1.3 RTW Comments

[Puzz3D]

I repeated Red Harvest's test on the flat map, large units with 3 standard archers in a line 3 deep vs an AI controlled sacred band infantry and got the same results. I did not put the archers in skirmish and fired until the sacred band touched my center archer. The had 74 and 75 men left at that point. I didn't count the number of volleys.

Something interesting happened in the 3rd run. The sacred band got stuck about 100 meters or so away from the archers and couldn't advance (raising and lowering their pikes) so the archers got to fire all 7200 arrows. They got 22 kills using all the arrows. That averages out to 0.24 kills per 80 arrow volley. If you shoot into the back of the sacred band with a single archer, you can kill essentially the whole unit with 30 volleys. That's about 2.5 kills per volley.

When the AI uses ranged units as melee, I think this is another consequence of the AI being designed to charge into melee with units that are slightly weaker than the target unit. In STW, the AI does not make direct frontal attacks unless it can beat the unit it is attacking. If its unit is weaker, it always attempts to make an indirect attack. The AI makes attacking decisions using the unit stats in effect at that moment, so there is no doubt about having a combat advantage unless the AI is being tricked into thinking it has a combat advantage by some weighting factor. I can see several tendencies of the AI in battle which suggest a weighting is occuring, and we know that CA used weighting in the auto-resolve so they aren't above designing it into the battle AI to make things more "exciting" I suppose.

[Veresov]

[QUOTE=Red Harvest]I've fooled around with a couple of 1.3 RTW campaigns so far, fought maybe a dozen large battles on VH/VH. I'll start listing some comments, and perhaps add to this later. I haven't looked at stats changes yet.
Adding Items:
10. Archers still charge infantry at times without really firing. This became apparent when the AI threw its Cretan archers into my meleeing Hastati. I could see this if the Hastati had been alone on a flank, but my unit was backed by cav, etc. What a waste of a great missile unit.

This you can fix by adjusting down the secondary combat attack strength of the cretan archers. If you look at the knife skill, it is quite high. The AI looks at this attack skill and thinks this is a good melee unit. Just lower this strenght and the cretan archer will become an archer. This is fixable bug.

[Jambo]

Isn't the idea and testing of an army entirely of skirmishers or archers rather pointless? I mean, neither myself or the AI ever seem to have armies so constructed. Anyway, if I had a whole army of skirmishers or cretan archers it's conceivable I might use a few in melee...

Edit: Hmm, I wonder if BI (or 1.4) is different from 1.3 in this respect. In my 4 or so BI campaigns I don't recall any unit type acting particularly idiotic other than the frustrating bug of pila throwers not throwing their pila before attacking.

[Red Harvest]

This you can fix by adjusting down the secondary combat attack strength of the cretan archers. If you look at the knife skill, it is quite high. The AI looks at this attack skill and thinks this is a good melee unit. Just lower this strenght and the cretan archer will become an archer. This is fixable bug.

I'm not so sure and killing a unit's melee is not really a fix, it is crippling one part to get the unit to actually behave like a ranged unit. I experimented with this a lot in earlier versions or RTW, but can't recall what happened when I tried that type of test.

I've run tests with three Carthaginian javelinmen now, and they do the same as archers, line up three units deep and surge forward, then run before throwing. Three units deep is the worst possible combination: it leads to FF issues, and masking issues, and most importantly it prevents them from using flanking/rear fire. Plus they fatigue faster than the pursuer, and all at the same time, because they keep moving back and forth (since they are faster.) A worse deployment would be almost impossible to find. It gives up all the advantages of having fast, multiple ranged units.

What ends up happening is that if the melee unit pins one of the javs, the other two try to attack sequentially before they've fired all their javs. That is illogical, they should empty their supply before attacking. That is how skirmishers were meant to work. Avoid melee until the enemy has been depleted and can be broken. Even if the ranged unit has a very high melee attack, it should seek to deplete its ranged weapons as much as possible before engaging. It should only melee if it is a "sure thing," forced, or ammo depleted. In MTW the jav units had trouble skirmishing, but archers actually worked.

