Perfection is not attainable, but if we chase perfection we can attain excellence – Vince Lombardi
Ever since Adolf Anderssen sacrificed a B, both R's and then his Q to mate Kieseritsky in the Immortal Game in 1851, analysts have been plumbing its depths trying to find flaws. Reti determined Kieseritsky missed a draw and GM Robert Huebner found a whole mess of mistakes on both sides. Then there was the Bird vs. Mason game in New York, 1876 which won a brilliancy prize...it turned out that Bird's 29.a5 was NOT a brilliancy because Mason missed 29...Bxd4!!
In his book Chess Lists 2nd edition, Andy Soltis lists eight more flawed brilliancies Pollack-Lee (London 1888), Mieses-Bardleben (Barmen 1905), Pillsbury-Wolf (Monte Carlo 1903), Leonhardt-Tartakower (Carlsbad, 1907), Schlechter-Salwe (St. Peterburg, 1909), Averbakh-Kotov (Zurich, 1953), Piket-Nunn (Wijk aan Zee, 1990) and Lautier-Shirov (Horgen, 1994).
Writing in his recent Chess Life column Soltis pointed out that a growing problem for annotators and authors is that many of the best known instructional games are flawed and that makes writing anything instructional difficult. He said that when he wrote The Inner Game of Chess back in 1994 he had to rely on human analysis because engines weren't very strong. But, when he revised it last year, modern engines found a lot of errors.
Soltis added that even if one one does find a relatively faultlessly played game, it usually has little instructional value. The reason is that engines often find refutations that involve strange, hard to understand and explain moves. This is why I find my Fritz program's Shootout so valuable. When the engine plays strange moves or gives an evaluation that “doesn't look right” I run Shootouts and watch the “flow” of the game. It's this general drift, not so much specific moves, that sometimes reveals where one side's advantage lies.
True, the most interesting games are often flawed by bad play by either one or both sides. Strong players can now use engines to check their judgment and spot hidden tactical errors and top level GM's can use computers to test their openings and generate ideas. When engines suggest surprising moves that look ugly and seem to violate principles to humans, their evaluations are usually right though. The result is that the games of today's generation of elite players sometimes seem to resemble computer play which is boring to the point of making you want to scream.
The moves Fischer and Tahl played may have been inferior to those Kasparov, Anand or Carlsen would have played but that doesn't diminish their achievements. I disagree that the moves played in those flawed brilliancies may not be considered instructional and they are a lot more enjoyable to play over.
There's some instructional value even in imperfect games. Why? Joseph B. Wirthlin said it best: Great sculptors and artists spend countless hours perfecting their talents. They don't pick up a chisel or a brush and palette, expecting immediate perfection. They understand that they will make many errors as they learn, but they start with the basics, the key fundamentals first.
I think what makes people fascinating is conflict, it's drama, it's the human condition. Nobody wants to watch perfection. - Nicholas Cage
I don't love Photoshop; I like imperfection. It doesn't mean ugly. I love a girl with a gap between her teeth, versus perfect white veneers. Perfection is just... boring. Perfect is what's natural or real; that is beauty. - Marc Jacobs