Please, fix the combat odds!

[Fosse]

Round I:
All of my axes won at 98.8% odds.
All of my swords lost at 1.2% odds.
42 of my Infantry won at 50% odds.

Round II:
One of my axes lost.
One of my swords won.
55 of my Infantry won.

And in both rounds all of the rest of my units died without the enemy taking a scratch!

I generated 5 great generals.

 

[Mr. Do]

Fosse, can you elaborate upon, "And all of the rest of my units died without the enemy taking a scratch!"? That sounds very weird...

No. Not sounding good at all. You win WAY too often. 17 victorious battles... and you lost two that are roughly 1/8 and 1/7. According to those odds you ought to have lost 1 for every 7.5 you've fought. You won 17 plus these 2. That makes 19. 19 battles and 2 losses. And you OUGHT to have lost about 1 for every 7.5. So you OUGHT to have lost 2.5 battles. But only have lost 2. That's... broken!

I've not been noting down the battles where I lose on odds of 70% or less. Also currently I had taken part in about 20 battles in total, yet the chances of losing a 1/8 battle followed immediately by a 1/7 battle is 1/56. Have to wait and see what happens from here. Unfortunately the game decided to crash right after the last post, so I haven't kept going for now.

You see, the only way the results posted so far make sense is if you assume that every single one of those battles has also been at 80-90% odds.

 

[PennHead]

Setting aside all discussion of selective memories, pointing out 95% means you lose 1/20, etc, if there have been losses at 100% as a few have reported, that means there is a problem with the odds. 100% means 100%. I'd imagine that it's just really close to 100% and being rounded up, but reporting a 100% chance when you can lose is wrong.

 

[Lord Parkin]

Yeah, I agree with that. 99.95% should not be rounded up to 100%, it should always be rounded down to 99.9%. The game is wrong to report anything as 100% if you are not guaranteed to win EVERY time, regardless.

By the way, here's one of my own tests which I did a while ago, sort of along similar lines. Not quite as extensive, but still useful results to compare. I put 100 Axes against 100 Swords. Results attached (73 wins for the Axes, at 74.9% odds).

 

[Hey Joni]

I appreciate all tests that have been suggested, thanks for the effort! At least for me this thread is becoming educating.

Last night I ran a lot of tests, and, seeing other posts, I can agree that the 100 vs. 100 method works as it should.

However, I have one objection: this test has all the battles happening in one single turn. And the games we play take place in multiple turns. So I would like to see a test that has the 100 vs. 100 units but continues for, say, 50 turns. That means there are going to be only 2 combats per turn, for 50 consecutive turns.

I thought about the best way to do that, but couldn't think of an easy one. The best thing that I thought of is using hotseat but I don't know whether you can access the world builder in hotseat and I was already tired, so I left that for tonight... I'm taking suggestions for such a test, or even just feedback on my logic.

If the idea sounds stupid, have in mind that I am not a programmer, just a curious fan of the game.

As for those who have lost 100% odds - it never happened to me (yet) but if you fought first strike units, they might have reduced the odds and then won. If, however you fough non-first strike units, its a bug.

 

[MrCynical]

As a matter of fact you can't access worldbuilder in hotseat, but I ran the axes vs. swords test by the cumbersome method of popping in and out of worldbuilder for 50 turns.

Results:

Presented odds of axes winning: 74.9%

Axes win: 72

Swords win: 28

Doesn't seem to be anything in the idea that multiple turns are of any relevance, nor indeed any real reason to think this would be a factor in the first place. Given this takes vastly more time than single turn tests, and there is no visble effect, or reason to expect an effect, on the results, I'll go back to the normal method.

As to first strike units, the combat odds are supposed to take this into account. It didn't work properly in ealrier versions, but I think they fixed this for warlords (and in any case, they're very unlikely to give 100% odds. With the old problem they simply overran to give odds of greater than 100%). I can't say I'm that bothered by the losing at greater than 99.95% issue. It's rare, and fairly obvious that it's a rounding issue.

 

[werdna]

What difficulty level did you try it on?
I don't have problems with odds on lower levels (prince or noble), but on emperor I lost a lot 90%+ battles and most 70+

 

[Hey Joni]

Wow, MrCynical, it must have taken an eternity!