I did the 1 vs 1 cav test, equites vs. Balearic slingers. Predictable. The slingers walked up nearly the whole way, charging my cav without firing. Same as in 1.2. It's just a dumb thing to do, as a comparison of stats shows the Balearics have no business getting into a melee fight, especially head on. So the Balearics get slaughtered while inflicting about 10 casualties. Under my control I inflicted over three times as many casualties and nearly won the fight. Nothing special, I turned off skirmish and kept shooting until the equites hit my line.

Simply put, the AI is not inclined to make use of its ranged attack if it is moving.

Isn't there an additional problem about the AI not using Pila now? Hasn't been an issue for me so far as I've been playing vs. other non-pila cultures.

[Red Harvest]

Isn't the idea and testing of an army entirely of skirmishers or archers rather pointless? I mean, neither myself or the AI ever seem to have armies so constructed. Anyway, if I had a whole army of skirmishers or cretan archers it's conceivable I might use a few in melee...

Not really, I've had groups of ranged units (especially purchased Balearics as Carthage) get caught during overland transfer or after being dropped almost randomly before I lost my boats. I've also had to put together scratch forces with nothing other than a few javs and perhaps a unit of infantry. By the same token, I've seen all skirmisher/ranged unit type armies fielded by the AI. And in certain regions, the most likely composition of a brigand army is a bunch of skirmishers (like the Illyrians, etc.)

I even used three of the merc Numidian javelins (foot--the weakest unit in the game except for peasants and maybe townwatch), a family member, and a couple of hastati and a unit of Samnite spears to take Carthage which was held by two elephant units, a family member, and various infantry and skirmishers. Carthage sallied, and the "weighting" showed me to be heavily overmatched by about 3 to 1. I beat the main units as they sallied into my waiting skirmishers and infantry. Then I "skirmished" the elephants to exhaustion after they made a long trip around from a side exit. While they mauled my skirmishers and some hastati, the fatigue was too much, and after a lot of jav and pila vollies as well as melee, I routed them back into the city, and took the square. The skirmishers did most of the hard work.

However, that isn't the point. The point is trying to see what the AI thought was the proper thing to do with ranged units vs. melee and cavalry, and to figure out why it continues to charge its skirmishers into the main line. It didn't sit and camp, it didn't advance to the edge of effective range and begin firing. It didn't divide in two to keep the hapless enemy trapped between two fast units. The AI doesn't use ranged units as a screen most of the time.

And when I put my ranged units out front in classical period style, the AI charges at them like a red flag to a bull. I have to back them closely with spearmen because I know what is going to happen with them out front. The spearmen don't deter the AI, but they do make short work of any pursuing cav.

[screwtype]

I guess that's the difference. I don't take my best general and steamroller the AI.

I got two stars just from Autocalcs! ROFLMAO.

[screwtype]

Fatigue is a big factor for the AI. When it has multiple armies, it doesn't wait to connect its forces and attack as a combined army. This leads to defeat in detail of fatigued clumps. Fatigue saps morale and it reduces kill rate as well as running speed. So when the fatigued army reaches the player's line it ends up routing easily. (Many units don't even reach the line.) My units behave the same way if I treat them like that.

Okay, but the problem is not just fatigue, it's lack of coordination between the AI armies. They need to start closer together or something. These sally battles are really anticlimactic.

[screwtype]

I may do what Jambo suggesed and turn off fatigue to help the AI. This will actually just bring the AI's units onto an equal footing with my units in erms of fatigue since fatigue is never an issue for my units because I play with no timer and can rest as much as necessary.

Yes, it's true that fatigue is not much of an issue for your own units. It's not hard to manage them until they are fresh. Heck, even when they are "exhausted" they still seem to perform pretty well at times.

I didn't realize fatigue was such an issue for the AI. Maybe I'll try turning off fatigue too, although I'm reluctant to do so, because fatigue should make the game more challenging, not less. But from what you are saying, this looks like another design flaw.

[screwtype]

No, it isn't presumptuous, the AI sucks with standard armies too, this is confirmation of what I see in battle. This is a simplified test. If the AI can't handle the force properly in the simplest test, it isn't going to do much better on the field. Hence, the AI charging it's Cretan archers into my Hastati without firing a shot. It is the same problem we had with the AI before in RTW. In fact, the test gets right to the heart of the matter. It's not like I've run one test and made the conclusion. I did the same sort of thing in 1.1, 1.2, and in MANY, MANY battles. I ran the test to try to understand what I was seeing in battle.

The RTW AI can't use missile units with effect. There are a host of issues about that and some of them clearly reveal that those doing the AI design didn't figure out a way to render the skirmishing of the time. Heck, it even showed in the scripting of the demo.

Examples: Look at default army formations, the skirmishers are often in the rear, bass ackwards.

I've not yet done a javelin test of the same, but in a recent battle vs. a nearly all javelin army (with rebel bodyguard general) it used its javelinmen as melee vs. my infantry. It didn't skirmish and try to inflict maximum casualties or disrupt me before being forced to engage.

It comes down to this: Do ranged units use their weapons effectively under AI control? The answer is a resounding "No!" They still close to skirmish range, then go "Oh crap! What am I doing here?" I haven't retested in 1.3, but in previous versions they did the same thing vs. CAVALRY 1 vs. 1.

That's bad. It makes you wonder how they could miss such an obvious problem.

In the previous games, missile units would sometimes appear to choose melee over ranged fire at critical moments. But if they can't replicate that behaviour in RTW/BI for some reason, it would be better if they just made it that all AI controlled missile units must fire all their missiles before engaging in melee.

[IceTorque]

Bridge battles in 1.3 are worse with the AI's archers drowning themselves,
In one battle i ordered my general and one cav unit to chase the fleeing army across the bridge my general went straight onto the bridge but my cav unit ran to the right of the bridge and drowned themselves with only a handful
surviving.
Also when training cav units i am getting family members instead of the cav unit but these family members do not appear on the family tree, as i am playing a modded version could anyone confirm this in vanilla ?

[Red Harvest]

Did a few more tests 1vs1: Some equites vs. vanilla slingers, vanilla archers, and forester warbands, and dacian archer warbands. Their are a few patterns emerging:
1. High melee archers like the foresters seem undecided. They will pause and stand well within their range, rarely firing, mostly just standing to receive the charge. They will win in melee thanks to their spears and high attack, but a wide horse formation on the charge can inflict very high casualties on them. Still it's a waste not to make the most of the high missile attack and range. When I play as the foresters side, the cav don't even reach my lines most of the time, and when they do they are easily beaten.
2. Most mid or low melee/moderate missile attack archers will stop at about the limit of their range, then restart walking. Sometimes they stop and fire a few rounds. Once, one stopped at the very end of its range and let loose quite a few vollies causing me heavy losses before melee. (Never could get it to do that again--I think it happened because I used 3x and didn't go back to 1x before they hit the outside of the range.) Sometimes they don't fire at all and wait for the charge.
3. Low missile attack/low melee slingers advance, but never try to shoot, they stop at about their range, then turn to run. Occasionally, they kill 1 or 2 mounted men in melee. Other than that, they are dogfood.

AND THE MOST STUNNING/AMUSING OF THE TESTS...an accident...I took the field with 1 Roman archer vs 1 Roman archer by mistake:
My counterpart walked up through a barrage of fire. He didn't run, he didn't stop at range. He just walked up to within about 40 yards being massacred. After losing about 40% of his force, he decided to go to loose spacing, then walk back to a range of about 70 yards, and trade vollies. By the time he reached that point he had less than half his force left. After a few more vollies it ended with all but 6 of his guys dead. I "suffered" 4 casualties out of 81.

[Red Harvest]

Wanted to add a positive:
15. I've been listening to the speeches again. They are far more varied and colorful than when I started playing. There are several that I've never heard mentioned before (I had heard quite a few zingers, and some related here.) Something about slapping an ass (as in donkey I presume) comes to mind.

[screwtype]

Yeah I've heard some new speeches. But I'm not sure if they are really new because I've been playing the Julii for the first time, and maybe they always had different speeches. I always played the Brutii before.

Pity about the ranged units though. They sound just as broken as before.

[Garvanko]

There are a couple of new variations in the speeches, I believe.

[Jambo]

Well, I'm not sure how much to read into all this. You may have a few skirmisher-only battles, etc, in your campaigns, but they are extremely rare in mine.

I'd be more interested to hear results of tests between a collection of units which are more likely to be seen on the campaign map, e.g. how about a few archers, a few skirmishers, a warband or two and a general? If the missile units still behave ridiculously then, then I'm sure CA might be more inclined to take these results seriously.

As it stands a 1 vs 1 test isn't going to make too much of an impression as it puts the AI in an unusual position which it wouldn't commonly find itself. The 1 missile vs 1 cavalry clearly suggests that to avoid being ran down whilst trying to skirmish the cavalry, the AI has switched the missile unit's skirmish (and most likely FAW) options off. Whilst this seems to happen rarely in battles, I do actually remember seeing this myself (possibly at the end of a large battle). The reason I didn't think anything bad from this is probably because the missile unit had behaved normally up until the point it had been targeted by the cavalry at the end of the battle. NB I'd need to test this more to be absolutely sure as my memory of the exact situation isn't that great.

Judging from the various abilities in the descr_unit file, I'm certain that skirmish and FAW are mutually inclusive and sadly inseparable. e.g. the "thrown" ability means both FAW and skirmish are "on" by default and without it both are off by default. One of things I'd greatly desire as a modder is the ability to preset units with FAW but not skirmish (e.g. for Plumbutari say), something which at the moment isn't possible. :/

[Puzz3D]

I tried Red Harvest's test of 3 AI archers vs 3 hastati (two in front with the general behind) on the flat map. The archers move into a column, but individually stay in 3 deep formation. They advance into firing range of all 3 hastati and start shooting. The front two archers are overlapped, and the general remains some distance behind. The AI archers distribute their fire evenly amoung the 3 hastati so that they all suffer about the same kills. Once my hastati are each reduced to about 40 men (half strength) the archers charge them frontally, but are routed very quickly in the melee. I think this result indicates that the AI charges because it thinks it will win the melee. It actually isn't even close to being able to win the melee with 240 archers against 120 hastati.

I repeated the test this time with my hastati 3 abreast. The archers moved on top of each other and set up further out because it could target all 3 hastati from further away. One archer ran out of arrows, advanced and waited while the other two continued to shoot. However, their arrows were less effective from this distance, and the hastati were only reduced to about 58 men each. At this point, the two archers which still had arrows moved up to where the archer without arrows was standing, and the archer without arrows moved into a flanking position on the hastai. The other two archers opened fire from this shorter range on the center hastati while the first archer made its flanking charge on the lefthand hastati. The charge was repulsed. The hastati which had been attacked pursued, but I halted it and returned it to the line. It took losses down to 40 men from arrow fire on the way back to the line. At this point the second archer (the AI general) ran out of arrows. The AI setup a frontal attack with the general on the left and center hastati in combination with a flank attack on the left hastati by the archer that had routed and rallied while the 3rd archer supported with arrow fire. This attack was repulsed, and the hastati pursued. When they got close to the 3rd archer, it attacked the hastati trying to stem the pursuit, but was routed. At no time did the AI use fire arrows. I think this shows a pretty sophisticated AI except for the moving on top of each other, but it does overestimate its ability to win the melee with archers vs hastati.

I repeated the test with 3 archers vs 3 archers but I didn't shoot. The AI did the same thing as in the previous test, but this time they charged frontally and did win the melee easily since my archers were decimated to about 20 men each by arrow fire. So, the AI does shoot, and it correctly judged it's ability to win the melee in this case. Of course, it's going to win the melee because these are equal strength, low armored units in this test, but it's interesting that the AI waited as long as it did before charging.

[Jambo]

Other than the column setup, that's pretty much what I would've done given the same situation and no retreat. There's little point in the archers not attacking as retreating in a custom battle serves no purpose. It would be interesting to see if the same is apparent in a campaign game where retreat is sometimes an option.

