Log in

View Full Version : Battlefield difficulty test results.



Hashashiyyin
12-06-2006, 02:23
After some of the data files were released, I was hoping for some news on the different battlefield difficulties. There seems to be confusion (at lest on my part) on wether or not the AI gets moral/combat bonuses on the different battlefield difficulty levels. I've heard that VH does not give bonuses and other say it does. I started out believing that there was no bonus, until I had a few odd battles in a VH/VH campaign that seemed to point to the AI getting bonuses. Impatient, wanting to start a new campaign, I did some testing of my own and here are the results.


Test set 1:
Battlefield - morocco, clear, morning.
Playing as the English against the French.
Each test was with one unit of dismounted feudal knights per side. I wanted to see how a unit stood up to its carbon copy on different difficulties. I did a very few tests on medium, more interested in VH battle difficultly and the medium difficulty tests were more of a control group.

Medium: 10/10 battles in my favor. Not really a contest. I wiped the floor with them in every test. Obviously medium does not give the AI any bonus. It's so easy; your own troops are probably getting a rather large bonus at this level.

Very Hard: 100% victory rate for the computer AI. Most of the time losing by 20 or so men. One battle was very close and both sides took heavy loses, but for some reason when I was winning, my troops routed. Tried the test in many different configurations of attacking and defending. For attacking, I ordered my unit with a single "single-click" attack command and let the AI do the rest. For defense, I tried to counter charge the AI with a single-click command at the same place in the enemies charge.

This made me think. Not only was the AI smoking me, it was doing so rather hard. This and some surprising campaign battles I've had, made me feel like my worst feelings were true, the AI still cheats. However, I notice at lot of the time, my army was getting hung up on terrain, or had to walk up the side of a mountain first. This leads me to my second set of tests because I wanted to see the results with terrain factored out.


Test set 2:
Grassy Plain
As England against France, 1 unit of dismounted feudal knights each.
Again, tested many different configurations of attacking and defending.
Very Hard results: +90% success rate for me. Almost a complete turn around of what I saw in morocco. Results were so in my favor that I repeatedly checked to make sure I was on Very Hard. Only time I lost was once when, for some reason, my charge was not uniform. The other lose was a test of my troops taking a charge w/o counter-charging. Most fights were total victories on my side, with at least 30% of the unit's numbers remaining. When terrain factored out it seems as if your units have a slight bonus on very hard.

Now I was rather confused. With the terrain taken out of account, it's more then clear the AI's troops get no bonus what so ever. In fact, the results seem to point to your own troops having a slight advantage. Thinking back to getting the opposite results in morocco, I figured it might have been user error. Back to Morocco I headed.


Test Set 3:
Morocco, clear, morning.
Again, England vs. France, 1 unit of dismounted feudal knights per side.
Again, many configurations of attacking and defending were tried. This time with much more attention to the terrain during deployment.
Very Hard results: +95% victory rates for the AI. Same results as before. I was however able to win one battle, seeming more out of shear luck then anything I personally did.


So to recap:
With terrain affecting units, AI wins 95-100% of the time. When terrain is taken out of the equation, I won 90% of the time.


Analysis:
It seems really odd the total flip flop that occurs between carbon copy unit interactions when terrain is a factor/not a factor. The one thing that became clear during the testing on VH was the difference between the AI's and your own ability to micromanage troop movements and charges. We still have the same basic method of controlling our troops as in Shogun: Total War. While this system has been upgraded with ever successive Total War, we are far behind the AI's ability to micromanage troop movement/charges and very far behind on the speed and fidelity in which said commands can be issued.

Thus the advantage the computer has on VH is not in moral or ability, it lies in its advantage to interpret and correct for the terrain at a much great rate and fidelity. VH is going to be the most realistic battle in terms of unit to carbon copy unit balance. That is, on VH battles, you and the computer should be on an equal footing, ability, fatigue, and moral wise. The computer does have an advantage in its ability to micromanage its troops. So the overall effect is that the computer has a slight advantage on the VH difficulty.

This advantage, however, does not take into account tactics. The superior use of tactics on your part will (almost) always win the day against the AI. Your ability to intelligently move, flank, harass, etc. far outweighs the slight advantage the AI has in micromanagement. This is, of course, assuming that you are fielding an army capable of said tactics. A strait lineup of equal infantry vs. infantry on VH is more then likely going to go drastically in the AI's favor. Unless you get really creative with tactics on a strait infantry vs. infantry (obviously, carbon copy units) fight on VH, you can expect to lose. But with skirmishers, heavy and light cavalry, archers, etc. you can more then make up the difference with solid tactics.

Final Word. VH is the most realistic battle setting as long as you understand and compensate for the AI's far greater ability to micromanage units and correct for terrain.


