That's an interesting point which I'd wondered about after posting. At what point should it even out? eg Is there a value (of attempts) which once you have reached it your results should be near 50%. Do you have to have 1000 attempts, 10,000 or just keep going. Obviously the more you try the nearer 50% you should reach, but practically, if the value is infinity or 10,000 (for example) I think that for the average player (1000 attempts in 2 months is pretty impressive btw) if that value can never be reached then our results will be either quite a bit higher or lower than a 50% average. Hope that makes sense. Maybe there are players who consistently have a better than 50% success due to the number of attempts made.
I'd be more than happy to write a program that will generate a distribution chart to show likely outputs and let you see when it will even out
For the record, 1000 in 2 months isn't much, I've done over 100 just today lol
- - - Updated - - -
Not to sound grumpy but why is this post being continued? It's been stated that it's a 50% chance and as the evidence given by Zade we can roughly confirm it's 50% yes some days you might shatter 12 out of 15 orb attempts and on other days you might succeed on 15 out of 16 attempts.
Cos arguing with Internet strangers is better than arguing with my neighbours
- - - Updated - - -
Sigh, me being me, obviously couldn't help myself, any opportunity to play with graphs!
I wrote a simple little program that uses the same algorithm as Kevin to decide whether or not a gem becomes an orb, and ran it a total of 11,111,000 times (to generate the graph, I actually ran it a lot more times than that after generating the graph, just to confirm some things)
First, the blue squares show the distribution of success rates when you try to upgrade 10 gems, sometimes it is as low as 20% (2 orbs out of 10 gems) and sometimes as high as 90%! (9 orbs out of 10 gems). If I ran the program again and again it would certainly be possible that you could get 10 out of 10 or 0 out of 10 (chance of either of those things happening is 1 in 1024, so if you do 1024 gems you'll probably find on one occasion you'll get 10 success in a row and on one occasion you'll get 10 fails in a row). I actually ran the program in a loop a further 100 times and saw exactly the pattern that the maths would predict.
Now doing 100 gems at a time as in the orange diamonds, you see the distribution is much closer to the middle, it rarely deviates more than 10%, so from 100 gems you will almost always get between 40 and 60 orbs
Doing 1000 gems the deviation is more like 3%, so you're almost certainly going to get between 470 and 530 orbs out of 1000 gems.
At higher orders it's almost bang on 50% all the way