Random Numbers

[fuzzykitty]

I am working my way though Bad Moon Rising capaign 2 on Wesnoth 1.6 and it occurred to me that I was seeing very irregular odds patterns.

In one scenario, on a whim, I decided to see just how bad a 70% hit rate was. By attacking and reloading after the first strike attempt I have been going for the past 10min unable to hit. If I let the combat play out neither side hits very much, maybe 30%. No, it's not my perception. I have also noticed that at different times the game will go through rashes with stellar hits, even for 30 and 40% odds. Sometimes several attacks in a row just can't miss.

I realize that odds have been discussed and complaints debunked before, but come on. This looks like a bug in the random number generator or the game is inaccurately reporting odds.

[Zarel]

In one scenario, on a whim, I decided to see just how bad a 70% hit rate was. By attacking and reloading after the first strike attempt I have been going for the past 10min unable to hit. If I let the combat play out neither side hits very much, maybe 30%. No, it's not my perception. I have also noticed that at different times the game will go through rashes with stellar hits, even for 30 and 40% odds. Sometimes several attacks in a row just can't miss.

What a coincidence! This sounds exactly like how probability works in the real world!

[ancestral]

Gambler's Fallacy: "I have also noticed that at different times the game will go through rashes with stellar hits, even for 30 and 40% odds. Sometimes several attacks in a row just can't miss."
If there's a 30% chance to hit and it has hit 4 times before therefore it must not hit the next time? Incorrect, as the next roll is always another 30% chance to hit.

[Yogibear]

Replays, please . If you want us to listen any further, that is.

[HomerJ]

Rodnh.

Greetz
HomerJ

[hiro hito]

What a coincidence! This sounds exactly like how probability works in the real world!

... and how it should not happen during the game!
only 2 turns with big RNG unbalanced can screw the game...

[Noy]

What a coincidence! This sounds exactly like how probability works in the real world!

... and how it should not happen during the game!
only 2 turns with big RNG unbalanced can screw the game...

Troll much?

[fuzzykitty]

What a coincidence! This sounds exactly like how probability works in the real world![/quote]

Umm, no. That's not how probability works. Probability should be readily apparent within a reasonable sample size. Some people argue 13 samples are sufficient while others advocate ~20 assuming you are not looking for subtleties.

The reality is that the so called "random number" feature is not really random and as such it will adhere to it's own distribution. If the random number generator were truely "random" then a unit with a 50% hit chance should hit at a rate approaching 50% after 100 tries. Try flipping a coin in a well controlled experiment and you will find that after 100 tries you approach the 50% probability of heads or tails. That's probability in the real world.

What you are describing as probability is the idea that if something has a 1 out of 10^21 chance of happening, seeing it happen every time for the next 10^21 tries is perfectly reasonable. The problem is that you are confusing a theoretical possibility with reality. Probability is an attempt to apply a theoretical outcome to a real world situation. This is why the afore mentioned condition will NEVER happen in the real world.

[Zarel]

Umm, no. That's not how probability works. Probability should be readily apparent within a reasonable sample size. Some people argue 13 samples are sufficient while others advocate ~20 assuming you are not looking for subtleties.

Yes, well, I see a significantly lower sample size.

You appear to act as if you understand probability. I'm going to assume you've taken a statistics course (which, by the way, you don't need to understand probability to pass). How's about you run an actual statistical hypothesis test on Wesnoth?

Go do that, and report the results. Remember that to reduce bias, you have to predict the statistical anomaly that will occur before you test for it: You cannot run the data, find a statistical anomaly, and then test the same data for that anomaly (it screws with alpha values by lots).

The reality is that the so called "random number" feature is not really random and as such it will adhere to it's own distribution. If the random number generator were truely "random" then a unit with a 50% hit chance should hit at a rate approaching 50% after 100 tries. Try flipping a coin in a well controlled experiment and you will find that after 100 tries you approach the 50% probability of heads or tails. That's probability in the real world.

