Combat Odds

[DemonMaster]

i tested it with 600 combats, swordmen versus warriors on grassland.
the combatodds were 75,9%, so there should be 455,4 successful attackings.
i got 451, thats only a different of 4 and as percentage -0,73%. that sounds good enough.

You tested 600 combats, that must have been boring?

 

[Nicopol]

there is no problem with normal units like swordmen.
the problem is with ÜBER UNITS with almost all promotions and a lot of unique equipment.
it's not seldom that they lose 99% battles therefore i alwasy give my "heros" the immortal promotion with wb because wb is way quicker than a reload.

maybe the first strike(s) killes the combat odds...

 

[eerr]

there is no problem with normal units like swordmen.
the problem is with ÜBER UNITS with almost all promotions and a lot of unique equipment.
it's not seldom that they lose 99% battles therefore i alwasy give my "heros" the immortal promotion with wb because wb is way quicker than a reload.

maybe the first strike(s) killes the combat odds...

full first strikes are not messing it up

first strike chances are what mess it up, as it is calculated only for the average number of first strikes, and not for each number individually

 

[Sensemann]

You tested 600 combats, that must have been boring?

nope, i have done it very quick. i created a line of 100 swordmen and warriors, saved it and turn quick combat on. after combat i looked at the statistic and load again. took me 10 min.
the difficult is to add the drill iv and commando promotion to the units, so that i can test it for this promotions. exist an easy method to do this? can i edit the xml, so that the standard warrior get drill iv and commando promotion?

 

[eerr]

nope, i have done it very quick. i created a line of 100 swordmen and warriors, saved it and turn quick combat on. after combat i looked at the statistic and load again. took me 10 min.
the difficult is to add the drill iv and commando promotion to the units, so that i can test it for this promotions. exist an easy method to do this? can i edit the xml, so that the standard warrior get drill iv and commando promotion?

i think its between <freepromotion><freepomotion> or something in the code, try comparing with a unit that does get certain promotions free
(i don't think its between <bfreepromotion> because thats boolean for the real magic-user, monument given deal)

 

[Mithrus]

Another issue that hasn't been explored (or at least openly discussed) is what condition is the winner in at the end? FAR too often, the victors that had a slim chance of winning are walking away nearly unscathed, and the victors that should have won with ease are near death. I would think the odds of winning would be a reasonable marker for what condition the unit could be expected to be at if they win. You have a 1% of winning, yeah, I think you shouldn't be near full health at the end.

I wonder, someone before had posted changing the max hitpoints from 100 to 1000, and it having a positive affect on the outcome (I think first strikes needed work IIRC). This topic hasn't been discussed in a while. Has anyone tested this lately to see if it improves combat odds accuracy?

 

[Sensemann]

ok i have done another test. this time the warriors have drill i-iv promotion and fight against swordmen. test result:

combats: 700
combatodds: 47,5 %
successful: 347
should be: 332,5
difference: 14,5 (2,07 %)

thats a greater gap, now i'll try it with firststrikechance only.

 

[Adrogans]

I have to say I am loving these test results.

 

[Crazy_Ivan]

So far the test results look normal enough to me, with the standard variance that happens with the randomXrandom effects of the actual combat.

 

[Sensemann]

yeah crazy_ivan is right. but one should calculate how often we have to try, so that the error falls beneath 1% or perhaps less. but here the result for warrios with 10 firststrikechance versus swordmen:

combats: 600
combatodds: 50,8 %
successful: 312
should be: 304,8
difference: 7,2 (1,20 %)

seems ok for me. last test i do, warriors with 5 first strikes.

 

[cabbagemeister]

Something that I know is calculated wrong: if you use go-to from more than one square away, and the attack path crosses a river, the river crossing modifier is ignored in the odds.

I have also noticed this, although I can't tell if it's just the odds display calculator that's leaving out the river crossing, or if it really isn't getting factored into the battle. This might be a good factor to include in the 100 swordsman/warrior tests.

 

[Sensemann]

done. result: (warriors with 5 firststrikes versus swordmen)

combats: 700
combatodds: 50,7 %
successful: 375
should be: 354,9
difference: 20,1 (2,87 %)

mmh it seems, that the firststrike and firststrikechance get little better odds. but to be sure, we have to calculate the error, first. if i have time tommorow, i'll try this.

 

[Sensemann]

I have also noticed this, although I can't tell if it's just the odds display calculator that's leaving out the river crossing, or if it really isn't getting factored into the battle. This might be a good factor to include in the 100 swordsman/warrior tests.

it is displayed wrong. tested it with monks versus swordmen. the combat odds say that there is a 90,1 % chance from 2 tiles away the river and 66,6 % directly beside the river. result:

combats: 500
combatodds: 90,1 %
successful: 339
should be: 450,5
difference: -111,5 (-22,30 %)

and 339/500 = 0,678 => 67,80 % combatodds, its close to 66, 60 %
so its a bug and should be posted in the bug report thread.