OK, FOR NOW I declare that what happened to me was very unfortunate, instead of a bug, cause it's stupid to argue against pure fact (not to mention all the math zealots that jump on you ). This doesn't mean I'm giving up, I won't stop here and intend to make a survey of all combats in my next game. I will be fully convinced only when I see it by myself.

It will surely be a spreadsheet with columns for different odds and their results... This will definately be a boring job but if someone is willing to do the same in the name of justice, pure curiosity or whatever, please post your result here. (I sure wish now was the time to do my statistics report, instead of some time ago, I would have a great topic ).

 

[kcbrett5]

As for your second point - it proves nothing! Winning often at tiny odds should not compensate you for losing often at great odds! Though I agree that I win on 0.3% a lot more often than I should, this doesn't mean that the odds are displayed correctly. In fact, this is just another illustration of my point - the odds are not the measure for victory.

But why should the odds guarantee you victory? I dislike even being shown the odds. Why should you know in advance that you are guaranteed to win? Life is far more random than that. Part of a good strategy is to have enough units to overcome a bad result. Or you can gamble and hope you win. But reloading if you lose then is just cheap. Don't come back later and tell me you always win on emperor level if you are just reloading everytime you lose an important battle.

 

[kcbrett5]

Make another example. Toss a coin. Write down the result. IF you get head-tails-head-tails-head for all eternity, then write a new post. However, if you by chance happen to get head-head-head-head-tails... that's bound to be a broken RNG in nature, right?

Oh boy. Don't start arguing about heads and tails on this forum. You will blow their minds!!!!

 

[Buffbot]

I'm not really sure why Firaxis keeps a huge chance-based model for their battles. The only reason given by proponents is that it gives an outplayed and outmanuevered civ a chance to win by getting lucky. I hate gameplay mechanics that are designed to give losing sides a chance to win due to luck. Something like a pure hitpoint based mechanic found in RTS games would be better, imo.

I can think of another pertinent and logical reason for it. NUMBERS. In Last Samurai Tom Cruise is telling Katsumoto about Thermopylae where 300 men held off an army of a million.

Now in Civ, the odds aren't gonna be THAT bad, but given enough troops, the larger numbers even when significantly outgunned will eventually kill the defenders through attrition.

Now in your example that wasn't the case, but after looking at the combat model and how they figure battle odds, if you have even ONE attacker advantage, if the defender gets hit at all, that will adjust the odds for any subsequent battles immediately.

I use that specific fact to assault max fortified, walled longbows with warriors or other non-iron/copper troops. Given the numbers I need in the stack, it's a bloodbath for me, but if I get the computer hit once it makes the job easier & easier until the city is mine. It's even easier if I can catapult the city defenses to zero beforehand. I have watched the odds go from 0% to 30-40% just from taking down the city defense %.

While it sucks to lose when the odds are grossly in your favor, I have learned to deal with it. (Usually by using colorful metaphors and creative invectives )

Since I am still mired in Noble level barbarian horde management, I frequently get to enjoy the long odds and wastelands that occur after the barbs come strolling through my lands. Maybe I need to post a thread on how to deal with them!

 

[blitzkrieg1980]

One quick thing. If you're probability display shows 100% it doesn't matter if it was rounded. You SHOULD NOT lose 2 battles that closely together. EXAMPLE: If the probabililty was actually 99.95% and was rounded, this means that 1 in 2000 battles with these odds will be lost. Losing two battles with these odds means that 4000 battles will be fought during the game. I think the battle engine should be corrected so that anything with 99.9% and higher should be rounded to 100% (not just the display, but the actually battle variable.

yet the chances of losing a 1/8 battle followed immediately by a 1/7 battle is 1/56.

This isn't actually mathematically sound. One battle has no bearing on the next when it comes to overall odds. Each battle separately has its odds. 1/8 battle loss means that the next 7 with those odds should be won. 1/7 battle loss means that the next 6 with those odds should be won. The two occuring one after another being 1/56 isn't mathematically sound. By this theory, the 2 "100%"(99.95%) battles being lost in a row would have a chance of occurance equaling 1/4 million when it is actually 2/4000 (= 1/2000).

