I believe combat is rigged in this game and it ruins it for me

[Blake]

Originally posted by eonwe
I did some tests using world builder and created an ocean-map with 250 green battleships for me and 250 for Asoka who I was in war with.
The map is on monarch difficulty-level and option to generate new random seed on reload.
I ran it through only two times, each time attacking each one of Asoka's battleships with one of mine.
I didn't attack any wounded ships nor did I attack with a ship that had already attacked because I only played one round.
So every battleship had one chance to either win or lose.

Results were following: on the first time I was left with 227 battleships and on the second time with 218 battleships. I calculated probalilites to be left with X or less battleships using binomial model and calculating probability density functions using maple, with n = 500 and p=q=0.5. These resulted probalities of 0.027184509 and 0.002395787. Quite improbable one could say.

Prepare forehead slap in 3..2...1

Collatoral Damage

[xxFlukexx]

Blake: I was a bit hard on ya, my bad. Actually the odds for winning a 40%, 50% and 89% battles all in a row are not that unlikely, about a 17% chance to win all three or something. Now you did get a pretty good streak going and that is pretty noteworthy. The result I have a problem with is the first even battle where you didn't even get hit and destroyed the enemy. Now that result in an even 1 on 1 battle like you had it is absolutely absurd to me to be in a strategy game. I don't want me to win a super lucky battle and I don't want the enemy to win a super lucky battle, I just want a fair fight. And because of crazy combat results you can plan your heart out and get destroyed and that to me just plain sucks. I think I'll turn on the reload seed option in all my games now. I really don't like to turn on any options, but as long as these complete bull-s results come in, I'll use it.

The thing is I don't even notice 'lucky' results when I am the favorite, because I brought enough troops for everyone, if my catapult wins with a 2% chance, who really cares? My grenadiers are just going to whine because I spent all this time and money on them and they aren't getting any action. While on the defensive, a case of bad luck can really screw you over, especially in the early game and that's where my beef is.

[xxFlukexx]

For those of you who don't want to read my test: I don't think the AI is getting any combat odds advantage but I do think there are still some problems with the combat system. Below is my test and some strategy musings.

This is a test for 100 unpromoted pikemen vs 100 unpromoted Knights. It is on Immortal difficulty and the "new seed on reload" box is checked.
Chance of the Pikeman winning is 72.8%.
Here is my test before the attacks (note if you reload you will probably get different results than me due to the option set):

http://apolyton.net/upload/view.php?file=68058_testaccountBC-4000pvk.Civ4SavedGame

Here is my test after the attacks:

http://apolyton.net/upload/view.php?file=68058_testaccountBC-4000pvkfinal.Civ4SavedGame

After all is said and done on my test:



Pikemen alive: 78
Knights alive: 22
5 Pikemen went unscathed.

The remaining Knights got hit a total of 58 times. Since it takes 5 hits to kill a knight, knights lost 78*5 + 58 = 448 rounds.
The remaining Pikemen got hot a total of 125 time. Since it takes 6 hits to kill a pikeman, pikemen lost 22*6 +125 = 257 rounds.
Barring me counting wrong (which is completely possible) that makes 705 total rounds.

Knights won 257/705 = 36.5% of the rounds
Pikeman won 448/705 = 63.5% of the rounds

Pikemen won 63.5% of the rounds but won 78% of the battles.
Knights won 36.5% of the rounds but won 22% of the battles.
I have gone over the above reasoning a few times and I think it is correct.

Using Statisical calculations which I won't bore you with, the odds are well within the boundaries and I cannot say that the odds displayed by the combat calculator is incorrect. I think that the odds are correct and consider the matter closed, the AI gets no bonus to combat rolls of the dice or anything, but it sure as hell *feels* like it sometimes.

Assuming that the true probability of a pikeman winning a round is around 63.5%, the odds of an unscathed pikeman is 10.3% or about 1 in 10.
Assuming that the true probability of a Knight winning a round is around 36.5%, the odds of an unscathed knight is 2.4% or about 1 in 41.
Hmmm, I only got 5 unscathed pikeman and 0 unscathed knights, but that is not very suprising.
I believe that the true odds of a Pikeman/Knight winning a round is a bit off from what my test found.
Longest streak was 15 wins in a row by my pikemen, spread over 4 total battles. The streak goes on longer than the screenshot, you can try to fish it out in the long combat log if you want.

Now I did run the test a few more times seeing if I could break 28 wins for the knights and I did not even come close, one test did have 26 wins for the knights but that is it.

This is very surpising to me because if there's one matchup that I lose repeatedly it is my pikemen vs enemy knights. What I found was the exact opposite, I win more than I actually should. I should win 2.7 times for every loss and 3 Pikemen should barely take out 8 Knights if they have time to heal after every fight.

