Possible explanation of bizarre phalanx AI

[econ21]

In vanilla RTW, one of the worst features of the battlefield AI is how it mishandles phalanxes. In the hands of the player, the phalanx formation is so strong it can walkover nearly all non-phalanx infantry when fighting head to head. (In the German campaign, my spearmen could wipe out Roman infantry with virtually no casualties). But with the AI, the phalanx lines seem to break up as they approach the player's battlefield and individual units meander off in different directions, allowing them to be fought piecemeal, flanked and defeated in detail.

I wonder if the explanation for this is that the AI phalanxes are looking for flanks or weaks spots, rather than just trying to march over the enemy battleline as the human would do? When MTW came out - I think it was after a patch or two - I was very impressed with the way the AI would handle its royal knights. They would often manage to hit a unit I had supposed was in my battleline just slightly in the flank and smash it. (This is probably my fault, as I tend to leave slight gaps between units rather than form a seamless line).

I noticed the same thing today fighting Gauls in RTR - they did terrible damage to my line by not attacking frontally, but verring off at the last minute and hitting other units to the side, slightly flanking them. It's kind of like English 19th century bayonet tactics - apparently, soldiers were trained not to stab the enemy in front of them, but the one to the right who was attacking their mate, because that would be unexpected. As I did not expect it from the Gauls, it was very effective - allowing many of my front line units to be rolled over. (The AI also used its general's heavy cavalry very well, supporting a flanking maneouvre that tied up my reserves with very adept repeated charges.)

This made me think - perhaps that's what explains the phalanx behaviour? The problem is, royal knights have the speed to get away with this and perhaps so do Gaullish infantry. But because phalanxes are so painfully slow, this tactic comes unstuck.

Anyone got any thoughts? (Next time I'll try more solid battlelines and see what happens.)

[Ahmose]

Good explaination. Now I understand how the phalanx suck in the hands of the AI. I imagine the CA team in thier working " ok this tactic to the close combat infantry , what about the phalanx infantry? " the team leader replay " just copy the tactic to all units , don't bothered me " :dizzy2: . Really , RTW needs more effort to become the supreme game. thx Simon Appleton for explaination.

[katank]

Good thought.

Problem is phalanx cohesion. When ordering a phalanx to attack a static phalanx, the attacker invariably loses due to lack of maintainance of phalanx mode.

I see enemy phalanxes half out of phalanx attacking my solid wall of spears obliquely and getting mauled, routed only to rally to get slaughtered again.

There should be an option to force the phalanx to march straight at enemy and maintain lowered spears regardless of anything.

[Spartiate]

I think that the barbarian factions should be able to train enmass every turn if their population allows for it but only historically accurate troops.The Gauls for instance did not have any trained troops per se............rather they had tough experienced warriors.........not soldiers.I would allow all factions to do this also but for the Romans i would limit this to militia troops.I haven't really thought this through yet and i know i have just made an unfair generalization of the Gallic units but i do feel that CA have given abilities to some nations troops that they never had.

[Red Harvest]

Yes, Simon, this has been discussed and perhaps even understood for quite awhile. Not much we can do about it, unfortunately. The AI does the same thing with other units, but it does not effect the others in the way it does phalanx units because the others are faster, and less facing dependent. Within some radius each unit tries to do some sort of "best match up". When lines meet the AI happily marches up and then rearranges its line at point blank range. Generals who did this historically almost always met disaster and were considered indecisive twits. It is insane to do this in the last 50 yards or so, but that is precisely what the AI does. If you try to match your units to the enemy a few yards before the AI reaches the critical distance, you get the same chaos in your army as well. This is part of the AI that left me speechless when I first figured it out. I just couldn't believe it was that poorly structured and out of sync with the rest of the combat engine. I suspect this AI "thought" process applies to many missile units too. Why else would elite archers march up to 50 to 70 yards before opening fire on heavy infantry? It is just mind numbingly inept and a waste of good troops. Incidentally, troops tend to switch to phalanx (from raised spears) at the same distance. So it seems to be a proximity formation check for all units as well.

A positive of RTW to MTW is "intercept courses" (although sometimes a pursuit or direct heading would be better.) Unfortunately, when this pursuit is done as melee occurs it often sends units off marching parallel to the line, so that they get hit on the flank. The intercept course magnifies lateral moves compared to MTW and this weakens the RTW AI. It could probably be fixed without getting rid of intercept courses, but is going to require an added layer of depth which CA seems unwilling to add at the moment. (I really, deeply want to be proven wrong on this projection when BI comes out.)

Now if I understood why skirmishers charge my heavy infantry and cavalry charge pikes...actually, I think I do. I pull my skirmishers/missile units back through the lines as they become threatened. The AI gets "target lock" and follows them blindly. I'm not trying to trick it or "game" it, I'm just trying to use traditional tactics. Again the AI is lacking a decision making layer that it sorely needs.