RTW v1.5 custom battle and replay test

[Puzz3D]

I set up a medium difficulty, large unit sizer, custom battle on the grassyflatland map with 4 hastati + 3 archers vs 4 hastati + 3 archers. There were no upgrades used. Fatigue is turned off since the AI doesn't know how to manage fatigue.

With the default deployment of 3 hastati in front, 3 archers behind them and the general's hastati behind the archers, I advanced to within shooting distance of the enemy hastati and archers which were in the same formation. My archers began shooting the enemy hastati while the AI archers switched to loose formation and then started shooting my hastati. Once my hastati were each reduced from 80 to about 68 men, the AI archers moved forward and started shooting my general's hastati. From what I have observed, the AI tries to distribute its archer kills evenly over all the hastati. My archer units also took losses, but I don't know if they were specifically targetted by the AI. While still shooting at my general, the AI hastati moved forward, threw their pila and charged two of my hastati. My hastati didn't throw their pila because I forgot to put them on fire-at-will. Their archers continued to shoot at my general's hastati and possibly my archers. One of the Ai's hastati had been severly weakened by my archers, and that one flanked my line trying to attack my archer on that end, but my archer routed it with arrow fire. My two engaged hastati lost and then the AI hastati turned on my remaining hatati line unit. I sent my general's hastati forward to fight, but the AI used fire arrows to rout my one remaining line hastati by firing into its flank, and my general's hastai eventually lost the melee with the AI's hastati. The AI also used fire arrows at this time to rout one of my archer units. It then advanced its hastati and drove off the units I was able to rally. The AI's archers never advanced further and stayed at a distance shooting for the entire battle which lasted about 10 minutes.

The AI played quite well in this test battle with the simple unit selection which was equivalent for both sides. The replay played back correctly.

[Red Harvest]

Tried the same test, used my standard formation of archers in front. Tried on offense and defense. Slaughtered the AI both ways while giving only a few instructions. (E.g. Group march order, followed by sending each hastati to attack another hastati when the AI hastati started to attack.) Didn't even bother to give follow up instructions as the AI army was beaten by then.

The AI simply cannot deal with missile units in front of the player's line. It won't engage in intelligent counter missile fire/attrition. On defense it will march its archers foward and try to reach range to shoot at the hastati while taking whithering fire. Sometimes it fires a volley or two at the achers before moving forward toward the hastati. Meanwhile, my archers shred its ranged units. Then it sends forward the hastati, who get shot up along the way and rout as my FAW pila slaughter them when they try to engage piecemeal. Result 5:1 casualty ratio.

On offense the AI marched the hastati forward first getting shot up for awhile while its archers took position. It then sent the hastati toward my line, so I gave the standard attack command and engaged 1 vs 1. Took some missile casualties this way, but easily won the melee. Result 3:1 casualty ratio. Would have been far worse for the AI, but I took a "hands off" approach for the most part, and didn't do any 2 vs. 1 attacks or slam into the flanks or rear, or even chase around archers so that they could be mown down.

I don't see that much has changed in this kind of test.

[Red Harvest]

Earlier I did a test of AI hastati captain with an archer auxilia vs. my triarii captain and roman archer--archer in front on defense. The AI marched it's longer range auxilia forward through my fire, getting slaughtered without stopping to use its superior range. My triarii charged the hastati when they approached...melee was in the hastati's favor. My archers routed the auxilia, suffering 3 casualties. When the triarii routed my archers finished off the hastati as they tried to attack.


Puzz3D,

Your previous comments in other threads about the AI failing to account for missile casualties still hold true. The AI often ignores them with its own missile troops so that it can target melee units. Not a smart move as its ranged units are killed off before they can cause much damage. I don't know if it is because they have lower cost or lower melee stats or what, but the AI does not consider them worth targeting directly. It concedes the missile fight before melee.

[Puzz3D]

Puzz3D,

Your previous comments in other threads about the AI failing to account for missile casualties still hold true. The AI often ignores them with its own missile troops so that it can target melee units. Not a smart move as its ranged units are killed off before they can cause much damage. I don't know if it is because they have lower cost or lower melee stats or what, but the AI does not consider them worth targeting directly. It concedes the missile fight before melee.

The AI is shooting at what it considers the most valuable enemy units. I think it's looking at the melee stats to make this decision and also the armor rating. If you have two hastati and turn one of them around, the AI will shoot the one which has its back turned to it. I think by increasing the ammo and effectiveness of archers, RTW has exposed weakness in how the AI handles ranged units. It was always there even back in STW, but there shooting a 500 koku warrior monk with a 300 koku archer made a lot of sense since the warrior monk was the most powerful unit and it had low armor. Archers in STW have 28 arrows and with 4 second reload it uses all of it's ammo in just under 2 minutes of continuous fire, and it can kill an entire warrior monk unit. This is the AI that has been ported over to RTW.

