Numbers might be off on account of PRD.
Numbers might be off on account of PRD.
I guess not. On a side note, I was able to crit 15 times in a row with 5 crystalis just now. 0.25% chance. Maybe I should shoot for a high score?
If there was say a 85% crit chance cap and it was 15 times in a row, that would be more believable (if the 20% had stacked).
edit: a few shots after I was able to get 12 in a row, followed by a 10 in a row. It just seems to be so consistent. I'm just using 5 crystalis as well.
I think I'll try using 4 daedalus' and compare the results.
Was able to crit 18 times in a row. Anyways, this is all too inconclusive. I have no evidence that can claim it stacks or doesn't stack. My guess is it is as HaiHai said.
Last edited by mibuwolf; 07-02-2013 at 12:20 PM.
That's a good point. I wouldn't be suprised if one item critting means all the other items consider themselves to have "not crit", and increase the likelyhood that they'll crit as a result. This would mean that with 6 Crystalys on every attack there would be least one Crystalys that thinks it hasn't crit in the last 5 hits, and has an enormous chance to hit as a result. (and even if that large chance doesn't occur, you still have 4 more crystalys with a non-trivial chance to crit to back it up)
i just did a rough test with 6 crytals. 11 non-crit out of 56 hits. so 80.36% crit rate. what's the theoretical value? ~73%?
oh ya they stack linearly... ur math teacher is crying.
I'm neither a statistician nor am I terribly familiar with the specifics of PRD, and my understanding of maths is quite a bit worse than what it used to be. But, from my understanding of the various systems, I believe that even with PRD, the long-term crit rate should still be 1-(0.8^(number of Crystalys)).
Tests wouldn't be very rigorous without at least a few thousands of results, I believe.
You have been calculating about how many consecutive crits there are and what the chance is on it happening.
This means you could have counted the crits and everytime it was fewer than 5 you would have counted again just to see how many consecutives crit there can be. And after that you calculated the chance on it happening. The chance on having 10 crits in a row with a test sample of 100 attacks and 10 attacks are significantly different.
But I am not sure if you did it. His statement however holds "more value". (What made me say this is, I also tested it a moment ago. There were 2 crit attacks and then 4 crit attacks when I started my own experiment. (I counted from the very beginning.) Totally different than your results.)
The correct approach of determining the chance of critting which works for both PRD and PRNG is to count all attacks and crits.
The chance on critting should be at around 74% (= number of crit attacks / total number of attacks)
To his last sentence: If they are stacking linearly it means there is a 100% chance to crit with 5 crystalys.
James "2GD" Harding's career as a Dota 2 host on Valve events
* 25.02.16 - † 26.02.16
Spoiler: