Removing the randomness

[Bringa]

I think what I will do in the future is maintain a second version of Smarter Orcs with the fair combat settings enabled by default so people can experiment with that without having to change anything in their xmls. I've yet to hear from anyone that they tried the fair combat settings and went back to vanilla because they were missing the randomness, but then again not enough people have played with the fair settings.

 

[Chip56]

Actually he does get more damage on average.
Here you can get a combat calculator where you could change the hp even to 1000 (however you shouldn't set it to more enemies than 3 when you use high amounts of hp otherwise it hangs up):
http://c4combat.narod.ru/c4c.htm

And here a example why you get more damage on average:
strength 5 vs strength 5 both 100hp:
of course 50% win chance for both since they both deal 20 damage and have 50% chance to hit. Average Hp left if you win:49.21875
Explanation:
you have a:
3% chance to get 0 hits->100hp
7,8 to get 1->80hp
11,7 to get 2->60hp
13,7 to get 3->40hp
13,7 to get 4->20hp
results in average 49hp. remaining thats 49% of the hp you beginn with.

Would both units only have 20 hp to beginn its a 50% to win
but if you win you would remain with 20 hp always.-> average damage on victory 0!

edit: One more example : both units with 40 Hp:
Chances to win again 50%
CHance to get damage:
25% to get 0->40hp
25% to get 1hit->20hp
so the winner has an average of 30hp remaining.->thats 75%
end of edit

You can to the math for every amount of hp yourself (or you could just use the calculator, i tested the results they are correct) and you will see that the average damage on a victory increases with hp because the battle lasts longer and the chances to get no or min. damage decrease rapitly.

 

[Vulcans]

i tried the link you provided.

Ok, st3 vs st4

With 100hp system
Chance to win=78%
Average hitpoints left=54

1000hp, both units, strength 30 and 40
Chance to win=99.8%
Average hp left=43.9

I’d much prefer 99.8% than 78%
Also the “average hit points left if win” is a bit misleading. it decreases as you are having more chance of winning, because you are also averaging over times you only just win where you would have lost otherwise.

Eg if you are about 90% lucky in your rolls: in the 100hp scenario you would lose, so the battle wouldn’t be counted in “average hit points left if win” statistic, while in the 1000hp scenatio you might be lucky and just scrape by with 5 hp left, this is included in the “average hit points left if win” therefore skews the figure of “average hit points left if win”, reducing it because you just scraped by with a win, which is better then losing and not being counted in the statistics of "average hp if win".

in situations you lose you have 0 hitpoints. Therefore if you wanted to compare damage received you would need to have

average hit points after battle="chance to win" * "Average hp if win"
so for 1000hp
99.8*43.9=43.8hp
And for 100hp
78.2%*54.34=42.5hp

Therefore the 1000hp scenario you receive less damage in the battle.
It’s just the 21% of the time that you win and just scrape by is included in “average hp left if wins” while that 21% of the time in the 100hp system you wouldn’t have won, you’d be dead, therefore it’s not included in “average hp left if wins”

Obviously this increases the difference between different tier units. A 4st unit is nearly guaranteed victory over a 3st unit. I thought one of the reasons that gave all units +1st was to decrease the difference between different tier units.

the more i look at it the more it looks like just changing the balance of the game to favour the stronger units (which is normally the humans has as we don't waste time creating useless units, and focus on heros).

 

[Bringa]

Vulcans, you're still missing the point. This is not the same as changing the system to make certain units stronger at all. I think the easiest way for you to appreciate this change is to just testdrive it. I posted a fixed xml on your FFHOTM thread; if you use that instead of smarter orcs (you can always go back to unchanged smarter orcs by installing it again) and run a game, you'll quickly feel the difference. You're looking at single combats here; we're looking at the whole game. In FFH as it is now, your hero is useless until he gets a certain set of promotions (let's say shock I & II vs melee units), at which point he becomes invincible. With fair combat, the hero is useful from the beginning, but will never become invincible. Combats always hurt him and superior numbers always threaten him. This does require quite some different playing; mostly it requires you to play much more strategically.

