Help to measure up the tactical AI
What I would like to ask here from anyone who is interested in this, is to fight 4 battles per scenario in order to get an idea about the difficulties of these scenarios and how well the AI can handle them. I will post up the second scenario as soon as there are enough input for the first.
I would like to ask three things after each battle: (a) screenshot of the end screen with the kill/loss ratio; (b) screenshot of the detailed kill/loss stats per each unit (in order to see which unit types get the most kills); (c) a subjective assessment of the difficulty of the given battle, where 1 means the easiest and 10 means the most difficult.
Of course the last one is subjective but that is exactly I am interested in. That is, how difficult each battle feels for an "average" player.
Also, you dont have to be a tactical mastermind, in fact it is not a problem if you are not a tactical genius at all. It follows that there is no shame losing any of these battles, in fact I lost one of the following engagements on the first try. Also important that I am interested in the first try so there is no need to replay these battles. Obviously most of us could considerably improve his performance playing the same scenario time and time again but that is not what I am interested in.
1, First Scenario
You have four battles here. In all battles the AI army (HRE) is the same.
BG means general's bodyguard, pick the earliest available BG unit. Also all battles are on grassy flat, medium difficulty, normal size (otherwise it is difficult to compare the kill/loss ratios).
You can play it with any version of the game (1.1 vanilla, 1.1 modded, unofficial 1.2) just make it clear which version you are using.
AI army, HRE:
1 BG
4 imperial knights
2 mounted sergeants
4 dismounted imperial knights
2 dismounted feaudal knights
3 armoured sergeants
4 peasant xbow
Player armies:
a, Danish:
1 BG
3 huscarls
4 dismounted huscarls
4 viking raiders v1
3 militia spearmen
2 norse archer
3 xbow
b, Milan:
1 BG
2 man-at-arms
5 dismounted man-at-arms
6 italian spearmen militia
6 genoese xbow militia
c, Hungary:
1 BG
3 feudal knights
3 hungarian nobles
6 magyar cavalry
2 pavise spearmen
2 dismounted feudal knights
3 xbow
d, France
1 BG
4 chivalric knights
2 gendarmes
2 feudal knights
2 mounted sergeants
3 dismounted feudal knights
3 armoured sergeants
3 xbow
thx :bow:
Re: Help to measure up the tactical AI
i might do that tonight, am i correct in assuming i dont put any armor or xp upgrades on both mine and the AI's units?
Re: Help to measure up the tactical AI
I wasn't sure what you meant by 1-to-10 difficulty rating - this could mean either :
a) 1=the easiest, 10=the hardest battle I could still win
OR
b) 1=I can't lose, 5=as God wills, 10=AI can't lose
I gave the ratings in the results table below according to (a). If you meant (b), just divide rating by two.
https://img256.imageshack.us/img256/772/resultsma2.jpg
This was played with 1.1 + ProblemFixer
Here's the screens :
milan1
milan2
milan3
hungary1
hungary2
hungary3
france1
france2
france3
danish1
danish2
danish3
---
Edit : just realised I made a mistake : used late BG instead of early. I hope it doesn't impact your test too much :(
Re: Help to measure up the tactical AI
No screens, but here are the results:
Kills/Captures
Danes 546/282, HRE 747/24, Average victory
Milan 576/188, HRE 769/13, Average victory
Hungary 534/543, HRE 134/0, Heroic victory
France 488/320, HRE 393/0, Clear victory
This was played with 1.1 and my own personal mod, which among other small things ups the morale, makes the charges more consistent, and fixes the shields through upping defense - battles feel much better that way, aren't as likely to produce premature chain routs, and are also quite bloodier and longer. I pretty much balanced it so that the battle outcome in an even battle is much closer and the price I have to pay in casualties is heavier. I ended all battles as soon as they were won.
Subjective difficulty (10 is the most difficult):
Danes: 5; It could have gone either way, mostly because of their heavy cavalry punching holes in my lines and DIK/DFK being very tough. I managed to wrap up one flank and then sandwiched the other.
Milan: 5; Also quite close. For a moment after their first charge I thought I was going to lose, but DMAA held and slowly ground away and I finally managed to sandwich and kill his general. I am also not a big fan of commanding infantry-heavy armies.
Hungary: 1; I completely encircled them with HA, shot them to pieces while they just stood there, and then enveloped them from all sides. If I did not act so aggressively at the end (many ended up completely surrounded and fighting to death), I would have had fewer losses, but there's no way I was ever going to lose this sort of battle with the AI as passive and unresponsive to being shot as it is.
France: 3. I had plenty of cavalry, which means low chance to lose. I actually provoked them by charging enemy xbows with a unit of MS, and lost quite a bit of my main line in the initial charge of his knights, but there was a gap between his charging cavalry and infantry, which means I first took care of the cavalry with my combined infantry and cavalry, and then smashed the infantry with the return/flanking charge of my own knights. His knights didn't flank, they just smashed into my main line in all four battles.
As I said previously in the AI discussion thread(s), the battles like these are fun. I won them all without too much effort and attention to detail, but it was fun and tense - they are just rarely seen in a campaign.
