Yeah, basically, but I was using log_2(x) instead of log_10(x) in my made up example. Here’s some actual data. What they mean by “unlimited attempts” is that there is no filter on the sample solutions, so they are all submitted to codebase. I expect false positives/incredibly slow to be more likely than performance actually increasing without limit.
To check my understanding:
If C was 20, then 100,000 samples would be enough to solve approximately 100% of these problems?
But instead C is something like 6-7, because 100,000 samples gets you to 33%?
Yeah, basically, but I was using log_2(x) instead of log_10(x) in my made up example. Here’s some actual data. What they mean by “unlimited attempts” is that there is no filter on the sample solutions, so they are all submitted to codebase. I expect false positives/incredibly slow to be more likely than performance actually increasing without limit.