Conquer Club

Dice analyzer is broken again.

zip_disk wrote:Dice analyzer is broken again.

damn, was just about to download it.

Is the link to it broken? Or it just doesn't work? (might be the new dice look update. In this case I'll have it working in a few hours, when I get home)

I've updated it, but I couldn't test it for auto-attack. Please tell me if it works for it in case you have a larger battle.

It does. Thanks.

thx for testing that

for some reason, my dice analyser shows that ive won 4 out of 601 1 vs 1 battles. I know that isnt the case. Is it because the dice analyser is malfunctioning, or more likely, my computer is malfunctioning?

I love my dice.

Attacker dice distribution Defender dice distribution
1s █████████████████████████ 198 / 1264 (15.66%) █████████████████████████ 96 / 622 (15.43%)
2s █████████████████████████ 205 / 1264 (16.22%) █████████████████████████ 104 / 622 (16.72%)
3s █████████████████████████ 216 / 1264 (17.09%) █████████████████████████ 130 / 622 (20.9%)
4s █████████████████████████ 213 / 1264 (16.85%) █████████████████████████ 108 / 622 (17.36%)
5s █████████████████████████ 223 / 1264 (17.64%) █████████████████████████ 93 / 622 (14.95%)
6s █████████████████████████ 209 / 1264 (16.53%) █████████████████████████ 91 / 622 (14.63%)

the second I take that thing off mine get all shitty.


I need it back.. but it wasnt worth it.

Bad Speler wrote:for some reason, my dice analyser shows that ive won 4 out of 601 1 vs 1 battles. I know that isnt the case. Is it because the dice analyser is malfunctioning, or more likely, my computer is malfunctioning?

Hmm, mine is ok, and I've been monotoring it since the latest updates. That might of appeared when the changes have been made and you used the old dice analyzer in the new page format. Keep an eye on it, I'll test it again, but I don't think I'll find any problems right now.

yeah, that might have happened. I reset the stats, cause somethings messed up.

Just finished the test again, it works well, it must of been that.

I just looked through the posted results and I see represented exactly what I see in games:

Defender rolls slightly more sixes than he should! I know the question has been asked a thousand times, but isn't it possible that because of a rounding error (or some other math anomaly) that defender throws slightly more sixes?

In particular, check out zip_disk's data: > 18% 6's for defender with 2000 rolls!

I know there must be some comp sci and probability/statistics majors out there. I myself am a comp sci / EE with a little background in probability. Would lack mind posting the code in question for peer review?

2000 isnt that many rolls to base a conclusion on. Here's a larger sample:

These are closer to actual odds, but are still off a bit, and are actually biased towards me, the attacker.

Thanks for the post - those are much closer to what is expected!

I just added the analyzer, but it will be a long time before I get anywhere near that many rolls!

mentioned it before, might as well request it again.

I think you should add the expected values next to the values you really got so you can see how much you got burned.

ok, I'll add them in the new version. If you have any ideas of things nice to implement tell me now before I start implementing changes.

Thank you
cyberdaniel

Now the sample size is starting to get interesting. There are a few major deviations in there. I'm sure they'll continue to even out as you approach 30,000.

As someone who plays with stats for a living, I could pretty much make that data say whatever I want. How about on MWF, the attacker is favored, and on TH, the defender is favored, and weekends are a free-for-all?

lol, but the goal of this is to prove that the dice are random without cheating. Anyone can cheat to say it isn't random or it is.

Attacker Dice Distribution

2s (exp) 16.67%
2s (act) 17.09%
2.5% error

4s (exp) 16.67%
4s (act) 15.85%
4.9% error

Defender Dice Distribution

1s (exp) 16.67%
1s (act) 17.74%
6.4% error

Battle Outcomes

3v2 (exp) 37.17% / 33.58% / 29.26%
3v2 (act) 38.22% / 33.52% / 28.26%
3.4% error on 2339 rolls

3v1 (exp) 65.97% / 34.03%
3v1 (act) 65.89% / 34.11%
error within margin on 2750 rolls

(Not really enough rolls to bother with the other battle outcomes.)

Conclusions

There are more than enough rolls for a statistcal analysis. Any sample size larger than 2000 should fall within a +/- 3% margin of error. Any sample size larger than 5000 should fall within a +/- 2% error.

The Battle Outcomes are almost within statistical margins. Defender wins twice on 3v2 3.4% less often than he should - no doubt due to rolling more 1s than expected.

The Dice Distributions listed are not within statistical margins, and defender's 1s are particularly egregious. I have to admit, I expected defenders 6s to be the problem, but I stand corrected.

While these numbers really seem to point to something being amiss with either the randomizer or the code, it may not be enough off to justify the time it would take to track it down, given the number of other important items on bugs/suggestion list.

how are you determining your expected values??

This link on CC (one of many, actually) http://www.conquerclub.com/forum/viewtopic.php?t=6119 references the following table from Wikipedia:



I certainly could have made an error in transferring the numbers, or in the math, so let me know if you find anything I did wrong. The 16.67% for dice distrubution is just 100% divided by 6 possible outcomes.

Just wait till i get more rolls, then we'll see how the odds shape to be. Also, cyberdaniel, it would be great if you added percentages for the battle outcomes as well. Thanks.

before i removed my thing after the update screwed it up, i had no yellow in those bars at all. i either won or lost outright