Promotions:
For Knights, shock is just not worth it vs pikemen. Your knight is going to need 2 more promotions to gain an advatage against an enemy combat promoted pikeman. A combat I + shock Knight vs a Combat II Pikeman has a 33.8% chance to win while a Combar II Knight has a 31.9% to win, and it is about the same vs a Combat I Pikeman. For Knights who want to just raid enemy territory I would recommend the Flanking (+withdraw) promotions. Flanking II gives you a 30% withdraw so if you lose ~72.8% of the time, you will withdraw 21.8% of the time so you have a 27.2% chance to win + 21.8% chance to withdraw and hopefully get the heck outta there.
For Pikeman, you want Combat II then Formation to make it a mounted unit killing machine. A Combat II + Formation Pikeman will win 89.6% of the time vs a Combat II Knight (you will probably see mostly Combat I and Combat II Knights vs the AI) Because Pikemen will on average kill 2-3 Knights then kick the bucket, you want both Theocracy and Vassalge(or Pentagon or West Point) to get your Pikemen up to 8 xp so they have a good shot at getting formation, once you have formation you can relax.

[Dis]

I'd like to know how the pikeman destroyed my tank in the game I just finished. . The graphic shows little explosions at the tip of their pike. Interesting. .

okay, okay I'm just being an arse. I was playing warlord level. Which I'm way overpowered for. But it was oh so fun taking over the world. got my highest score ever, over 31,000. And got the Augustus Ceasar title. Something I can't do on higher levels.

[Blake]

Originally posted by xxFlukexx
Blake: I was a bit hard on ya, my bad. Actually the odds for winning a 40%, 50% and 89% battles all in a row are not that unlikely, about a 17% chance to win all three or something.

"It's only not unlikely because it happened".

There's no way that a single unit has a 17% chance of beating "2.5" units with equal strength.

If they were all seperate battles then the odds are not that unlikely, the unlikely bit was actually that it had such good odds in the later battles, it should have taken damage in te prior battles and thus had dramatiaclly reduced odds in the later battles.

The calculations for the probability of winning multiple battles almost blows my mind, but i'm interested enough to try calculating it...

Taking the simple case of totally equal fights, in sucession.
First battle there is a 50% chance of loss.
There is then 5 possibilities for winning:
20, 40, 60, 80, 100 hitpoints. The probability of each needs to be calculated. Each then branches again, with 1 branch for loss, and then a branch for each possibility of hitpoints remaining. Now there is some pretty heavy pruning thanks to the losses, and the fact like there's only two possibilites for a 20hp vs 100hp battle, loss or 20hp remaining... at least i think so... strange things go in with low hp fights.

It hurts my head, altough it probably could be calculated, especially with a combat simulator that gives each possible outcome and probabilities, because there are way too many calculations to do by hand.

[eonwe]

Originally posted by Blake

Prepare forehead slap in 3..2...1

Collatoral Damage

In case you didn't check my save (which seems to be true based on your comment) I didn't put battleships to a stacks.
Each battleship was in it's own square and I attacked each square only once. So there were 500 1 on 1 fights between battleships and I don't see how collateral damage factors in.
I did run it third time and got 210 left alive for me. Probability for being left with 210 or less battleships is 0.0002002 assuming binomial model fits this problem.

Attached is a screenshot of this setup.

[Blake]

Okay, sorry.

I loaded the game and ran a test. I ended up with 133 battleships. I dunno what's going on with your game, are you sure you didn't have it show the TOTAL number of visible battleships?!

Since half of Asokas will be hidden by fog of war, so you'd see all of yours (expected 125) and then about half or 2/3rds of asokas (expected 75-100) for a total of around 200-225, that is the value the mil adviser will give by default, if not set manually to your units only.

Sorry. It's the most logical explaination I can come up with.

edit: Oh you attached the results save as well. I just checked it and confirmed that I'm right adn it was showing Asoka's ships too, with the actual result being the very probable 122. Looks like you get to slap your forehead after all .

[eonwe]

I loaded saves for my three test-runs and yes, I had somehow managed not to uncheck asoka in military advisor to show up only my ships.

So the total number of ships is 500, 250 for each side. In those three tests I ended up with 115, 120 and 121 ships left, which all are easily within 95% confidence intervals of this binomial distribution.
So there doesn't seem to be anything unusual going on here based on this test.
So I don't there that there's anything else left to do than .

Well, atleast I changed my status from lurker to poster .

[Aro]

Welcome to the battlefield, eonwe...

[Blake]

Welcome eonwe. I've done just as dumb things when testing stuff, it's that problem humans have with fixation.

By bothering to actually do a test on that scale you make a very good first impression .

[Aro]

Originally posted by Blake

By bothering to actually do a test on that scale you make a very good first impression .

Indeed.