Re: Help to measure up the tactical AI
Well, I guess a battle plan description and brief account is in order to clarify the results I posted above.
1. Danish
Deployment : Double line with a thin, loose formation of spear militia in front (charge breakers), the axemen at the back, the shooters on flanks oblique, cavalry even further out. Looks someting like this :
- - \ =========== / - -
Account : I marched the army as quickly as I could towards the enemy, and just charged the axemen through the sparse spearline at whatever was in front of them. Cavalry moves to flanks and rear of the enemy. Archers & XBows : plan was to move them outwards and shoot along the line from both sides once the battle was joined. The xbows had not much of an impact though, as they got charged and killed quickly.
Feeling : Tense battle, any mistakes and it could've gone bad.
2. Milan
Deployment and account : similar to the above.
Differences : Solid spearline this time, with the DMAA in square blocks spaced behind. Xbows lasted somewhat longer being protected from the inital charge with a countercharge of my own heavy cav. They still didn't played the major role I remember them from campaign battles.
Feeling : Quite close. I actually thought I would lose this this one until almost the end.
NOTE : In both of the above battles you'll notice there's one cav unit that scored ~300 prisoners. That's the "router patrol", a unit sent behind the enemy whose sole purpose is to make sure the infantry units breaking from the main line don't rally and go back to fight. All along the battle line I tried to create local advantage by numerical superiority or threatening rear charges. This caused the foot knights to rout early (30-45 men), just to be ran down by the patrol. This tactic requires a lot of micro (50-60% of the time spent manually driving that one unit), as the router chasing doesn't work well on auto and the only way to be effective is to use the click-ahead trick.
3. Hungary
Deployment : Back at the map's edge. Double line, xbows front, knights & spears behind. Heavy cav on flanks. HA's in squares, one behind another, further out on flanks and ahead, close to the enemy. Like this :
Code:
h h
h h
h h
h h
B H x x x H H
== == == == ==
Account : Run the HAs to the sides of advancing enemy, shooting along their line and rear on FaW ; their heavy cav gives chase in vain. Used my heay cav to charge whatever DFK or XBow strayed too far from support. When the line joined, close all available cav (ie those not skirmishing away from HRE knights) from flanks and rear and mass charge then pull back, chase routers and charge again anything that's still standing. When the main line's gone, gang up on the knights.
Feeling : Almost too easy. HAs are the most powerful M2TW unit type.
4. France
Deployment : Three lines : heavy cav at the front, xbows middle and foot at the back, alternating spears with swords in line. Mounted sergeants and additional heavy cav out on flanks, facing inwards (sort of wide U shape).
Account : Heavy cav frontal charge, HRE xbows gone. Pull back a short distance then charge again, since the AI is slow to react and the dismounted knights line is still exposed. Pull the cav to the flanks, while the former flanking unit move to the rear of the advancing enemy. Let the xbows shoot whatever they can, then pull back. Engage the line, matching spearmen to cav and knights to spearmen. Rear charges, sandwich the HRE knights, mass rout, game over.
Feeling : Fun, intense cavalry battle, moderate difficulty. AI a bit dumb, moving xbows too far ahead an failing to push spearmen out to counter the charges.
Re: Help to measure up the tactical AI
Gentlemen, thanks for the tests and for the detailed battle descriptons. :bow: Very interesting. ~;)
@ Ars Moriendi, I was thinking about the first kind of rating (your (a) rating), so it is fine. Of course a defeat can have different difficulties too (like losing a close fight or losing it big time) but I expected people to win. Of course, I proved myself wrong as I lost one of these battles ... ~:joker: but let's stick to this rating as I still expect most of the people to win all 4 battles.
Also, late BG not much of a problem, I guess you could have done the same with an early BG (i.e. chasing routed units). Of course early BG is still the "default". What I am interested in though is that whether you could do this (i.e. send your general behind enemy lines to chase routers) without any harassment, or were your BG harassed by enemy cavs?
Re: Help to measure up the tactical AI
Quote:
Originally Posted by Cheetah
What I am interested in though is that whether you could do this (i.e. send your general behind enemy lines to chase routers) without any harassment, or were your BG harassed by enemy cavs?
Well, they were occasionally intercepted by some none-routing unit, but nothing that wasn't quickly solved with the help of another unit breaking from the main fight to "sandwich" the offender.
The trick is to keep them far enough to be safe, but not too far so the routers have time to rally. Also, as these battles went, most of the HRE cavalry was busy from the beginning - either deeply embedded in my line from the initial charge or chasing my xbows.
The large number of casualties these units scored (~350 for the BG with milan, ~250 one of the huscarls) was also helped by the fact they were in the ideal position to mow down large numbers of enemies in the final mass rout.
Another point : some fast, light cav might be better for this role, although probably the first contact with HRE knights would've been the end of them.
Two questions :
- what is it that you hope to achieve with these tests ? :inquisitive:
- when do we get the second scenario ? I haven't played the game in months (still waitin for the darn patch :furious3:) but seeing your invitation I just couldn't resist to fire up some custom battles...
Re: Help to measure up the tactical AI
Quote:
Originally Posted by Ars Moriendi
I haven't played the game in months (still waitin for the darn patch :furious3:) but seeing your invitation I just couldn't resist to fire up some custom battles...