[Red Harvest]

Say what you want, but in large battles (again last night) the AI is prone to charging some of its archers into melee where it has no business doing so. That's not sophisticated that is "stuck on stupid"...especially when they are charging legionary cohorts. It is failing to use its strengths and instead is playing to its weaknesses.

It is pretty clear that the triggers/tests for ranged combat don't work properly. Including the pila...tested the pila bit based on others posts about the new problem, the AI doesn't use their pila in RTW 1.3. That makes it rather easy to win 1vs1 hastati/hastati etc. Average kill at contact knocks them from 81 to 71 men. Side notes: both the pila pause and pila ability to stop a charge cold are still there--and this latter really costs the AI when it is charging my pila hurlers.

[Jambo]

Yes, the pila issue really is a major greviance. :/

[Red Harvest]

Judging from the various abilities in the descr_unit file, I'm certain that skirmish and FAW are mutually inclusive and sadly inseparable. e.g. the "thrown" ability means both FAW and skirmish are "on" by default and without it both are off by default. One of things I'd greatly desire as a modder is the ability to preset units with FAW but not skirmish (e.g. for Plumbutari say), something which at the moment isn't possible. :/

They aren't completely inseparable, if you set them up like pila wielders they will have FAW, but no skirmish mode. Of course, that makes them melee primary, and AI units don't use pila now...just another thing that is broken.

I'd be more interested to hear results of tests between a collection of units which are more likely to be seen on the campaign map, e.g. how about a few archers, a few skirmishers, a warband or two and a general? If the missile units still behave ridiculously then, then I'm sure CA might be more inclined to take these results seriously.

As for CA looking into this and actually fixing it: highly unlikely. This has been reported for a long time. (Ignoring MTW's problems with javelin skirmish.) I went through the 1vs1 tests illustrating this about 9 months ago. We've discussed the skirmish problems here a lot. When you start adding a bunch of units, you give CA even more wiggle room to explain bizarre behaviour away. No, a simple test is best, particularly when it confirms what is see in larger campaign battles. If the AI can't do the simple properly, then it has little chance of handling the more complex. If anything, I suspect the more complex battles do a better job of *masking* the problem.

And the formations? That is the most telling part, why are ranged units behind the infantry line in typical default? In light of the problems seen with skirmishing and ranged combat two possibilites suggest themselves: 1. It was done as a crutch since the skirmishers don't work well in front. 2. Not much consideration was given to skirmishing's role on the battlefield in AI design. Both might be the best answer.

The whole tone of RTW battles can best be described in one word: CHARGE!!! Subtleties like skirmishing and ranged combat got shorted.

[Puzz3D]

Say what you want, but in large battles (again last night) the AI is prone to charging some of its archers into melee where it has no business doing so. That's not sophisticated that is "stuck on stupid"...especially when they are charging legionary cohorts. It is failing to use its strengths and instead is playing to its weaknesses.

Yes. Well I don't think the AI was actually coordinating its 3 archers in my tests. It appears to be using each unit individually which is given away by the AI moving all 3 archers to the same spot. It's as though each archer doesn't know the other ones are there. However, the first archer that ran out of arrows advanced and then waited until the hastati were reduced to 58 men. That happened to also trigger the other two archers to move closer.

"Stuck on stupid" is simply that either someone programmed the AI units to charge when they are weaker than the unit they are targetting or the AI is forgetting to take something into consideration such as armor rating which didn't contribute to melee in MTW but does in RTW. I wonder how players would react if the AI always withdrew from battle when it's forces were weaker?

If you are playing on VH, you're going to exacerbate this tendency of the AI to charge. Also, how is not throwing a pila and getting 10 kills on an 80 man unit enough to cause an otherwise equal matchup to loose on VH? One way of increasing the challenge would be to not throw pila yourself since the AI doesn't throw them, and that would solve the pila blocking the charge problem as well.

