According to Charles Sullivan, "The 'average grandmaster move' includes all moves I have analyzed. Of the 617446 moves analyzed, 541209 (about 88%) are moves of the world champions (plus Morphy)."

Anyway, as far as engines are concerned, I think we should respect Morphy's play. He was certainly not one of the '2100' players that Nunn has written about. And I believe that he could be much more than an average GM today -- but that's just my opinion.

Among other possible reasons, because, in a sense, this is cherry-picking. Kasparov, Smyslov, Lasker, Botvinnik and all the other players have dozens of two-years tables, one year, three years etc. It's easier to pick a very good one. Morphy does not have this privilege.

Morphy does not have this "privilege" because he quit. When we rate him, we rate him on what he actually did. There is nothing else to rate him on.

As I have pointed out earlier that "average GM move" does not refer to a random GM, but (mostly) to an average World Champion move -- and that there is no shame in a = score. (it only shows that his 'life-time' record is on the same level of the world champions':)

And that if we compare Morphy's record to the record of all world champions when they were of the same age (20-22), he is on equal foot too -- according to the engines.

alexmagnus, how can you possibly argue that retroactively calculated ratings "prove" (or suggest) anything at all about inflation (or not) in the actually fide-rated games 1970 to present?

let's start with the basic question (discussed at length in my player page a couple of times, btw) - what do <you> mean with "inflation"?

note also, that technically one doesn't "trace elo back" to some point - one has to start at some point way back and then calculate ratings going forward. but i guess you knew.

Inflation? Ratings becoming incomparble. In terms of <expected score>, which - if the inflation exists - is higher for the earlier player against a later one than his Elo rating would predict. Note, when referring to a hypothetical match between two players of different eras, I assume both have no time for preparation, i.e. everything both have is their respective knowledge at the moment the ratings are taken. I also assume that the later player doesn't know who his opponent is.

so, actual "practical chess skills" is your measure - higher rating <number> without better "practical chess skills" equals inflation. then, let's return to my question:

<alexmagnus, how can you possibly argue that retroactively calculated ratings "prove" (or suggest) anything at all about inflation (or not) in the actually fide-rated games 1970 to present?>

depending on the method used and the average level one defines, but also the selection of games one uses to produce "ratings" for games played 100 years ago, one can end up with pretty much anything, in my opinion.

how can the more or less arbitrary choices one does in such retrospective rating calculations, say anything useful about what <actually has happened> from 1970 and until now, with a given set of rating rules and rating rule modifications, etc. etc. ?