How Does Reading Ahead Correlate to Rank
In a discussion on KGS I made a tongue-in-cheek comment that the number of moves you are able to read ahead is approximately equal to the number of stones you would give a 10k. I then added that if you are ranked below 10k your rank is related to the number of moves you can rememeber (inversely of course) in your current game.
Obviously this is not true, but it got me thinking: maybe there is a correlation. So I thought I would throw this page out here to try to engender some discussion of the matter. Anyone have any thoughts on how read-ahead ability is related to rank?
Hmm, I can easily give a 10k six stones, but I can read much more than 6 stones ahead. Then again, tesuji and reading are the strong points of my game... (On every move?)
I have heard that Yi Chang-ho has said that he typically reads about 4-5 variations 100 moves deep and then 15-20 other variations about 20 deep before deciding on an important move. This is however something I heard from a friend... it could be all wrong.
I have seen that interview translated BUT when you think about it just a little bit, you quickly understand it just cannot be really true. He may see main variation 100 moves ahead, but he cannot read. Reading means analyzing variations. Now if you take 10 moves ahead with two options at every step that would mean about 1000 terminal positions, 20 moves ahead 1 million terminal positions, 30 moves ahead 1000 million terminal positions. I guess you can guess what happens around move 100. So to "read" 100 moves ahead would really mean picking just the main line.
adamzero I don't think I ever read less than 3 or 4 moves ahead, and my max is somewhere in the 20s or 30s, probably. I'm about AGA 2d now. I remember having once read (in a book? online?) about some well known pro (name escapes but I want to say Kato or Sakata) who claimed to read at least 70 moves ahead in any major fight.
- But I was thinking about reading ahead the number of moves in question on a consistent basis on most moves.
Rakshasa: I think it's hard to say how many moves you are actually reading ahead. Much of the reading and discarding of moves is done subconsiously, you just "know" that a certain move won't work even though you haven't read it out consiously. (Even if you haven't seen that exact position before) At what point of your reading of a variation this happens is propably related to your strength in fighting. I'm a KGS 2d and in a fight i can easily read 15 moves ahead with many variations. It's mostly a matter of taking the time and effort if you have the ability to keep the moves played in your head.
adamzero Just 'knowing' isn't reading. It's pruning the decision tree via pattern matching and heuristics like 'shape'. It's only those things you don't know that you have to read. Reading unconsciously, even if there weren't a clearly different and plausible explanation, seems something of a contradiction in terms.
Bill: After brief pondering, I think perhaps how far you can see ahead may be a better indicator. In recent days I have noted that I can often see about 10 plies ahead. That does not mean that I can read that far, because I am talking about only one line of play. By see I mean a process that takes about one second or less, and does not involve conscious search.
By contrast, I think that maybe Rui Nai Wei can see around 16 plies ahead. In discussing the first environmental go game she came up with a combination that long very quickly. (She had missed it during play.)
In discussing a later game, Jiang Jujo, in an aside to me, mentioned a ko in a corner that, as I recall, Rui had found in analysis. He seemed quite pleased. I had also found it, at a depth of 21 plies.
I mention that because I may be using a slightly different notion of reading. I am assuming an accurate assessment. Also, depth is not the only factor. Breadth is very significant. If there are no branches, it is not too hard to go quite deep. The corner we analyzed had enough choices that going 21 plies deep was an accomplishment. I don't know about Rui, but it took me a long time. ;-)
kokiri I'm not going to disagree that good reading and rank are fairly well correlated but rather offer a warning that that doesn't imply cause and effect. i.e. stronger players may be better readers by and large, but that doesn't necessarily imply that by becoming a stronger reader we will all become better go players.
Reading is like the foundations of Go, upon which other strategic, shape, whole board etc skills are built. Just like with a house, trying to build a big house without decent foundations is a nonstarter but building great foundations isn't going to mean much unless the house itself is also well built.
Reading out a corner situation will show whether another move is required, but even if it is, there is the strategic decision of whether that is necessarily the biggest move on the board.
I don't think i'm by any means alone when i say that i read at all maybe 1 move in 5 or even less.
Alex Weldon: I'm a KGS 1 dan now (well, actually, my rank has drifted up to 2 dan, but I don't think it'll stay there when I play more rated games), and I barely read at all, in any conscious, deliberate way. There's obviously some high-speed reading going on at the subconscious level, because I know, when playing a move, what the position will look like in a few moves, but it's rare that I stop and concentrate on reading. Obviously, I should take the time to read things out more thorougly - it would undoubtedly make me at least a stone or two stronger, and eliminate the blunders for which I'm infamous. But if you're looking for an example to show that reading ability is not always a good measure of strength, I'm it.
Bill: But it appears that you can see a few moves ahead, Alex. :-)
Black to play.