Same here ~:)
Re: Help to measure up the tactical AI
Quote:
Originally Posted by Ars Moriendi
- what is it that you hope to achieve with these tests ? :inquisitive:
Well, I would like to know which setups are easiest for the human player, and which setups are the most challenging, hoping that the devs will take notes and will try to guide the AI recruitment policy (and perhaps the upkeep costs) accordingly. ~;)
The second scenario:
Human player has the same armies as before.
AI army has a slight change as it has 4 more melee troops (2 feudal knight and 2 dismounted feudal knight) instead of the 4 xbows:
AI army, HRE:
1 BG
4 imperial knights
2 feudal knights
2 mounted sergeants
4 dismounted imperial knights
4 dismounted feaudal knights
3 armoured sergeants
Re: Help to measure up the tactical AI
Re: Help to measure up the tactical AI
Oops, i did it the wrong way round--kept the same hre army against the four armies you specified as the opponents :embarassed:
Re: Help to measure up the tactical AI
Evaluation for the second scenario : Harder than the first. If I had not the experience from the first scenario with similar armies, the outcome would have been worse.
1.Danish - identical deployment, similar battle. More difficult, although the results are almost the same.
2. Hungary - identical deployment, battle not so similar. First, the AI just stood there for a while getting shot, whereas in the first battle it attacked right from the start. Second, more cavalry for the HRE allowed the AI to chase my HA more effectively. Horse archers still dominate the battlefield, but they're difficult to manage against large numbers of cavalry so this time more maneuvering was required. I took more than twice as many casualties.
3. France - identical deployment. I started with a frontal charge (4 Chiv Knights), destroying 4 units of dismounted knights. I pulled back to prepare a second, but there was no time for it because the AI countercharged his imperial knights throughout the line. All the rest of the battle was cavalry maneuvers : charges, counter-charges, flanking, surrounding etc. The few casualties my infantry had were probably caused by boredom, since they were never issued any orders...
4. Milan - here's the ugly part : I lost the first time, the results listed in the table are from the second run. I know we're supposed to give the first try only, but I was so pissed I just clicked "Play Again" before thinking about recording the results. Sorry.
...
That being said, the problem here is the uselessness of the foot missile units. The first three battles can be won without much of their assistance, but the Milanese is close to being unwinnable (well, for me at least). Not having any kind of terrain advantage and with a cavalry balance of 3-vs.-9 it's quite difficult to employ them effectively. My usual plans for missile units would be (a) deployed in front in a wide line, fire at the approaching enemy - which in these scenarios was useless considering the AI just ran at me and there was no way to delay them enough for the xbows to cause serious damage; or (b) deployed on flanks to move outward and fire at flanks and rear of the engaged enemy line - not achievable due to significant disadvantage in cavalry. I might as well start the battle with 6 less units.
The way I solved this in my second Milan attempt : a kind of deployment I've never used before. I put the italian spears in schiltrom and placed them with rather wide spaces between, like this :
Code:
* * *
m * m * m *
m m
x x x x x x
where * is schiltroms, m is dismounted man at arms and x crossbows.
The AI charged the schiltroms, and while they formed these "circles" around the spearmen it allowed the dmaa to engage in a sort of flanking. More importantly, the wide spaces allowed the xbows to fire between the engaged units long enough to kill significant numbers, while being protected from the initial charge. If you look at the results you'll notice they had the most kills (with the exception of the very special "router patrol", of course). Some of them were eventually caught by enemy cavalry going around the infantry formations, but they did enough of their job by then.
This solution allowed the foot missile units to change their role of "power amplifiers", meaning they just helped reduce casualties by killing from afar in battles were I had the upper hand and would have won anyway without them - to a more active part, being the little something that wins the day.
Re: Help to measure up the tactical AI
Are there any rules in terms of PAUSE or speed? Being able to hit Pause in a SP battle at any time makes things far far far easier.
Re: Help to measure up the tactical AI
Well, I managed to lose all 4 battles on the first try, and then win them all on a second try. I am very happy with the balance I managed to achieve in my mod. ~:)
I am not really surprised though. In my mod, DFK have total def of 32, and I had spear militia to counter that. I managed to win the encores by pretty much rushing a single flank as fast and as hard as I could, then rolling up the other. However, there was no way my classically-deployed combined arms lines could hold against a charge by an army the half of which is cavalry and the other half is elite infantry. With the higher morale, I simply ran out of men in the battles I lost when my lines were spread wider, even though half of the AI army was routing too.
This is also why I disagree with people who advocate upping the costs of cavalry to balance the game: this does not prevent a player from using it, but it does prevent the AI from buying overpriced units, which makes the game easier in fact, not harder.
In any case, I can conclude from these two batches of tests that:
- the AI has problems handling missile units, both using them and defending against them (especially when it has its own missiles too)
- the massed cavalry rules the day even with a simple frontal charge
- the AI's biggest problem is bad campaign map recruitment
- combined arms is still more fun than spamming the elites, and I'd prefer that the AI can do well with combined arms than to spam a single kind of unit
