"OO" algs

[qwr]

If you're measuring alg execution times, you should almost always just use the mean rather than cutting out best/worst. The outliers matter, because the alg isn't the full solve; a 5th-percentile-bad execution time on, say, an RUS U perm alg doesn't mean that the solve as a whole is also going to be 5th-percentile-bad and be excluded from the average.

Also maybe first quartile (the point at which 25% of solves are slower) could be a useful conservative measure of how risky an alg is. I know when tech youtubers review GPUs, they report the lowest 1% of framerates as a measure of how severe lag spikes are.

In general, the quickest way to understand a distribution is looking at a histogram. But I think the middle three quartiles (and mean, which would be close to median for symmetric distributions) is a good approximation.