I did an updated version of the 2018 cubetutor power rankings.
This time, I used data from both cubetutor and cubecobra.
The four metrics considered are:
1) Cubecobra ELO (strength of the card based on picks), 30% weight
2) Cubecobra # of cubes, 20% weight
3) Cubetutor pick %, 30% weight
4) Cubetutor # of cubes, 20% weight
Because these 4 metrics are all strictly positive, they are closer to lognormally distributed than normally distributed.
I took the log of each value, which greatly reduced the skewness and kurtosis, making the distributions much closer to normal.
Then for each metric, I Z-Scored it by subtracting the mean and diving by standard deviation so that each metric has 0 mean and 1 standard deviation.
The ranking of each card is based on the weighted combination of these 4 Z-Scored metrics.
For the # of cubes metrics I also made a small adjustment to make sure the metric had no correlation with the number of printings of the card (before the adjustment there was a 35% correlation for cubecobra # of cubes and 45% correlation for cubetutor # of cubes).
The cards evaluated are current through Ikoria.
My goal is to continue to update this ranking once per year.
We still have about 5600 votes from 2018 on over 2000 different cards from mtgsalvation users.
Even though we're missing some newcomers, I incorporated those votes into the numerical rating.
The way I did this was to again z-score the voting results to zero mean and 1 standard deviation.
Then for all cards that received a vote, I added on the average numerical rating (0.17) to the voting
z-scores so that on average the voting will not change the ratings and the combined rating will still be zero mean.
The voting result was given a weight of sqrt(#votes/(#votes+25)), so that it increases with the number of user votes.
The community already did a lot of voting on the 2019 power rankings (which is a separate project based solely on user top 20 voting rather than cubetutor and cubecobra data), but if there's still interest in voting I can make available a spreadsheet for that task so that users can add to the existing votes (either casting votes on newcomers or new users casting votes). I went ahead and add my votes for the new cards.
I currently have the results sorted by the "net" rating that includes both the cubetutor and cubecobra numerical rating plus the voting data.