How to "fix" drop rate without actually changing anything
Increase lose streaks and decrease win streaks to normalize the results. Keep a running log of the last X battles which tracks every award (although realistically we are just talking about shards and gear) for each of those X battles.*
Check the desired drop percentage against the log and if the drop rate would be inconsistent with the RNG reward, then change the result as required, and clear the log (for that item).
If the log is good, then just record the "natural" results and move on.
Basically you want a tally that looks like 100100100110010010 but not 111111100000000000 even though both yielded the same percentage and raw number of successes.
Tally # 1 will be satisfying to a human, Tally # 2 will NOT. Nothing changed but perception (or sequence ordering). Progression will remain the same, but frustration will drop and enjoyment will rise.
This is a fairly simple solution, anyone with a CS/math degree should be able to do the math and write a proof of concept in a single day. I'm not saying that implementation would be that simple, just the base logic and math.
* You want X to be as small as possible. The smaller the X, the shorter the streaks and the more predictable/regular the results. Hopefully your drop percentage isn't an irrational number or something like 9/32, that will make for big logs and big streaks.
Check the desired drop percentage against the log and if the drop rate would be inconsistent with the RNG reward, then change the result as required, and clear the log (for that item).
If the log is good, then just record the "natural" results and move on.
Basically you want a tally that looks like 100100100110010010 but not 111111100000000000 even though both yielded the same percentage and raw number of successes.
Tally # 1 will be satisfying to a human, Tally # 2 will NOT. Nothing changed but perception (or sequence ordering). Progression will remain the same, but frustration will drop and enjoyment will rise.
This is a fairly simple solution, anyone with a CS/math degree should be able to do the math and write a proof of concept in a single day. I'm not saying that implementation would be that simple, just the base logic and math.
* You want X to be as small as possible. The smaller the X, the shorter the streaks and the more predictable/regular the results. Hopefully your drop percentage isn't an irrational number or something like 9/32, that will make for big logs and big streaks.