But as I said, talking about the theory is pointless. Fire up a game and try it out. You'll see.

 

[Chip56]

average hit points after battle="chance to win" * "Average hp if win"
so for 1000hp
99.8*43.9=43.8hp
And for 100hp
78.2%*54.34=42.5hp

Yep you are right if you see the whole it means that a much stronger unit will be a lttle bit better in a single battle.

However it will be fairer in following examples:
5.01 vs 5
normal 100hp combat:
62.4% winchance for the stronger unit
average 42,6% hp left after win
so 62,4%*42,6=26,571%hp

1000hp combat:
62,1%
average 16,97% hp left after win
so 62,1%*16,97=10,53%hp left
So units that are only a little bit "better" are now more likly to loose.

 

[Chip56]

Oh and one more thing since you are guarented to get damage you will be weaker in the next battle. In the old system it is very likly to get no damage and thus fight the next battle with full power.

So if you are "95% lucky" with your strength 4 unit in the old scenario you will fight with 100% hp (you have a 6% to get no damage) in the next battle:
that means your victory chances for the second are 78%

in the 1000hp scenario the "95% lucky" will reduce you to approx. 70% hp.
And that means in the next battle you will fight with 70% and reduced strength and then your chances drop really fast.

 

[xanaqui42]

I posted a fixed xml on your FFHOTM thread; if you use that instead of smarter orcs (you can always go back to unchanged smarter orcs by installing it again) and run a game, you'll quickly feel the difference. You're looking at single combats here; we're looking at the whole game.

Two comments:

1) If you give me a link to said file, I'd be happy to link it to the main Smarter Orcs thread. EDIT - merged into a different modmod branch!
2) Your file should be as compatible with the most recent release (0.15) as the last one (0.14); there were no XML-relevant changes in 0.15.

 

[Bringa]

Right! Pm coming!

 

[Lord Yanaek]

OK i'll test the new mod and tell you really how i like it (or not, but i guess i will like it)

 

[Vulcans]

Yep you are right if you see the whole it means that a much stronger unit will be a lttle bit better in a single battle.

“Slightly” It is a HUGE difference!!!!!!! From 21.8% chance to losing the battle to 0.16% chance of losing the battle. That’s 136 times the chance.!!! Probability will mean it’ll last 136 times more battles, and in that time the unit will level and get even stronger. After those 109 battles it’ll be even more unbeatable. And the AI doesn’t know how tocompensate for it, it will just throw more cannon fodder not realizing it doesn’t have a chance.

I donn’t know about you, but I don’t want my st4 unit to be guaranteed victory against a strength 3 unit. It takes the fun out of it. I think 78% chance of victory sounds more reasonable then 99.8

However it will be fairer in following examples:
5.01 vs 5
normal 100hp combat:
62.4% winchance for the stronger unit
average 42,6% hp left after win
so 62,4%*42,6=26,571%hp

What type of example is that? I didn’t even know there was a 5.01st unit. The difference is insignificant, and barely noticed in calculations. That is a mutual destruction scenario. Hence in mutual destruction scenario when you are doing large amounts of damage per hit then each blow does more, so if the survivor wins by 1 blow, then it’ll have more leftover hitpoints. While if each blow does less damage, then the resolution is greater, so in a mutual destruction scenario the survivor will e nearly dead.

Oh and one more thing since you are guarented to get damage you will be weaker in the next battle. In the old system it is very likly to get no damage and thus fight the next battle with full power.

In one situation with larger amounts of damage per hit (eg 15 damage per hit) he could possibly he could have 50% chance of 0 damage, and 50% chance of 15 damage. On the other side, with finer resolution he could have 5, or 6, or 7, or 8 damage. Sure, he has a chance of not getting any damage, but he also has a just as much chance of receiving one blow, which does more damage then the average damage he would receive in a 1000 hp scenario.