[Red Harvest]

When the AI uses ranged units as melee, I think this is another consequence of the AI being designed to charge into melee with units that are slightly weaker than the target unit. In STW, the AI does not make direct frontal attacks unless it can beat the unit it is attacking. If its unit is weaker, it always attempts to make an indirect attack. The AI makes attacking decisions using the unit stats in effect at that moment, so there is no doubt about having a combat advantage unless the AI is being tricked into thinking it has a combat advantage by some weighting factor. I can see several tendencies of the AI in battle which suggest a weighting is occuring, and we know that CA used weighting in the auto-resolve so they aren't above designing it into the battle AI to make things more "exciting" I suppose.

I'm sure weighting factors into a lot of the AI combat decisions. You can see some of the weighting by looking at the bar in 1vs1, and it often does not do a good job of representing the actual balance in dissimilar units, particularly when one is a missile unit. CA has a really obvious weakspot with ranged units of any type. I highly suspect their weighting is flat out wrong for melee. Pri/sec bug or the strange charge effect types of issues with the weighting are possible.

In weighting, I'm not sure how much the lethality, charge, and combat bonus parts factor in, nor discipline, morale, stamina, etc. I know fatigue plays a significant role as the bar shifts as fatigue changes. (It is a huge effect down in the "very tired" and "exhausted" range.)

Regardless of the weighting though, there is a bigger issue: Failing to use missile primary units as ranged attackers first and foremost is a major flaw. Whether or not they can win the melee is secondary. If they can win without losses by engaging in ranged combat, then why melee at all? Resorting to melee should be last resort, or used when the weighting (one that actually is correct) says they have a huge advantage.

P.S. Wouldn't surprise me if CA made a simple divisor or sign error. The game/stats have been full of these. Just for the sake of argument the melee test might have been intended as:
If missile unit melee > 2 * target unit melee, then attack
However, the actual formula might have gotten garbled so that the 2 is a divisor or on the wrong side, e.g.
If 2 * missile unit melee > target unit melee, then attack.

[Red Harvest]

If you are playing on VH, you're going to exacerbate this tendency of the AI to charge. Also, how is not throwing a pila and getting 10 kills on an 80 man unit enough to cause an otherwise equal matchup to loose on VH? One way of increasing the challenge would be to not throw pila yourself since the AI doesn't throw them, and that would solve the pila blocking the charge problem as well.

Don't know, haven't tested the pila on VH, I do testing like that on medium. Haven't yet checked to see if stat bonuses other than morale are being given for VH. Obviously spotting the opponent a 15% manpower advantage on medium is decisive--even accounting for the random incremental combat results.

I haven't yet run into the pila issue in campaign since I haven't been fighting Romans or Spanish/Spanish merc. Without others reports I wouldn't have known about it yet. It will weaken the Roman factions against others however, and there is nothing I can do about that.

[Jambo]

They aren't completely inseparable, if you set them up like pila wielders they will have FAW, but no skirmish mode. Of course, that makes them melee primary, and AI units don't use pila now...just another thing that is broken.

Hmmm, not sure about this. The "prec" ability (i.e. throwing pila before attacking) indeed no longer works, but a unit with the "prec" ability definitely doesn't start with FAW "on". If I want FAW on for my legions I have to manually put it on. As far as I'm aware all "prec" does is give you the javelin symbol when the cursor is hovered over an enemy unit, which will make it throw a javelin before charging.

So in that respect FAW and skirmish are inseparable, at least for the AI. We of course have the option of deselecting and selecting whatever combination we like.

[Puzz3D]

The whole tone of RTW battles can best be described in one word: CHARGE!!! Subtleties like skirmishing and ranged combat got shorted.

That's right. So, now I have adpated to RTW v1.3 gameplay with not very smart AI, and I try to play in such a way that the AI has a chance of winning. So for example, if pila don't work for the AI, I won't use them either. The big obstacles to enjoying the game for me in RTW v1.2 were the suicide general, the poor spear performance, the messed up charge which caused units to rout too fast, and the save/load breaking sieges. I haven't figured out what to do about AI sallying, but my solution might be to auto-resolve.

