But if their strength is changing over the course of the year, it's a bad assumption.
This, and other points made by Kirby, reflect what I think is a valid point of view. And, by the merest fluke, there was some support from that on the very day (but afterwards) that I posted my doubts.
I discovered (to my horror) that a very famous (or, better, notorious) game by Inoue Genan Inseki against Honinbo Jowa was not in the GoGoD database. It is now! In the course of transcribing it, my eye naturally wandered to the surrounding commentary. The significance of this game was that Jowa had included it, along with some others from the same short period, in a now famous book. Genan was incensed because Jowa had not included any games of his (Genan's) from the period just after, when he had made a significant improvement. He believed therefore that Jowa was trying to belittle him.
Now if we follow the grading-list maths argument, and we could view only the Elo-type number and not the actual games, yes, it sounds plausible to say that Genan's poor grading in 1822 based on that year alone could be massaged by using his games in 1823 and 1824 and so better reflect his true strength in 1822. That is, Genan was worrying about nothing but the merest blip.
But Genan himself would say hooey to that, and if you look at the actual games you'd be inclined to agree with him. The famous game in question is famous/notorious because it shows Genan (then pretty strong at 5-dan) getting into all sorts of bad-shape tangles under 6-dan Jowa's relentlessly forensic investigation of his weaknesses. But most of all, it features an eye-popping example of a "White to live and die" situation - White can live but in the process wipes out a huge territory of his own, so loses the game. Even amateurs, reading Jowa's book, could not fail to be swayed by such an egregious example. Furthermore, Genan himself claimed he did make a big improvement
after that game (and who are we to gainsay a man who reached Meijin status not too long after.) To repeat, it would apparently have been quite wrong, in Genan's view, to massage his 1822 rating with 1824 games.
In the first case I was talking about, Hyeon Mi-chin's, even on the figures from go4go alone, I posit that her case might be similar, simply because in the relevant period she apparently did not improve much, i.e. took a rather long time to go from 1-dan to 2-dan. As it happens, and as I have already mentioned, the extra games for that period that GoGoD has, also suggest her rating should be downgraded. On top of all of that, it is well known that players do go through several slumps in a long career. Why try to hide them?
At any rate, it seems to me there are two scenarios: (1) a player's rating can be massaged by using later results, presumably in the hope of ironing out irregularities, or smoothing the curve, and (2) a snapshot based on games up to a specific time but not beyond can be presented.
I'm in mirkest glen on this. I've read books like Freakonomics and I'm aware of statistical traps it is possible to fall into unthinkingly. But my intuition (that word again!) still tells me that for historical assessments at least, the raw snapshot is more reliable (even though it can produce glitches such as red-eye) than cropping and photoshopping the image.
).