There are two unwanted factors in combat. One is the significant jumps which come from changing the number of hits required to make a kill.
http://www.civfanatics.com/civ4/strategy/combat_explained.php
this causes jump points in probability when the number of required hits to make a kill changes, and also the number of round losses you will survive through changes. This is unwanted, and I support the project to remove these significant jumps (eg huge change in combat odds compared between 1.38 and 1.39 ratio)

then the other factor is the probability of a weaker unit beating a stronger unit (I think you shouldn’t have nearly guaranteed victory unless you’re near twice the strength, not when only being slightly stronger)


the number of combat rounds drastically changes the effect of randomness.

Let’s take a 6 sided die.
The chance of getting a 5 or higher=2/6=33%

Now increase the number of rolls (combat rounds)
The chance of getting 10 or higher on 2 dice
=4/36=11.11%

The chance of getting 15 or higher on 3 dice
16/216=7.4% chance

The chance of getting 20 or higher on 4 dice
32/1296=2.47%

See as you increase the number of rolls (combat rounds) you change the probability of winning/losing. Sure the 5vs 5 is the same, it’s a 50-50 chance, but it changes the graph of how fast the odds change with combat ratio changes.

What you are effectively doing is making battles more Boolean, which unit is stronger will win.

the AI can not adapt to the new combat rules. An AI strength 3 unit will attempt to take on a strength 4 unit, as it knows it has 21.8% chance of winning, which is worth a go if someone doesn’t have a reserve defending the city.

But a single 3st AI unit won’t normally bother attacdking a single st6 unit because the chance of success is under 1%. So the programmers have made the AI evaluate opponents on the standard combat system, st3 will attack st4, but won’t bother attacking a st6.

Now changing the combat to a 1000hp system changes the odds of a 3Vs4 battle to the same odds of a 3Vs6 battle in the 100hp system. But the AI doesn’t know about this change to make combat more Boolean. So it’ll attack anyway thinking it has a chance, but not knowing that it’s total suicide.

The combat can not be fixed by increasing the number of rounds!

using floating points or other methods could be tried, but the number of rounds should not change. otherwise we change the whole combat system

 

[xanaqui42]

There are two unwanted factors in combat. One is the significant jumps which come from changing the number of hits required to make a kill.
http://www.civfanatics.com/civ4/strategy/combat_explained.php
this causes jump points in probability when the number of required hits to make a kill changes, and also the number of round losses you will survive through changes. This is unwanted, and I support the project to remove these significant jumps (eg huge change in combat odds compared between 1.38 and 1.39 ratio)

Note that Smarter Orcs has poorly-tested functionality that is designed to attack this issue. If in GlobalDefinesAlt.xml, you alter COMBAT_DENOMINATOR to a value greater than 1 (and no other changes), you get greater accuracy in damage during combat.

The reality is that this change helps only marginally, even with a fairly large denominator (say 10). The number of rounds still predominates; this change does little other than make slight shifts in the position of these jumps, and pushes the average HP of the victor in these cases towards 0.

Do you have a suggestion for what you think might "fix" combat?

 

[xanaqui42]

Now changing the combat to a 1000hp system changes the odds of a 3Vs4 battle to the same odds of a 3Vs6 battle in the 100hp system. But the AI doesn’t know about this change to make combat more Boolean. So it’ll attack anyway thinking it has a chance, but not knowing that it’s total suicide.

Note that the AI has been modified to take into account the various XML changes that Smarter Orcs supports in combat (oddly enough, that was one of the easiest parts of these changes). If there's an issue with this, that's a defect.

 

[Chip56]

I didn’t even know there was a 5.01st unit

Okay than one actual example rather than a theoretical (would have been better to say 4.95 vs 5 because thats possible):
Abashi the Black Dragon : 20 strength
Basium: 10 strength +combat 5= 20 strength

Thats a nice 50% chance for everyone.

And now basium with only 99HP instead of 100HP (got some damage and then healed to 99 thats possible)
Basiums strength is now 20*0.99=19.8
So hit chances are now: 19.8/39.8=49,74% for basium and 50,25 for Abashi
But thanks to the way damage is calculated basium will only inflict 19 damage vs the 20 of Abashi.
So Basium needs 6 hits Abashi 5.
->chances to win are 62% in favor of Abashi.
And thats in cases of "equal strength" units.

