Stirring stuff but....

[Tora]

I've founded a School for Worse Than Useless Scipii Generals whereby those with no redeeming traits whatsoever get sent out into the desert on field trips to sharpen their skills with brigands and the odd Egyptian patrol.
One such party consisting of two Armoured General units was caught by an Egyptian army of 1860 men - Nubian spearmen, desert cav, bowmen, skirmishers. The two Scipii generals, total 96, isolated and routed the enemy units one by one, killing 1654 for the loss of 29, the Egyptians making no attempt to stick together.
While it was one of my most satisfying Total War battles it begs the question, are Roman cavalry overpowerful or is the AI simply as much use as a chocolate teapot? Or both.

[Sandwichwarrior]

it begs the question, are Roman cavalry overpowerful or is the AI simply as much use as a chocolate teapot? Or both.

I'd say the AI gets the greater share of blame. While Roman Calvary is incredibly powerful, especially for a unit that appears so early in the game they aren't invincible and the AIs inability to win despite 20 to 1 odds should tell you all you need to know.

[Kaldhore]

I did the same with one of my oldest generals - I was waiting for him to die with all the useless retinues I had given him. Just sat in the spannish regions waiting to die.

Meanwhile the last spanniard forces marched toward Numantia and seeing my general - they attacked him. I was just about to clock off for the night and I thought - what the hell lets make a battle out of this. He had 29 men, against 1600 Bull warriors and Iberian Inf.

The battle lasted an hour and their army was routed, I had 7 men left killing 1300 (approx) of the spannish.

I too was kinda elated and grief stricken at the same time.

[The Stranger]

whatta f*ck, i knew cavalry was good, ohh wait you didn't modded your generals to 1 hitpoints, but wait 1600 bull warriors with pilum that pierce armour and 2 hitpoints, that's 3200 infantry, this is not good, CA failed again.

[Kaldhore]

Basically I did something similar to Tora. I moved away from the army - sometimes galloping sometimes trotting.

This split the army up considerably. I charged individual units before most of them could throw anything, withdrew - charged - withdrew - charged till rout then I purposefully didnt chase down the enemy. I waited then till I had routed just about all the units individually - and they had reform and come back. Then I charged a few of them in one long run - bingo chain rout.

[The Stranger]

CA really screwed this, in real life only the sarmatian knights from king Arthorius would do this

[Old Celt]

The AI is by no means great. But as much as the AI isn't great, it usually handles basic tasks reasonably. I think what happened in your battle was a case of unexpected circumstances yielding unexpected results. Most of us would not send 2 valuable generals to fight a full army without support. That particular circumstance was probably not playtested, and so, appropriate tactics although available, were not incorporated into the AI behavior. The AI probably looks at leadership units during assessment, but fails to combine that information with the number of units and troop strength.

[Red Harvest]

The AI is by no means great. But as much as the AI isn't great, it usually handles basic tasks reasonably. I think what happened in your battle was a case of unexpected circumstances yielding unexpected results. Most of us would not send 2 valuable generals to fight a full army without support. That particular circumstance was probably not playtested, and so, appropriate tactics although available, were not incorporated into the AI behavior. The AI probably looks at leadership units during assessment, but fails to combine that information with the number of units and troop strength.

I wish I could agree on this, but I've done so much 1v1 testing that the AI just makes me numb with its inability to reach even mildly logical conclusions. In vanilla 1v1 on the grassy flatland map, I can use the long range archers to cut down units like medium cav (Roman Cav.) Why? Because the AI saunters up slowly...by the time the remaining half reach my troops I charge into melee and kill them all. No fancy footwork, no attempt to deceive or exploit the AI.

So what does the AI do with roles reversed? It never fires a shot. It tries to charge up to my approaching horsemen, then at about mid range, it halts and starts to perform a firing sequence. Of course, my cavalry are in mid charge at that point, and the long range archers are now too close. As I begin to charge the archers have a change of plan and attempt to run away. Of course it is too late for that and they are cut down without inflicting any casualties

And with cav vs. the phalanx? Well..the AI tries to disengage its cav and charge again...when it is winning in melee and has gotten past the pikes. This leads to rapid routing of the cav who are getting poked in the rear haunches.

[Old Celt]

I don't doubt what you're saying, Red Harvest.

My point was that the logic probably doesn't even have correct assessment for circumstances like 1 or 2 units attacking an entire army because it wasn't something most people would do. That the AI does dumb things at many other times for other reasons is a certainty.

[RJV]

I don't doubt what you're saying, Red Harvest.

My point was that the logic probably doesn't even have correct assessment for circumstances like 1 or 2 units attacking an entire army because it wasn't something most people would do. That the AI does dumb things at many other times for other reasons is a certainty.

Perhaps sometimes attacking an entire army with a couple of units is somewhat unusual, but CA have been making these games for years, they should KNOW the different sorts of battle that take place. Having a very small army (less than 50 men) kill off 1300 opponents is just plain silly. Even if they'd hard coded something in there which let the little army have a good go at it before finally succumbing to the overwhelming odds would have been better.

It's just another example of frustration.

[Old Celt]

Perhaps sometimes attacking an entire army with a couple of units is somewhat unusual, but CA have been making these games for years, they should KNOW the different sorts of battle that take place. Having a very small army (less than 50 men) kill off 1300 opponents is just plain silly. Even if they'd hard coded something in there which let the little army have a good go at it before finally succumbing to the overwhelming odds would have been better.

It's just another example of frustration.

Oh yes I agree. My belief is that this is a case of the most extreme sort of AI weakness discovered so far, and that generally, the AI plays better. This weakness can be worked around like many others by simply not indulging in the behavior which exploits it (i.e. don't attack troops through the wooden walls with spears just because you can).

There is no question the AI could and should be improved. I think the only hope for that will be the expansion pack, because we will be paying for it. Assuming that members of CA staff read this forum, after some of the discussions about the lack of QA with definitive examples provided, it is almost incomprehensible to me how they could have released 1.2 with its plethora of "kindergarten" mistakes. Simple typos and reverses of greater than less than signs are just inexcusable. Even the cheapskates where I work would hire an intern to peruse the code for stupid errors like that. I'm mad about it, but when all is said and done, the game is still playable, just not nearly the finished product it should be. CA really should be ashamed at the quality of the 1.2 patch. I sure would like to read what CA staff has to say in response to this.