I've continued reading The Rookie - the travails of a weak chess player who late in life embarks on a very frustration-ridden quest to become strong enough to hold his head up in chess circles. On the way he meets several people with similar ambitions. So far none of these people have reached their goals, and on current progress in the book it doesn't look like they ever will.
To that extent daal may be right that he will never reach shodan (but I'm still chary of accepting it's a case of "can't"). However, the book mentions a new graduate in his twenties who decided he wanted to be a grandmaster before he settled down. He worked very hard and made significant progress, reaching some sort of respectable milestone (FIDE Master?) before he gave up. He gave up because he had done enough work to realise with horror how much work would still be needed to reach his goal. He didn't want to "waste" his time any further.
I think we can all empathise with that. But the story made me realise there's another way of looking at this problem that I don't actually recall being discussed properly, even in chess. Turn it round and ask: if I do X amount of work, how much can I improve?
Obviously that's very hard to answer, but a couple of interesting points emerged from my thoughts on it. One is that you need to do quite a bit of work just to stand still, and people are apt to forget that. They wrongly equate it with lack of progress.
Second, from my experience and apparently also the experience of many on this forum, certain types of work don't seem to work for most people very well even if done intensely. Straight off I'd cite tsumego problems, learning josekis and memorising games. In these cases the amount of work needed to improve appears to be huge with little return.
On the other hand a relatively small amount of work on studying professional games seems to pay off quickly.
There are provisos to this. One is that studying pro games doesn't mean reading commentaries (that's useful but with rather low direct returns). It means studying raw games in in bulk.
I have tried this myself in the past and certainly found it was the best way forward, but I never put much effort into it. I have only two data points for cases where much effort does apply. One is the traditional Japanese advice to play over 1,000 pro games to reach shodan (in the old sense, i.e. modern 5-dan). Judging by very many anecdotes I have read, this does seem to work, though neither the proverb nor the anecdotes explain how you do the work involved.
The other one, now well known in our circles, is the case where T Mark went from 2-dan to (strong) 4-dan just by transcribing the games of Go Seigen. Coincidentally, that was just short of 1,000 games. He went on to transcribe many thousands more games yet did not make such a big or rapid improvement again (he was certainly 5-dan strength at one point but never qualified as such).
This is the more interesting case because I know quite a lot about the details.
Some of these details:
1. There were no commentaries attached.
2. He transcribed complete games (i.e. he did not follow the common practice of just playing over the fuseki, or just the fuseki and middle game).
3. This was his first big session at playing over pro games. He had previously transcribed all the games and commentaries in Invincible but (a) the number of games was not very high and (b) the commentaries and especially the variations got in the way.
4. Because it was his first major work on a single mass of uncommented games, he was very focused. Every week when he handed over the games to me to add the player/event data (we were still on floppy discs in those days), he would, with shining eyes, describe some new strategy or tactic he'd observed and we'd discuss it avidly. This seemed to reinforce what he had learnt, though it didn't have any obvious benefit for me!
5. He did this intensely over a relatively short period, and he concentrated exclusively on a single player.
6. He never had any goal of improving in mind, or even of studying. He just wanted to produce discs for sale. It was only very late in the process that he noticed his tournament results took a dramatic upward turn, but he also noticed he was able to talk about the game in a different and more interesting way.
7. He never separately studied tsumego, joseki or counting (though he had done some of this in his kyu days).
Assuming all these factors did in fact explain his improvement, we have to ask why his improvement stalled after that, apart obviously from the law of diminishing returns.
I'd cite as the most important factors that he became "too" good at transcribing and there were too many humdrum games. When he was working on Go Seigen he was still a novice at transcribing, and was relatively slow. He therefore had to spend much longer on each game. The time was spent on hunting for the next move on densely filled diagrams. His brain was working subconsciously somewhat like AlphaGo doing pattern recognition. By the time he had finished the GSG collection he had got the average time for a transcription down to under 30 minutes whereas for me it was still about 45 minutes. Eventually he trimmed his time down to about 20 minutes, but even after about just 800 games he had become adept at predicting accurately where top pros played their next moves, so that he could find them quickly on the diagram. But once he had become adept, his brain did not have to work so hard. Eventually his transcription speed also slowed significantly.
Another factor with the GSG games, which actually made transcription harder, was that Go in particular made many unpredictable moves. I highlight this because I believe one of the most significant learning experiences for all of us are the "I didn't know you could do that" moments. In fact I think that is ultimately why studying many pro games pays off so much: in every facet of the game we see a host of moves that we would never have thought of for ourselves, which not only adds a new weapon to the armoury but also provides interest and thus motivation. This is why the "humdrum" games I mentioned add a negative element to the equation.
The skill set that Mark acquired was such that he became perhaps the best lightning player in Britain and one of the best in Europe. I think that could be expected from the neural network training kind of work he was doing. What was surprising, though, was that he also became very, very good at tsumego and tesujis (but mainly of the type where you have find a surprising move - presumably he'd seen them all before!), and that he was rather good at instantly estimating the count.
Another factor that applied after he completed the GSG collection is that he never again worked as intensely on a collection of a single player. He did many collections, of course, some much bigger: Shusai, Kitani, Takemiya, Fujisawa, Sakata, Hashimoto, Yi Ch'ang-ho, Cho Hun-hyeon, etc etc. But at that stage he was dipping in and out, doing yearbook games of many different players in-between. My speculation is that this blurred the focus: no longer could he build up such an accurate network of how an individual player played and how his moves could be predicted. Having seen so many styles he was predicting too many possible moves during the transcription process. These moves were clearly not bad moves but they slowed him down and introduced inconsistency into his thought processes.
What tentatively I draw from all of this is that the time needed to play over several hundred games of one player at a smooth and moderate speed, but attempting to predict the next move, gives one measure of how much work is needed to make significant progress. But it must be focused and fairly intense. With a relatively small number of games no measurable progress can be expected (an unformed neural network is presumably almost as useless as no network).
But once this hurdle is crossed and improvement is made, further progress may well require (in addition not instead of? a different kind of study. The neural network has been formed and can now only be tweaked.
As to how long it would take you to show a very significant improvement if you followed T Mark's method, he generally aimed at a minimum of transcribing four games a day (7 days a week) and sometimes did as many as 10, but he didn't quite manage that in the early days - apart from being slower at transcription he was still working for a living. From what I recall it took him a little under six months. So I take that as the baseline. Anything shorter or less intense than that means, if you are like me for instance, you will just stand still.
|