tpj data files is too risky. It's hard enough to keep this train on the tracks as it is!

FYI, I have implemented exceptions in v1.23.00 for Turkey selecting Esc Assault and Esc Counter Attack at 1:1 at least. Gives the AI some more aggressiveness without being too reckless. We'll see how it goes...

[Daniel Amieiro]

thanks pzgndr