Hope this helps you when deciding on your next campaign difficulty level. Please feel free to test my results and post yours.:beam:

R'as al Ghul
12-06-2006, 12:31
Your analysis concludes that the AI is superior on VH because of tactics and terrain. I don't see how tactics are involved when you charge two units frontally against each other. Did the AI, in any of those tests, flank you or did it always charge you headon?
Also, in how far do you think did the terrain influence the battle?
You took two equal units, both are/ should be affected by the heat that you most probably have on the Morocco map. I don't see any advantage for either unit. Was one of the units charging downhill or was the height equal for both? If the AI always charged downhill into your unit then it's clear why they won 90% of the time.
You need to offer more information for these tests to be conclusive.

R'as

maestro
12-06-2006, 13:47
The problem is that, like every other TW game, it's way too easy on VH/VH. Tactical and Strategic AI is still nowhere near challenging :wall:

Fridgebadger
12-06-2006, 15:43
One thing i've noticed on VH is the effect of fatigue - my units get exhausted *very* quickly, and then barely recover even when they're standing around for half the battle.

Now, if anyone can find out the exact effect of fatigue... is that just moral, or does fatigue effect stats as well? In which case, CA have just slipped in the VH stat penalties by the back door...

Slaists
12-06-2006, 16:15
If I understand Hashashiyyin correctly, what he meant by superior AI with difficult terrains is actually pre-battle positioning. Hashashiyyin metions, he frequently would have to tire his troops before being able to fight the AI on mountainous terrain.

I have a different question though: did Hashashiyyin try to do both sides of the battle: i.e., attacking and defending? It seems, the pre-fight exhaustion would be a factor in a scenario when the player attacks. What happens when the player defends (and has the luxury of choosing the more advantageous terrain)?


Your analysis concludes that the AI is superior on VH because of tactics and terrain. I don't see how tactics are involved when you charge two units frontally against each other. Did the AI, in any of those tests, flank you or did it always charge you headon?
Also, in how far do you think did the terrain influence the battle?
You took two equal units, both are/ should be affected by the heat that you most probably have on the Morocco map. I don't see any advantage for either unit. Was one of the units charging downhill or was the height equal for both? If the AI always charged downhill into your unit then it's clear why they won 90% of the time.
You need to offer more information for these tests to be conclusive.

R'as

BuckNekkid
12-06-2006, 16:28
I have yet to lose a battle on VH. I'm not bragging, I'm complaining. I SHOULD lose once in a while. Someone mentioned that the AI was actually smarter on Medium. So I've been playing on that, and so far the AI has been more aggressive. AI on VH is just too passive (whether I have fire superiority or not). Medium is still too easy, but more fun.

I really feel that little work was done on the AI between RTW and M2TW, even though that's what the vast majority of posters asked for. Shame.

Bob the Insane
12-06-2006, 16:36
Well I have noticed on VH battle that you men will get fatigued just marching uphill. And a substantual hill will leave your men exhausted and on VH they do not recover very fast (if at all)...

I find it is better to use multiple units of the same type for these tests as it evens out other variable and allow the effects of morale to be seen. Simply leaving them in the default formation (which will match the enemies) on the grass plain and group the units and tell the group to attack the center of the opposing force and let the AI get on with it.

You have me intrigued though, I will have to do some tests myself tonught...

Quillan
12-06-2006, 16:49
I'm not so certain that's what he's referring to. I noticed something in a fight last night against the Danes with the Spanish. The AI was hiding most of its army in the woods. I unloaded my artillery on them until they ran out of ammo, then started closing with the remainder of the army. They came boiling out at that point. I had a 5 unit line of Tercio pike in the center of the line, with 3 units of Sword & Buckler men on each end, and musketeers out front. The norse war clerics avoided the pike and went after the other stuff. One of the cleric units came out of the woods at a dead run (full speed), across a field with a couple of dead trees in it, and hit one of my S&B units. I went from 90 men to 14 men in a matter of seconds. I can't get a charge that effective with lances over that terrain by single clicking and letting them walk, and here the computer does it at a dead run from an angle. I think that's what he's referring to by the computer compensating for terrain. Morocco has a lot of rocks in its battlefields.

Slaists
12-06-2006, 16:56
I concur, the VH battle AI playing the campaign is really easy... at least, before the "passive AI fix". I have yet to see the "fast reacting" AI or an AI being able to pull off a flanking manouver.

As to the Strategic AI: it seems, on "harder" difficulties it just becomes more aggressive. In MTW1, there was a debugging mode console switch that allowed the player to switch through various AI controlled factions and control them for some time. When I did that, I noticed just how retarded the strategic AI was. 100 turns into the game, almost all of the AI factions had ran their economies to the ground, no trade networks maintained, fleets fragmented and in weird places on the map costing money and not facilitating trade, almost no economic buildings built even for the factions that had huge economic potential. I wonder, what would be the picture if we did the same look at AI faction development in MTW2.