[Puzz3D]

Regardless of the weighting though, there is a bigger issue: Failing to use missile primary units as ranged attackers first and foremost is a major flaw. Whether or not they can win the melee is secondary. If they can win without losses by engaging in ranged combat, then why melee at all? Resorting to melee should be last resort, or used when the weighting (one that actually is correct) says they have a huge advantage.

In my custom battle tests, the AI did try to win by shooting. Even when shooting at a weak unit like archers, they didn't charge until my archers were down to 20 men. Against the hastati they either waited until I was reduced to half strength or they used all of their arrows before attacking.

[Puzz3D]

I did the same test with 3 AI Cretan archers (3/6/4 chg/att/def) vs 3 hastati (2/7/14) abreast on flat map. The hastati have not had fire at will enabled in any of these archer tests so they don't throw the pila.

In the first test there was a glitch where the Cretan archers turned 90 degrees and marched off to my right. At about 45 degrees to my right they got stuck and couldn't move although I could hear orders being given by their commander. After 5 minutes of that, I turned my hastati line 45 degrees to face the archers and they started to move as soon as my hastati started to move. After some indecisive right/left movement, the archers organized one behind the other and advanced into firing range. They distributed their fire so as to reduce the hastati evenly by shooting at the unit with the most men. When my hastati were down to 45 men the first archer attacked my left hastati (the general's unit), but was routed. It rallied, made a second attack and routed again with 21 men rallying off to my left were it set up to shoot. At this time the second archer attacked the two hastati on my left while the third archer shot my right hand hastati. The 21 man archer then made a flank attack and my general died. All 3 hastati routed quickly after that.

In the second test, the Cretan archers moved directly forward one behind the other into shooting range. Once again they fired until my hastati were down to about 45 men. This time the first archer attacked my right hand hastati which only had 43 men. The archer routed and the second archer then attacked my lefthand hastati (the general's unit) which had been weaken further by archer fire, although my right hand hastati was the weakest one except it was set slightly bechind the line. The first archer rallied with 36 men and also charged my lefthand hastati while it was engaged and routed it on contact. The third archer (the AI's general) was shooting my right hand hastati this whole time and had it down to about 20 men. The first and second archers engaged the center hastati and routed it, and then quickly defeated the right hand hastati.

It all seems pretty reasonable except that the AI charges when it can't win.

Note: I forgot that I turned off fatigue.

[Jambo]

Puzz, who's the defender and who's the attacker in your tests? Worth remembering that if the cretans are the attackers then the onus is on them to attack.

Furthermore, it's a custom battle as opposed to a campaign and there's little point in retreating to fight another day. It's win or lose on one battle and that might explain them charging and fighting even though they can't win. I've regularly seen horse archers retreating (not routing) in campaign battles where they can live to fight another day...

[Puzz3D]

I don't see any explicit attacker/defender designation in custom battle. I was at the top of the mini-map with the hatati in the tests. I repeated the 3 archers vs 3 hastati test with the hastati at the bottom of the mini-map, and the AI didn't do anything different. Since I have fatigue off now, the AI was able to rally more and make more attacks, but the archers still didn't win. Sometimes the AI would attack frontally and sometimes it would set up a flank attack. I tried the same test in RTW v1.2, and it was essentially the same result except the AI archers were more reluctant to charge the hastati. Archer kills with the arrows were less, but the AI set up to shoot from further away.

[Red Harvest]

Note: I forgot that I turned off fatigue.

That might have some impact on the decision making later, since it would effect "weighting"--which you can observe changing on the bar, but not for individual units above 1v1. However, that doesn't make it invalid as far as seeing what the AI tries to do.

The result with fatigue would likely favor the hastati: since the archers were working before charging, while the hastati were stationary (correct?), they would have reduced combat effectiveness and take higher casualties, as well as suffering from loss of morale. It could/would tip the result most likely. I haven't kept track of how winded (or not) archers become from firing in RTW. In MTW it was significant. I used to pull some of my arbs off of FAW and rest them while doing defense vs. multiple waves. And those duels with desert archers and horse archers...much more memorable than RTW.

The test still illustrates the root problem though: meleeing unnecessarily with arrows still in the quiver.