[Imbuing] Success Chances are Wrong

[Old Man of UO]

Has anyone else noticed that the number of failures is excessive high? Most of my imbuing I do near 50% success rate, mean on average there should be about 50% success and 50% failures. I am getting about 1.5X the number of failures compared to success... way above any statistical coincidence or even able to just blame the RNG.

I have many runs of 5-7 failures in a row (statistically a 1 in 32 to 128 chance). Last night I had 12 failures in a row at 50.8% chance of success - that is a 1 in 3385 chance.

So, even if you don't understand my numbers, the failures are way to high to be statistically possible. EA has a calculation error somewhere, either in the display or in the actual function.

 

[Barry Gibb]

That seems to be consistent with any of the other skills I have trained (regardless if they are Crafting, Casting, or Fighting skills).

The RNG is streaky at times. Are you also recording the number of times you succeeded 5-7 times in a row (which would carry the same probability of occurrence)? I don't mean to sound like a jerk, but you have to be fair with the math to show evidence of a problem.

My guess is that the same RNG is used for all skills (don't know for sure, because I did not program it, but why re-invent the wheel). People have always complained about its streakiness and there have been some adjustments in the past based on solid testing and reporting. But people still complain that it is too streaky. It usually boils down to a matter of perception, in that we typically remember negative effects more than the positive ones (in this case, the number of failures versus sucesses).

I suggest you submit all of your data to EA, along with your test methods, so they can investigate if there is a problem. Maybe try the Bug Reporting avenue, but this I am unsure of.

Stayin Alive,

BG

 

[Drawde2]

I understand how statistics works, so I wasn't saying anything. But I THINK you get more failures than you should. I haven't paid as much attention as I should to my success streaks, but they don't seem to happen as often, or as long, as the failure streaks. At around 60% success rate on each attempt.

 

[pgcd]

I agree with Old Man of UO, and in my case the numbers are even more blatant since I strive to have more than 60% of success. 5 fails in row are rather frequent (so frequent, actually, that I don't notice them anymore); yesterday I had at least three 6-runs and one 7-run during a couple of hours gameplay, along with the first 10 successes in a row (and I'm talking around 40 full points of skill that I trained).
Of course, streaks can happen - the point is that I am using a lot more resources than I should with my chances, and the simplest solution is that the calculation used for display are different from those used to actually see if you succeed.

 

[Old Man of UO]

...

The RNG is streaky at times. Are you also recording the number of times you succeeded 5-7 times in a row (which would carry the same probability of occurrence)? I don't mean to sound like a jerk, but you have to be fair with the math to show evidence of a problem.

...

I suggest you submit all of your data to EA, along with your test methods, so they can investigate if there is a problem. Maybe try the Bug Reporting avenue, but this I am unsure of.

...

I didn't record all of my data, just here and there and enough to know that it's statistically significant. The distribution is a binomial, not a normal distribution, and takes a lot of data. But when I get two or three streaks of 7 failures in a row while looking for 10 successes, or get 12 failures in a row, then that is significant. And I see this quite often. As far as being fair with the math, I ran these kind of tests for 25 years as part of my job... man can that be boring counting tablets in 1,000 bottles to see what the average and spread is!

The one problem with my test that skews the results is that I quit after 10 successes. A more fair test would continue past that. The best way would be to have one of the Devs run a test with about 1,000 repeats, something that's not easy (or at least not cheap) to do without unlimited resources.

I just don't believe it's the RNG being streaky, because my streak has lasted 3 weeks now.

 

[RainQueen]

I'm not sure if this is my imagination or not but it seems to work for me. If you get more than 2 failures in a row ........ stop and unravel something. Somehow this seems to reset the counter and most times you'll get a success on your next imbuing attempt.

 

[Storm]

old man is absolutely correct the fail rate is way off! we need a better rng for sure they have even stated the rng is not working exactly right!