A few posts ago I mentioned Paul Keres' book on the 1948 World Championship, which has been praised by greats like Boris Gelfand and even Garry Kasparov. It looks very good so far, but will it hold up once we sic the computer on it?
This leads me to wonder about older books in general. It would be interesting to see who were the most accurate analysts of the pre-computer era. Does over-the-board strength correlate with analytical precision? Keres has achieved high fame as an analyst; another candidate for top honors is Isaac Boleslavsky, whose autobiographical Selected Games was singled out by Bobby Fischer for its accuracy and objectivity. I'd expect them to fare well, likewise Fischer himself for his My 60 Memorable Games.
Other interesting characters, with my guess for their accuracy in parenthesis: (pre-computer) Kasparov (high), Anatoly Karpov (relatively low), Mikhail Tal (relatively low), Alexander Alekhine (good for his day, but lower than the post-WW II champions), Mikhail Botvinnik (high).
But how are we to test this? This looks like a job for Ken Regan and his wonderful Intrinsic Performance Ratings (IPRs). Unfortunately, entering the analysis into the computer is rather time-consuming, but maybe publishers can send him the PGNs, or readers who have created them already could do so.
Note: he hasn't volunteered for any of this - this is just a fun idea I had, made all the more fun by my not having to do any of the work involved. But if Ken's interested and others - whether publishers or curious readers - are willing to supply him with the raw material, then we may be on to something.