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

[Dis]

shouldn't 75% mean 75%?

I've kind of accepted the fact that I'm going to lose units when I shouldn't. I lost 8 cavalry my last game. I should have only lost 2. And that was when I attacked the grenadiers. The other losses came against samuraii, knights, and crossbowmen. I guess it's like the movie The Last Samurai. . Although it should be noted that technology did prevail in that movie at the end. (slight spoiler )

[Soren Johnson]

there are no AI combat bonuses vs. the player.

[Shaka II]

Thanks, it's what I thought. I seem to get as many good breaks as I get bad ones. The new probability indicator is great BTW.

Everyone has their own story of a heroic victory, here's mine:

My Quechua popped a goody hut, expecting a good surprise. Instead, out popped four barbarian warriors. It was a hopeless situation. As I waited for his defeat, I was pleased to see that he at least won the first battle. But then, miraculously, the jungle grew in where he stood, giving him a 75% defensive bonus. he won battle after battle, until all the barbs were gone. A truly heroic victory!

[Solver]

Well, if 75% is the chance to win this specific battle, it doesn't mean that 75 of 100 battles will be won. Randomness works in strange ways, and runs are common. In practice, you'll see 60 wins, or maybe 80 wins, but very rarely exactly 75.

[Lambiorix_be]

Originally posted by Solver
Wow... some people don't know what odds mean.

If your unit has a 75% chance of winning, it doesn't mean that your units would win 75 out of 100 such battles. Winning 60 of 100 would be significantly more probable!

I agree that you will not have 75 on 100 tries with odds of 75%. The more you try the more it will become like this say if you tried 1000 times with 75% chance of success you would have something between 700 - 800.

I disagree with the statement that with 75% odds a result of 60 out of 100 would be more probable. If the odds are 75% it would be as probable to have 90 of 100 as to have 60 out of 100. If the variation is always below the 75 on 100 then you may be sure that you do not have 75% odds.

But to test this you would need a lot of independent players that all do the same action the same number of tries. If you would compare the results of all those players then about half should have a variation above the 75 mark and half should be below the mark.

[MightyTiny]

Originally posted by Solver
Well, if 75% is the chance to win this specific battle, it doesn't mean that 75 of 100 battles will be won. Randomness works in strange ways, and runs are common. In practice, you'll see 60 wins, or maybe 80 wins, but very rarely exactly 75.

True, but a probablility of 75% success rate should mean that the more repetitions of the event, the closer the avaerage success rate of those events should come to 75%. Ten thousand events, and pretty close to exactly 75% of them would be successes.

So in the light of that, the numbers reported by Lansing do seem a bit unusual; especially since they consistently came out in the AI's favour, and the number of tries was large enough that I would have expected results a little closer to 50-50.

Lansing didn't say, so I'm wondering if he also took into account possible promotions - were your units and the AI's units exactly equally experienced in your experiments? What about the TYPES of promotions? Were they identical?

If experience, and the types of promotions was taken into account too, then perhaps this test should be repeated by someone. As the developer says that there's no biasing attack bonuses programmed in for the AI's, so I take that to mean that if there are such biases in practice, they are not intentional. Or perhaps Lansing's results were just a statistical fluke... maybe some math wiz can find the patience to calculate the probability of that. I can't.

[gdgrimm]

In addition to the effect of remembering the unusual and traumatic, but forgetting all the normal and expected, there's another factor that can mislead a player into thinking combat odds are rigged.

As a player, I almost never attack with less than a 60% of winning. In fact, I usually want a 70%+ chance before I attack. So on offense, I greatly increase my exposure to 'lose when you should win' and really never have the opportunity to 'win when you should lose'.

The end result will be that "most" unusual outcomes will be against the player, since the player doesn't even attempt the scenarios that would provide a beneficial, unusual outcome. That will leave one with the impression that the random factor of combat is against the player.

[Solver]

Yeah it's interesting... I guess the best way to do tests is different.

With the World Builder, give yourself 100 Knights and the AI 100 Knights. They don't get defensive bonuses, so the odds will be exactly 50%. Then see how many battles you win... the deviation from 50 there shouldn't be too large indeed.

[johnmcd]

If the AI gets a production bonus the further up the difficulty ladder you go you'll get to a point where you simply can't produce enough units to crush his ants underfoot in a turn.

There needs to be a second modifier to AI units to make up for the fact that more and more of them ceases to be an option at some point. The combat NEEDS to be rigged at the top levels to keep the game playable as otherwise their pillage code would be teamed with enough units to create a locust like invasion.

I do feel that any hidden strength modifier at the harder difficulties should be shown though.

[Solver]

The modifiers are in the XML files, they aren't hidden. Sure the AI can produce more units than you at the highest difficulties, but the human is always smarter than the AI, thus you should really be able to use your units in a better way.

[Krill]

But the human is always smarter than the AI, thus you should really be able to use your units in a better way.

I would disagree with that statement...

[TheHateMale]

I have utterly forgotten how to perform a binomial probability test, and it's been so long that I can't understand the mathematical notation for it anymore.

Could someone either find the chances of the tests happening, or preferably find a plain English version of it so I can program a quick simulator for it? I'll happily put in the time to find the significance of the results in the original post.

[johnmcd]

Originally posted by Solver
The modifiers are in the XML files, they aren't hidden. Sure the AI can produce more units than you at the highest difficulties, but the human is always smarter than the AI, thus you should really be able to use your units in a better way.

I'm nerdier than most and there is no way I'm rooting around in XML files to find out whether or not my perception of an advantaged AI is right and if so by what margin.

My point is really that there comes a point that it doesn't much matter how much better you use resources if those resources are simply overwhelmed, as will happen if only the production values increase. A farmer can out think any swarm of locusts, but there is no way he can swat them all.

[xxFlukexx]

Are you sure about that Soren? Then maybe the RNG is broken?



I shouldn't have a folder that I just created for these and already up to screenshot #15. Ok maybe I lose that battle 1/10 times and maybe I don't get many shots in, but never a shot in? And this happens regularly? Sorry, but when it takes 4 axeman to kill an enemy swordsmen with my advantage always greater than 50% I get suspicious.

[Lansing]

Originally posted by Soren Johnson Firaxis
there are no AI combat bonuses vs. the player.

Then something is massively screwed up. Too many people I know have noticed this, and I have received too many anomalous results (including in this test) for it to be just one long stream of player-killing variance.