No, the unit has a fairly high probability of hitting at a total rate of approximately 50%. That "total" is very important - the probability of getting so close to 50% is high only because it is a total over a large sample size. Individually, you're quite likely to see runs like 10 heads in a row, within your 100 tries.

What you are describing as probability is the idea that if something has a 1 out of 10^21 chance of happening, seeing it happen every time for the next 10^21 tries is perfectly reasonable. The problem is that you are confusing a theoretical possibility with reality. Probability is an attempt to apply a theoretical outcome to a real world situation. This is why the afore mentioned condition will NEVER happen in the real world.

Richard Feynman once said:

"I was walking to class today and the funniest thing happened: I saw a car with the license plate ‘ARW 357’. Can you imagine? Of all the possible license plates, what are the chances of seeing that one?"

Further reading: http://cubic.bioc.columbia.edu/~bigelow ... _joke.html

[HomerJ]

There's another popular example. Lets play a game with 32 unique cards. Shuffle. 3 People get 10 cards, 2 remain in the middle (Btw the distribution 3x10 +2 has nothing to with the experiment, just taken from a popular german game called "Skat"). Now look how the distribution turned out. Calculate the chance for this particular outcome. Chances are so low that the distribution you see is practically impossible.

Greetz
HomerJ

[fuzzykitty]

"..You appear to act as if you understand probability.."

Friend, you have no idea. You may think you understand probability but what you show you don't understand is the application of probability to a process and that's essentially what a game like Wesnoth is. If the random number generator cannot show the reported odds within a reasonable sample size then it is broken. Plain and simple. You can't just open a text book and expect to have a solution.

SSI and other hexbased strategy game makers resolved this problem fairly early on to where you don't see this kind of inconsistancy. In Panzer General II, for example, if you are given a prediction of the rough odds you will generally see that kind of performance. There is some variation but total systematic departure for samples >10 just doesn't happen. So I'll restate it: The Wesnoth method of resolving combat is broken.

[fuzzykitty]

"There's another popular example. Lets play a game with 32 unique cards. Shuffle. 3 People get 10 cards, 2 remain in the middle (Btw the distribution 3x10 +2 has nothing to with the experiment, just taken from a popular german game called "Skat"). Now look how the distribution turned out. Calculate the chance for this particular outcome. Chances are so low that the distribution you see is practically impossible. "

Sure, I'll agree. But then it shouldn't happen the next three times in a row. But in Wesnoth it can because as I said, the combat system is broken.

[Soliton]

So I'll restate it: The Wesnoth method of resolving combat is broken.

You realize that that is a very different statement from your claims of the indicated hit chances not being correct, right? If you want to prove those claims a look into a textbook would probably come in handy.

A more accurate statement would be that you don't like the combat resolution system of wesnoth. You could try the lessluck era or sauron's mod and see if you like that more. They're only for multiplayer though.

[Zarel]

"..You appear to act as if you understand probability.."

Friend, you have no idea. You may think you understand probability but what you show you don't understand is the application of probability to a process and that's essentially what a game like Wesnoth is. If the random number generator cannot show the reported odds within a reasonable sample size then it is broken. Plain and simple. You can't just open a text book and expect to have a solution.

SSI and other hexbased strategy game makers resolved this problem fairly early on to where you don't see this kind of inconsistancy. In Panzer General II, for example, if you are given a prediction of the rough odds you will generally see that kind of performance. There is some variation but total systematic departure for samples >10 just doesn't happen. So I'll restate it: The Wesnoth method of resolving combat is broken.

I ask you again to put your alleged statistical prowess where your mouth is, and conduct a statistical study on what's inconsistent about Wesnoth's RNG. Until then, your assertions don't really mean much.

[hiro hito]

You could try the lessluck era or sauron's mod and see if you like that more. They're only for multiplayer though.

... and windows system