The problem is that, like every other TW game, it's way too easy on VH/VH. Tactical and Strategic AI is still nowhere near challenging :wall:

Kraxis
12-06-2006, 20:46
As to the Strategic AI: it seems, on "harder" difficulties it just becomes more aggressive. In MTW1, there was a debugging mode console switch that allowed the player to switch through various AI controlled factions and control them for some time. When I did that, I noticed just how retarded the strategic AI was. 100 turns into the game, almost all of the AI factions had ran their economies to the ground, no trade networks maintained, fleets fragmented and in weird places on the map costing money and not facilitating trade, almost no economic buildings built even for the factions that had huge economic potential. I wonder, what would be the picture if we did the same look at AI faction development in MTW2.
I remember that... I modded the AI to put a lot more interest into ports. That little change did more for MTW than any of my other numerous changes. Suddenly the AI fielded massive fleets, long lines of trade (the the point that I actually made rather substantial amounts of money recieving the trade, which was something like 10-20% of the other end). And their armies were suddenly full of superb units. I got creamed the first time... Playing the HRE was suddenly impossible (because of the usual difficulties, and now with strong economies for the AI).

But this is interesting... I would like to know more about this. At least it confirms that CA wasn't lying when they said the AI didn't 'stat-cheat' on VH. But I wonder why the player would win so easily on Medium or on Grassy...

FactionHeir
12-06-2006, 21:08
In M2TW the AI still doesn not put much interest into trade and infrastructure. The only buildings I seemt hem build when I caspture their towns is inns, barracks and stables. At most a shipwright, a level 2 church, a guild, highest level siege engineer and high level farms.
They rarely build normal roads if at all and never paved roads. They never build ship trade buildings and no more than market for land trade. Most AI I see in the game has a meagre or bankrupt wealth rating too. And yet they can easily maintain their armies.

Kraxis
12-06-2006, 21:41
I agree about churches, siege engineers and money. There are seldomly cathedrals (I have seen enough abbeys) and the AI factions are always broke (but that is because they spend their money and don't save). And the AI always comes out with Ballistas long before me, same with catapults and trebs.

I have however seen a whole load of roads. In fact I lagged behind on the paved roads compared to several factions. And I was also behind on the seatrade. I usually see plenty tradefleets coming out of their ports, and when I capture their cities they have at least the first level of Merchant Harbour (or what it is called). They also tend to have Fairgrounds at least, which is good enough.
However I see them lacking in farming usually, generally never more than the first two (and usually only the first).

So it seems the AI does what it senses as the most sensible in each game. That might not be the same each time. That is good I think.

Bob the Insane
12-07-2006, 00:41
As to the Strategic AI: it seems, on "harder" difficulties it just becomes more aggressive.

Well trawling through the file I saw that on Hard and Very Hard your diplomatic relationship with the other factions will constantly degrade by a small amount every turn (thus requiring constanst small gifts to stay on an even keel). Of course this includes factions you have no contact with. In fact this also sort of occurs on Mediium as well...


To explain H and VH are normalised to Abysmal and will shuffle that direction every trun (with VH moving faster).

Easy is normalized to Perfect so relations will continually improve (provided you do not do one of the other many things to anny the AI).

Medium is normalised to Neutral so if bad relations will improve but if good they will worsen.

Add to this the AI is effectively set to Medium difficultly when the AI factions interact with each other.

And the relations for the top 3 scoring factions will tend to worsen with everyone as well.

Lastly when you fight a member of a particular religion everyone else in the religion will start to take a dislike to you.

You can see how easy it is to get bad relationships with everyone, especially on Hard or Very Hard.

Additionally look in the Campaign AI XML doc in the invasion_decisions section at the end is the following:

"SPECIAL CASE EXISTS HERE IN CODE FOR FORCING ATTACK ON HUMAN IF AT PEACE WITH EVERYONE FOR TOO LONG (hard = 4 turns, normal = 10 turns, easy = 20 turns)"

This is the only bit of that doc that mentions difficultly levels...

Kobal2fr
12-07-2006, 03:24
Hmmmm interesting. That would certainly explain the irrational attacks of allies, right out of the blue...

So, to keep peace with your neighbours, you need to keep a "pet war" going on somewhere. Good to know.

Whacker
12-07-2006, 03:32
Additionally look in the Campaign AI XML doc in the invasion_decisions section at the end is the following:

"SPECIAL CASE EXISTS HERE IN CODE FOR FORCING ATTACK ON HUMAN IF AT PEACE WITH EVERYONE FOR TOO LONG (hard = 4 turns, normal = 10 turns, easy = 20 turns)"

Do we know if this includes rebel cities and stacks? As Mr. Kobol2frenchytype pointed out, a "pet war" would seem to be the logical answer and a good idea here to avoid this trigger. If rebel factions don't work, then as an example if you are catholic, corral one of the Orthodox or Muslim factions to a single isolated province and keep them there, and never make peace.

Kraxis
12-07-2006, 13:27
Guys... the line means that very faction. Not the human player.

So if the human player is HRE and Denmark is suddenly at peace with everybody, then Denmark will attack the HRE after 4 turns at hard.

That sort of explains why I got jumped by all factions on my border as they were at peace with each other (there are of course more reasons).