That’s fair enough. I haven’t quite got my head around what the principal eigenvector actually tells you here, but tiers you’ve used it to create like mostly reasonable. Forming tiers from Nash isn’t an obvious process, anyway, with most of the cast getting zero weight.
You might be interested in the work vengefulpickle’s been doing some trying to separate the character strength and player skill effects in the historical data here. I don’t know what his matchup chart looks like at the moment, but it could be interesting to compare.