When I play against the AI, I don't use manual targeting. I use fire-at-will and let the AI decide what to target. That way I'm under the same restriction as the AI in regard to what gets targetted. Likewise, I use the default formation given to me at the start of the battle with melee infantry in a single line flanked by cavalry with ranged units set behind the melee units. My idea being that this is the formation with which the AI has been designed to deal. The one exception is that if the AI puts my triarii in the line, I'll place them in reserve near my general. So, for the most part I play the AI's game in an attempt to give the AI army the best chance of beating me. This is also why I turned fatigue off since the AI doesn't know how to manage fatigue. The AI doesn't know it's men are going to get tired if it moves them just like it doesn't know that enemy ranged units can inflict casualties.

Playing down to the AI is what I feel I have to do to get better battles. I don't increase the difficulty level. I restrict what I do.

[Orda Khan]

I think a saved Multiplayer battle is what is required to confirm that replays have been fixed. I noticed Custom battles replaying correctly before this patch

......Orda

[Puzz3D]

I think a saved Multiplayer battle is what is required to confirm that replays have been fixed. I noticed Custom battles replaying correctly before this patch

I'm not planning on testing multiplayer replays because the player who I used to do this with has uninstalled RTW. Both players would have to playback the replay and confirm that they both got the same kills, and that those kills matched the kills in the original battle. As wide a variety of units as possible should be used in the test.

[fallen851]

I have several replays where the AI skirmished against me quite heavily before advancing.

Give the AI more missile units than you, and it will burn up its missiles before advancing and closing for melee.

If you have more missile units (maybe the AI sees the pilum throwers as missile units?) they AI seems to advance no matter what.


Here is what I did.

Player- Greece
2 Spartan Hoplites (General imbedded in one unit)
1 Armored Hoplite
2 Hoplites
1 Militia Hoplite
1 Peltasts
1 Rhodian Slingers

Computer - Pontus
1 Heavy Chariot (General imbedded in this unit)
2 Cappodician (sorry spelling) Cav
5 Hillmen
6 Eastern Infantry
3 Peltasts
2 Persian Archers

Ok, so fine, I was trying to recreate a battle from the Persian wars... wrong time period, but the missle units worked well. The AI burned up all its missiles (mostly against the Spartans, but also a bit on the peltasts) before charging.


Try it.

[Red Harvest]

I've fought battles where the AI had several TIMES as many missile units as I did, and it still had trouble using archers/slingers and got the worst of it by far in missile play. It doesn't treat enemy ranged units with appropriate respect. And it spends far too much time moving, when it should be shooting.

A good skirmish/missile fight by the AI is a rarity.

[fallen851]

It doesn't treat enemy ranged units with appropriate respect.

This is very true. Perhaps the thinking was no need to kill missile troops, kill the melee troops, so when we close for melee we will have the edge?

Rome Total War has such a great graphics engine, but I really think it's a poor game. I could rant and rave for hours, but their are huge sections of the game that seem so cool, but are in fact useless, like spies, why even bother? And how about diplomacy, the only thing that matters is trade rights really. I just march a bunch of armies around, keep them well stocked with troops and kill everything. I don't care if the enemy knows how many Town Watch units I have in my capitial with their spies, they won't ever have an army that will get close to it...

I've decided to stop playing it, perhaps when EB comes out for 1.5 (it will be a long time...) I'll play again.

[Red Harvest]

Got another one last night in 1.5 in a campaign battle where the enemy archers would not shoot at all. It was really strange as they were positioned at the front of the enemy army. I had javs as ranged units...so I marched my entire army up to javelin range and began killing the archers. The archers stood there oblivious until the entire AI army began moving (and its "hidden" velites engaged.)

[Jambo]

Yeah it's a real shame the AI still has difficulty with ranged units. The problem isn't fixed with 1.6/1.5 and the AI still has a suicidal tendency to use its skirmishing foot units in melee situations when it clearly doesn't have to. Surprisingly the AI seems to be better with mounted skirmishing units...

[professorspatula]

The AI could definitely improve on its use of archers. I think it has improved somewhat with its javelin skirmishers though. I find myself having skirmisher duals much more, and the AI typically sends some infantry ahead of its skirmishers which makes my own skirmishers retreat, giving its own ones a chance to launch an uncontested volley of pointy spears. Hence I return the favour by sending my own infantry or cavalry forward to force their's back again. I've had a few battles now where much of the opening action in the battle is a series of skirmishing and pre-emptitive charges by both sides, and the main fight doesn't occur until most of the missiles are spent. The AI definitely seems more organised than before, though its still restricted by tight formation issues and there's not enough pro-activeness about it.

The AI is now basically at the level I'd expect it to be when the game was first released.

[Puzz3D]

The AI is now basically at the level I'd expect it to be when the game was first released.

That's a good point, and I feel the same way. Only now is the game deserving of the 92% rating it got upon release. Unfortunately, due the the massive number of problems that had to be addressed in the game over the past year, not all of the fine tuning that might have been accomplished could be done.

I'll continue to play down to the AI to make the game more interesting. At least the game is at a point now where I actually feel like playing it.

[Red Harvest]

I agree wholeheartedly with these comments. The game is just now reaching what I would consider a release state. This should have been the starting point for enhancement and fine tuning, NOT the endpoint.

If the game had been released in this state, I would have been happy with it, expecting patches to work out issues that were noted. I would have been ready to buy BI at release (as I did with MI and VI.)