"I think chessmetrics is surprisingly good"
After considering and compare Chessmetrics' highest rating scores for these four players:
Adolf Anderssen: 2744
Paul Morphy: 2743
Lionel Kieseritzky: 2734
Howard Staunton: 2706
My evaluation would have to be that Chessmetrics can be surprisingly bad.
Really? Is your point that Morphy is rated low relative to those other very strong chess-players...I'd have to agree... but Morphy is a truly special case. Statistical evaluations can't be better than the records they're drawn from and we just don't have enough games from Morphy played against top rated players to evaluate properly. Odds games and games against amateurs and second-raters make up a large amount of his surviving games.
Our judgement/estimation of Morphy's strength comes from his obvious sheer out-classing of opponents. You and I can see that Morphy spanks Anderssen, Europe's best or second best player, like a school boy -- we can see that Morphy's chess seems at least a generation ahead of Anderssen and other European masters. But Chessmetrics doesn't evalutate quality it just relies on cross referenced scores -- indeed this mathematical objectivity is in the vast majority of cases chessmetrics' strength. But Morphy's is a special case, because lacking enough good data, our subjective judgement of quality has to take the place of number-crunched results. So, Chessmetrics method doesn't do Morphy justice. I think that's to be expected. Europe's best didn't do right by Morphy, ducking him. (And Morphy didn't do right by Morphy, giving up chess.) And that leads to chessmetrics rating Morphy below his true relative strength (not that +2700 is all _that_ low!) So, I don't disagree that chessmetrics gets Morphy wrong... I just don't think it's surprising, or indicative of a larger general unreliability of chessmetrics' method. Like all evaluative methods chessmetrics has it's limitations and blind spots... you've identified one with Paul Morphy.
Or do you have other problems with those numbers?
Kasparov got nightmares until
now! And remembered it very well!