But then again, its been so long since I've done math!

 

[Civ4rulesTH]

OK. This was ridiculous. Six battles in a row on the same turn. For all six the chance was 95% or more. I lost every single one!!! Considering I should of lost 1 in about 970 battles with these odds it should of been 6 out of around 5700!!

 

[Nay]

It´s all about chances.
You win some, you lose some.

Last game on Emperor, we (2 humans against 8 AI) we got..... 13 techs from huts.
Including Metal Casting.

In other games i get 1, maybe 2.

Ive lost too many City Defense Riflemen in a City to Cavalry to think about it anymore.
It happens.

 

[Neo Guderian]

Buffbot has it right.

Odds should have little to do with your campaign planning.
There are two schools of thought when it comes to battle. Superior Quality versus Superior Quantity.

I tend to be a disciple of Quantity when it comes to civ as I set up Heroic and Ironworks in the same city because I want TONS of units.

When you decide to go on an offensive campaign, bring overwhelming numbers and always expect poor odds. This way, you won't be disapointed when you do have a run of bad luck.

Alot of builder types have issues with fielding armies of this size as the tendancy is always to build buildings before troops (I know, I suffer from this tendancy). In order to have the quantities I know are neccessary to deal with my enemies, I set aside 2 or 3 cities that do nothing but build units the whole game.

Combat odds should be an after thought to your well laid operational plan.

 

[Colin_McK]

Just like hey Joni, This happens too often for coincidence, so I save the game before I go into combat, especially with my highly promoted general units.

Once, when attacking an archerd city with my general swordsman, +10 combat, +20 city attack, +25 city attack +30 city attack +10 gunpowder units. 70.5% chance. Forgetting to use my catapults I attacked and lost.

Cursing, I reloded the game, and reduced the defences with my catas. 99.5%. Surely I must win now? nope. Cursing again, I decided to attack with my other 2 swordsmen. 90.5% for both, and both lost.

To fix this, just like judge d I just reloaded and waited a turn. Next time, all three units quite rightly won.

A good tip, sometimes it is worthwhile, as shivute said (especially playing the barbarian mod and the enemy has musketeers) to 'sacrafice' a catapult by throwing it at an enemy city knowing full well that it will be lost. Its collateral damage can be the tide-turner.

 

[Mr. Do]

This isn't actually mathematically sound. One battle has no bearing on the next when it comes to overall odds. Each battle separately has its odds. 1/8 battle loss means that the next 7 with those odds should be won. 1/7 battle loss means that the next 6 with those odds should be won. The two occuring one after another being 1/56 isn't mathematically sound. By this theory, the 2 "100%"(99.95%) battles being lost in a row would have a chance of occurance equaling 1/4 million when it is actually 2/4000 (= 1/2000).

But then again, its been so long since I've done math!

The main problem is that the way my post is written, it could be taken in two ways, one which is correct, one which is not. You chose the latter I was stating that the chances of [1/8 then 1/7] is [1/56], not that the chances of [1/7] after [1/8] is [1/56]... that would be silly!

Your maths are actually not sound: if a 100% battle is indeed taken as 99.95%, then that's the same as 5/10000 or 1/2000 chance of losing... so EACH battle's chances of losing are 1/2000, and if you set up a test, then the chances of both battles being lost is 1/4,000,000, you don't just add them together!

 

[Kartik]

Indeed, I lost three grenadiers to 2 chinese longbowman and a maceman, I was so angry, not to mention the grens had upgrades, I just quit after that.

 

[DemonMaster]

And (as mentioned earlier) Longbowmen have "First Strike". I don't think the AI calculate odds heaving FS in mind and FS can be extremely lethal.

 

[Reg Pither]

Just thought I'd add a couple of recent experiences to the pot...

I lost a General unit when they odds were 98% in his favour, but then had a catapult defeat a unit defending a city when the odds were only 0.2%! Got 9 XP for that kill, too.

So I suppose it all balances out in the end, and if there were no risks the game would be a lot less fun, IMHO.