 

[Sarisin]

I agree that the 60-70% range for combat odds is no-man's land if you really want to keep that unit. It just seems like you get so many losses in that range, but the tests are saying otherwise.

Well, I started a test at the other end of the spectrum. The results so far have been not quite what I expected, but it is demoralizing in a game sending your troops off to known slaughter.

I have had 25 battles where the combat odds were 10% or less. I won only one of those - it happened to be a Baby Spider with odds of 3.5% against a Warrior. I lost 24 battles. This is early game too (Epic speed), so no fireballs and other throwaway summons.

Yes, you do win some unexpectedly as you throw cannon fodder at the enemy, but so far only one victory for the underdogs.

 

[DrStalker]

I have replicable proof that first strikes are not always handled correctly when calculating combat odds.

Edit the XML so that warriors (strength 3) have 500 first strikes.
Use World Builder to create warriors and a Margalards (strength 22 orangutan , no special abilities) on adjacent grassland

The combat odds will display as <0.1%, but out of 100 attempts I won every one of them.


This is an extreme example, but it was designed to demonstrate a discrepancy in the way combat odds are calculated when first strikes are involved.

 

[eerr]

I agree that the 60-70% range for combat odds is no-man's land if you really want to keep that unit. It just seems like you get so many losses in that range, but the tests are saying otherwise.

Well, I started a test at the other end of the spectrum. The results so far have been not quite what I expected, but it is demoralizing in a game sending your troops off to known slaughter.

I have had 25 battles where the combat odds were 10% or less. I won only one of those - it happened to be a Baby Spider with odds of 3.5% against a Warrior. I lost 24 battles. This is early game too (Epic speed), so no fireballs and other throwaway summons.

Yes, you do win some unexpectedly as you throw cannon fodder at the enemy, but so far only one victory for the underdogs.

on most of the easier difficulties, you auto-win the first time you attack the barbarians
(not the first defense against them, however)

 

[TheJopa]

You right Kael and how often do we attack at 0,1% chance of winning. I did some time ago in the original forum some testing about this (this is an old debate) and the figures that I came up with was all within a reasonble range. I tested all sorts of odds, even 1% chance of winning battles and I came pretty close (3 wins out of 100). The only types of battles that fell way out of a reasonble range was when I tested naval battles. But I was so bored with the testing (I did the naval tests last) so I didn't redo those tests to see if the test result came out the same.

And lastly. It is boring to loose a hero when he has 99% of winning but have laughed more than once when it is the other way around (and yes I have experienced that more than once as well).

I also noticed one another thing: When attacking with fireballs (and earlier, meteors), I, of course, just throw them in at any odds, even as low as 0.1%. What I have noticed is that I *WIN* many 0.3% or 4% fights. Anyone else noticed this?

 

[Minor Annoyance]

it is displayed wrong. tested it with monks versus swordmen. the combat odds say that there is a 90,1 % chance from 2 tiles away the river and 66,6 % directly beside the river. result:

combats: 500
combatodds: 90,1 %
successful: 339
should be: 450,5
difference: -111,5 (-22,30 %)

and 339/500 = 0,678 => 67,80 % combatodds, its close to 66, 60 %
so its a bug and should be posted in the bug report thread.

You didn't need to do 500 combats to find that out. I did an attack from 2 tiles away that didn't list a river crossing and then looked at the combat logs and saw the river modifier. Also it's in the standard civ:4 too.

 

[Kael]

it is displayed wrong. tested it with monks versus swordmen. the combat odds say that there is a 90,1 % chance from 2 tiles away the river and 66,6 % directly beside the river. result:

combats: 500
combatodds: 90,1 %
successful: 339
should be: 450,5
difference: -111,5 (-22,30 %)

and 339/500 = 0,678 => 67,80 % combatodds, its close to 66, 60 %
so its a bug and should be posted in the bug report thread.

Yeah, in my mind thats not a big deal. Im more concerned with combat odds of units that are right beside each other, "Long distance" combat odds wont take river crossing into account (afterall how does it know for sure if you'll cross a river along the way).

 

[Minor Annoyance]

Yeah, in my mind thats not a big deal. Im more concerned with combat odds of units that are right beside each other, "Long distance" combat odds wont take river crossing into account (afterall how does it know for sure if you'll cross a river along the way).

Well it does draw the line for you that crosses the river, but I've made my peace with that having the pathfinder connected to the combat system is probably more trouble than it's worth.