Now in the 1000Hp scenario:
basium 990HP (99%)
Hit chances the same
Numbers of hits needed:50 to kill basium, 53 to kill Abashi (so only 1,5% more instead of 20%)
So now Abashi has only a 61,6% of winning.And he will take more damage on average if he wins: 15,9% instead of 41,5%HP remaining

The difference is insignificant, and barely noticed in calculations.

this causes jump points in probability when the number of required hits to make a kill changes, and also the number of round losses you will survive through changes. This is unwanted, and I support the project to remove these significant jumps (eg huge change in combat odds compared between 1.38 and 1.39 ratio

^^^My example was exactly because of jumping points. In my case it was a 20% increase of hits needed to win.

it will just throw more cannon fodder not realizing it doesn’t have a chance

And the second unit has HIGHER chances to win because the hero will be damaged-> reduced hitchance,damage and less HP.
To prove it our good 4vs3 example:
Old system:
6% for 100hp
13% for 83hp
16,7% for 66hp
16,7% for 49hp
14,4% for 32hp
11,1% for 15hp
and the 21% to die

Now the new system I will write down how much hp you will have remaining instead of those results. Instead of the best 6% of the old system i will write down what you would have got.
5,3%: average of 68,8% HP remaining
notes:

Spoiler :

if you want to check the results the last 1,55% of those 5,3% is(are?) hit number 21.
Seems the 70% guess in my earlier post wasn't that far of

10,68% average of 59,7% HP left
notes:

Spoiler :

I only cover the best 16% instead of 19% of the old system.Thats because the next hit (26) would add 3,8% and I dont want that you call me cheating because I took number to "early". However I think those numbers already show pretty nice that you WILL be much much more likly to loose the next battle. In more than 94% you will have less than 70%HP left.

19,6% average of 50% HP left
notes:

Spoiler :

Now i include the missing 3%
If you want to check:Last amount of hits included:31
We now have 35,58% of cases covered and every single one the hero will be MUCH weaker after the battle in the 1000Hp case in relation to the 100hp

15,6%: average of 43,9% HP left
notes:

Spoiler :

51,18% of possible results covered and he is still weaker in everyone.last amount of hits:34

13,63%: average of 38,9% HP left
notes:

Spoiler :

And now we come to the point where he actually gets better results in the new system. And that will not chance with the next 35% so I dont post these results here.The remaing 0,17% is death and thus the same

To end my post:
70% of HP remaining results in:strength 2.8 instead of 4
->hitchances against a new 3 strength unit only 48,27% instead of 57,1
and damage now 21 vs 19 (or 21 vs 18 depending when civ4 rounds) instead of 23 vs 17. Note that in the current version of civ the damage is the average of damage when max hp and damage at current hp.
And now our imba hero will need 47 hits to win but he will be dead after 37 (or 39). Good luck with winning the second battle. You will need it because now our "stronger unit==imba" system works against the hero.

 

[Bringa]

Would you guys just stop with the numbers? The situation here is too complex for abstraction. Just play a few games to get a feel for it.

 

[Vulcans]

there are TWO issues here.

I agree with you about the jump points, and how 1.00000000001 ratios should be close to 50% probability. I agree that this jump point prtoblem should be solved. the mechanics of jump points have already been described in the article i cited, and i don't see the point of continuing stating what's already been researched.

the other issue is the combat odds, and the chances of success with an uneven battle. and how changing the number of rounds changed the statistical probability. having a winning streak of 7 out of 10 hits is much more probable then a winning streak of 70 out of 100 hits. hence you are changing the probabilities and so how the different strength ratios relate to each other.

at the moment the odds of losing a 4Vs3 Battle are 21.8%
personally i think that sounds quite reasonable for the strength difference.

Now changing to a 1000hp system will change the odds of losing down to 0.16%. chance of losing.
So you’re 136times less likely to lose.

That is similar to a 7.4 Vs 3 battle in the 100hp system.
Chance of losing is 0.33%

So what you’re doing is effectively turning yout 4st unit into a 7.4st unit.

Hey, let’s turn our hunters into rangers. I could do with rangers moved forward in the tec tree to hunting.

And let’s turn our axe men into mace men with combat promotions.

