Log in

View Full Version : Creative Assembly Concerning Player Advantage



A.Saturnus
03-03-2003, 14:20
It has been stated that the player gets a combat advantage in normal difficulty setting just as the AI gets one on hard setting. I did some tests about this.
If it is true, we could predict that the player would win with a chance of at least 60% in a custom battle between two equal units.
Test:
First, I tested it with Chiv Knights. The custom battle was Agincourt, German vs. English chiv Knights, spring, lush grassland. 10 times the AI attacked, 10 times I did. No one used any visible tactics, both units charged on one another in wedge formation.
After that, I repeated it with CMAA. Settings were the same except that both units charged in line formation.
Results:
From 20 trials with chiv Knights, the AI won 17 times (9 attack, 8 defence). The average kill ratio of the AI was 29,4 ; standard deviation (sd) was 7,28
For the player, average 21,25 and sd 9,12.
Most of the time my general was killed on impact, so I don`t think these results are really reliable.
For CMAA, I registered when the general was killed. The AI won 12 from 20 trials (6 attack, 6 defence). My general was killed in 5 trials, the AI`s in 0 trials. Average AI kill ratio was 49,2 and sd 7,28. Player average 44,75 and sd 9,12.
If cleared for the CMAA-trials with a killed general the results are (15 trials remaining):
AI average kill ratio 46,4 and sd 11,96
Player average 45,6 and sd 7,61.
Conclusion:
Especially these last data suggest that there is
no advantage for the player. The differences between kill ratios are not significant. The player general seems to be killed more easily than the AI general.
Since the sds were all relatively high, the sample size of effectively 15 trials may not be enough.
It would be good if all similar tests in the future would clear the data for the killing of a general.

Red Harvest
03-03-2003, 18:19
Saturnus,

I've read that "normal" is equal, no advantage for either side. Hard & Expert give +1 and +2 morale or valour (can't recall which, and it makes a big difference.)

A better test map to use is one of the Steppesinland maps. There is one that is perfectly flat, and grass only (think it is 01.) Use summer to reduce unwanted rain. Avoid wedge. I typically use V1 or V2 with A1 or A2 since these are the most common levels seen in "real" unit builds and battle. The general should last a bit longer like that. In my tests the generals rarely die until the last 1/3rd of the unit remains and sometimes last until the last 1/10th.

Hamburglar
03-03-2003, 22:07
I have an idea....


Both of you were charging at each other with your general's unit, so him getting killed really just messes up your whole army and makes you lose.


I propose to redo the tests, except make each general's unit be a siege weapon so it can't move and join the battle. Make it something that cant traverse like a trebuchet.

When attacking, just attack from an angle so the AI treb doesnt shoot and when defending make your treb face the map edge or something so it doesn't interfere. This will get rid of the general's death I believe.

longjohn2
03-03-2003, 22:18
There's no combat advantage to either side on the normal setting.

Kraxis
03-03-2003, 23:59
Good to know LJ...

econ21
03-04-2003, 01:32
Excellent, Longjohn Call me a wimp, but I like two supposedly equal units to perform equally well.

I hope in RTW, CA will consider allowing various aspects of the difficulty to be scaled. I want that AI always to fight its smartest game; often would like it to be richer; but generally don't like to give it combat advantages.

A.Saturnus
03-04-2003, 12:34
longjohn2, how about the chance to kill the general? That my chiv knight general died most of the time didn`t look like chance.
Of course, doing the test with chiv knights wasn`t a good idea. The AI uses wedge, so I also have to. With CMAA, the dying of the general isn`t so much of a problem for testing. You can simple clean the data for it.

Puzz3D
03-04-2003, 16:06
Thx LongJohn

It always appeared to me that normal was no advantage to either ai or human player.


A.Saturnus,

My understanding from past posts by LongJohn is:

Easy: Human gets +4 morale.
Normal: no advantage.
Hard: AI gets 15% combat advantage.
Expert: AI gets 30% combat advantage and +4 morale.

