Closed cslarsen closed 6 years ago
Print out avg and stddev, for all time, and last 10, 100, 1000, 10000 races. Do this in a table so that you can directly compare the performance of various setups.
Now prints avg/stddev. What I'd like to see is various avg/stddevs in clusters of text size. So short texts probably have higher avg and lower stddev, while it's worse on long texts. Or a scatterplot of the same.
Closing for now. Will need time to find out other good statistics to present.
Confidence intervals, clusters and so on. Perhaps a linear regression on the wpm by the length of the text, or for wpm/precision. Or some correlation metric between interesting pairs. It would also be nice to capture in a much smarter way which words are typically hard to type. This depends on the immediately preceding words, so capturing that information would actually help to figure out what words are hard to type given the way the fingers are currently positioned over the keyboard.