And let’s just keep this our little secret between us humans and not tell the dumb AI.

 

[Vulcans]

and i care more about winning/losing the battle then if i have a few more hit points after a battle. after the battle there is often a promotion healing the unit, or you can always wait and heal, but if he loses the battle then he's gone, can't be healed, and you'll need another unit in the stack to finish the job attacking before the opponent heals from a promotion.

it's more important to win a battle, and whole discussions about if a unit has a few more hp left after winning a battle seem somewhat trivial compared to winning/losing.

i see battles more as 1v1, i only attack once with each unit, and have stack defenders/reserves for counter attacks. Therefore I consider battles on a 1 vs 1 basis in most scenarios.

 

[Chip56]

i see battles more as 1v1, i only attack once with each unit, and have stack defenders/reserves for counter attacks. Therefore I consider battles on a 1 vs 1 basis in most scenarios.

If you see it only that way there is no way i can ever convince you that 1000hp are NOT unbalancing.
Because I have to agree that in a single 1vs1 the stronger unit is better off with 1000hp.
My point is that you dont fight wars with only 1unit but with more and then the 1000hp scenario is better balanced.

 

[Vulcans]

If you see it only that way there is no way i can ever convince you that 1000hp are NOT unbalancing.
Because I have to agree that in a single 1vs1 the stronger unit is better off with 1000hp.
My point is that you dont fight wars with only 1unit but with more and then the 1000hp scenario is better balanced.

On the topic of stacks the same thing applies.

When defending: If the AI attack with 100 warriors, and against 100 hunters then not a single hunter will die. There goes 2500 hammers down the drain, without killing a single unit. There should be some small chance of casualty in the hunter stack.

When attacking: attacking is a 1V1 battle, you only care about killing the one unit. as you only do one attacking battle per turn. And with this system you’re guaranteed victory. then you can promo/heal for the next attack. your stack would have defenders, so it is safe from counter attack as above.

That means a stack of slightly stronger units will never take any casualties. That is wrong on a 1V1 level or stack V stack level

Anyway, I’ll have to leave this discussion for now. Please read my post 95 again. At the moment I don’t have time to get into this discussion. The imbalance is HUGE! I can explain it all more later if you want.

 

[Chip56]

warrior: 25 hammers
hunter: 60 hammers

So the Ai would have 2,4 warriors for every hunter (actually even more since they get production boni)
Also the AI pays almost no upkeep for mass units so that doesn't really matter when fighting the AI.
And 2,4 warriors will easily beat 1 hunter in the 1000hp system.
Yeah you will probably loose 100-150 warriors.
But you would have loost approximtly the same amount in the 100hp scenario.

Okay now some comments on your post 95 (before i refrained from it because I dont think we will come to some mutual agrement other than "let it be"):

hence you are changing the probabilities and so how the different strength ratios relate to each other.

Of course we change it. Otherwise we would get the same results, wouldn't we?

at the moment the odds of losing a 4Vs3 Battle are 21.8%
personally i think that sounds quite reasonable for the strength difference.

First I want to state that I respect your opinion and that you like some randomness. Thats fine.
I personally find 21,8% to high.
If I send 4soldiers/tanks/jet fighters/whatever in an open battle against 3 of the same type I dont really expect to loose in 21% all 4.
But yeah thats only my own opinion.

So what you’re doing is effectively turning yout 4st unit into a 7.4st unit

No because a 7.4 unit will have a 25% of getting no damage at all (even worse with drill!) and will have more than 70% in 76% of the possible battle outcomes.

And let’s just keep this our little secret between us humans and not tell the dumb AI.

I dont think the boys that programmed the AI wrote a little table with all possible combat settings and the AI looks at this table instead of calculating the possible results.

 

[Vulcans]

Of course we change it. Otherwise we would get the same results, wouldn't we?

i think we want the same results, just with the jump point discontinuities removed.

i think the suggestion unbalances the game. please feel free to do some calculations to understand how much it changes the balance, i guess i could write up a spreadsheet or something in a couple of weeks when i have more time to explain it all. at the moment i don't feel like explaining it all again in finer detail as i'm in the middle of other things.