Although the general starts with something like 25 hit points, I think there could be a fairly large statistical error associated with his being killed. Wedge puts the general at the point of the formation which increases his chance of being killed by exposing him to more hits, and the shift of 3 combat points from defense to attack (melee) reduces the number of combat cycles needed to kill by increasing the probability of getting a hit. You could turn morale off to minimize the effect of death of the general on morale or use a siege weapon as Hambuglar suggests and turn off fire at will. Definitely do tests like this on a flat map, and use the largest unit size to maximize the number of combat cycles. Also, meet the AI half way so that the fatigue levels are the same.

A while back CBR sent me a unit stat modified for +6 morale to all units for some MP battle testing. I've been using that stat in a SP campaign on normal difficulty, and the AI is definitely harder to beat in battle. It's more aggressive and coordinates frontal and flank attacks. Since the AI picks unit matchups very well, it can be a handful trying to meet all these incoming threats. Since I play mostly MP, I never pause the battle to issue orders, and this makes the battles quite difficult. The AI's tendency to limit how many men it sends to attack a province seems to fit well with the more aggressive tactics. It's definitly coming in with enough to win.

A.Saturnus
03-05-2003, 15:20
Puzz3D, I did this test because someone said the player would have a advantage on normal setting, don`t remember who. I don`t believe the general killing to be a statistical issue. My general was killed much more often while the situation was symmetrical. But I`m not sure, needs some more statistics to clear this.


...3...

Kraxis
03-05-2003, 20:47
That someone was me... It was because pre-patched I noticed that equal units would have a tendency to go to the player... So from now on I will test on Normal again.

Puzz the +6 Morale at Expert must be a deadly combination... I mean Peasants with Morale 6. Knights will never run and normal troops will fight until 8-10 are left...

But are those points just added to all units? Or is it some other feature? I think I will try a +4 Morale version.

Puzz3D
03-05-2003, 22:17
Kraxis,

Yes that's +6 morale to all units, but maybe +4 would work better. I'm using a stat that CBR sent me in which he added +6 to everything so we could test some low florin games MP battles, and it worked well online. I agree that, if you play at expert difficulty and the +6 morale, then the ai morale will be too high. I'm playing at normal difficulty and finding it quite challenging, and there is hard difficulty which would give tougher battles without any morale change. If you usually play the campaign on expert difficulty, then that should be producing tough battles already without any other changes. Like Saturnus, I'd like to get a tough battle out of the ai with equal unit stats and the higher morale seems to do that, but maybe more battles are in order before making a final opinion.

longjohn2
03-05-2003, 22:28
The chances of a general being killed are the same for both ai and player.
In custom and mp battles the general only gets 6 hit points. In campaign games, he can have a lot more, depending on his VnV's and whether or not he's a king. I think his command stars also play a role in how many hps he has. Of course, if he gets hit by a cannon ball, hps count for little :-)

Kraxis
03-05-2003, 22:32
Quote[/b] (longjohn2 @ Mar. 05 2003,15:28)]The chances of a general being killed are the same for both ai and player.
In custom and mp battles the general only gets 6 hit points. In campaign games, he can have a lot more, depending on his VnV's and whether or not he's a king. I think his command stars also play a role in how many hps he has. Of course, if he gets hit by a cannon ball, hps count for little :-)
Yeah... Command grants 1 life per level, that was at least what Gil said... It is all compiled in the unofficial FAQ at the com. http://www.totalwar.org/forum/non-cgi/emoticons/wink.gif (I knew something a dev didn't... cool http://www.totalwar.org/forum/non-cgi/emoticons/biggrin.gif )

Puzz
So you didn't test a +2 Morale scenario like all the MPers wanted?

Puzz3D
03-06-2003, 20:32
Thx LongJohn.


Kraxis,

I don't recall "all" the MP players wanting +2 morale, although I do remember I suggested +2 back on Sept 9th when we were still using v1.0. That was a conservative suggestion on my part because we didn't yet have a lot of experience with the game, but it was clear that morale was more fragile in MTW MP than it was in STW MP. Under v1.0, 10k became a bit of a standard for MP games. On average, that was almost +4 morale above a 5k game (5k * 1.5 * 1.5 = 11.25). Under v1.1, 15k florins is more the standard with many players using more than that. At 15k you are on average +4 morale above a 5K game (5k * 1.7 * 1.7 = 14.45k). The move to 15k was a consequence of upgrade cost of valor going from 50% in v1.0 to 70% in v1.1. So, I would say that most MP players want to play at something like +4 to +6 morale above valor 0.