There's no real branching involved, once the first move is played to check the move works and the variation is 9 moves or so deep.
I, a low kyu, can pretty much see what the correct move is straight away - it's one of a couple of moves that look right, but i still have to read the variation to the end to check that the move works.
there are two things going on, i'd say, one, seeing the right move, and two, following it to the end of the sequence to check it. Both are required but i'd say i value the former more than the latter.
Here, by read i really mean read out i.e. follow to the end of the sequence to check it works, rather than read in the sense reading many variations of of corner to try to find out the status. In such a situation with any real branching, i lack the concentration, tend to get bored and play on intuition based shape unless it's a game that i'm taking fairly seriously.
N.B. this example is from the excellent Tesuji and anti suji of Go, I have only used the original diagram; I trust I'm not overstepping the mark.
Bill: Well, my immediate experience was curious. Once I read 3 plies, which was pretty quick, but of the 1-2-3 category, I saw that White was screwed, because of two additional weak points. As it turns out, those are the sites of Black's penultimate plays, if White plays to the bitter end. But I did not see (visualize) the final position or later play at that time.
Rich: It's strange you say that; in this case, I read it out to the end, but recently I cut off a group with this squeeze/ladder tesuji and had a very similar non-visual reading. It actually struck me at the time that I could 'see' it without reading more than a couple of moves in. Of course, I then read it through slowly and carefully, and wasted a further five minutes reading and rereading everything ten times over, like the timid kyu-player I am :)
Dave: Personally I do not think that quantifying the ability to read ahead has too much meaning.
Reuven - I really don't understand - 6k-2d? ^^* - I mean, I can read it out withotu any effort at all - It's hardly even reading and my strongest Stable account on kgs is 9k.. ;p As for fighitn in general, I rarely put effort into them fights like some of the ppl who commented before me, I do most stuff as I feel... But when reading is nessaccery and I feel like studying a possition, my best is some of teh 3d problems on GP... But as somebody said before, ppl have different strenghts in differenty stuff... Which probably is the problem with the rank-reading ratio..
This position appears in "Reducing Territorial Frameworks". Shuko writes, "If Black were not confident of saving his group, he would have to flee to 'a' with 5, leading to White 5, Black 'b'." I can not imagine how much reading went into deciding that was safe and could survive White's cap at . However, I do not think that the key point was the number of moves ahead that Hosai read :-)
Dieter here are my thoughts
I have added to the position. I almost physically feel that White has a strong group there and Black is cut off. There is no reading involved. I know by experience that this combination of BambooJoint and table shape is almost certainly impossible to cut.
I think that this ability to feel this particular shape hasn't changed between my being 6 kyu and 2 dan. What has changed is the way I treat this position: before I would probably try something like a to cut off White regardless of reading. Now I would be more inclined to investigate the escape at b first.
When is not present, as in the original problem, I do feel that White has a weakness. I think my sensitivity for this weakness has considerably improved between 6k and 2d. I am more inclined now to start reading into this problem, but of course we are encouraged by the problem setting.
As for the real reading, I currently read up to and consider itdone. The remaining moves are not necessary. I think that's the seeing Bill talks about. I branch once at and consider it at instead, only to see that White's problem remains.
I think that as a 6k I'd probably repeat the reading a couple of times to make sure. Not anymore: reading it once suffices.
Alex Weldon: I read this without difficulty. Took me maybe 15 seconds to do it, including double-checking. What I meant before about not reading is that, in an actual game, I probably wouldn't have read it out; I would have seen the squeeze, played it instantly on the intuition that it would "probably capture, and if not, strengthen my group in sente, so it can't hurt," and then, if W was foolish enough to connect with , see that yes, in fact, captures. Maybe in a tournament game I'd take the 15 seconds to read it out, but usually I'm too lazy.
I guess I'm not really evidence that you can get to shodan without being able to read. I don't think I could know the things I know and not have picked up some reading ability along the way. Maybe what I meant is that it's possible to play at shodan level on intuition, without deliberately reading during the game.
Kokiri: What's interesting to me is that with moved; whilst the weakness still exists and black can make some use of it, the stones are not captured anymore. I'm not at a point where i can imagine seeing this difference; it's always going to involve at least a few seconds of reading unless i make an almost jedi-ike jump in my perception.
That said, I'm pretty confident that i would see the opportunity in a game. This is another skill - knowing when to read and when not to bother. Sometimes when playing someone weaker than oneself who plays slowly, one wonders what exactly they're thinking about.
I think i'm a good example of someone from a maths background, who's reasoning is therefore good, and who has pretty good reading skills. i think my strategic game is weaker, however.
B (2k): Just wondering: If you read ahead, do you actually think also in words: "black, white, black, ...." or do you just visualise without thinking in words? I do the former, maybe it would be more effective just doing the later?? (brain regions, no time spend for mentaly vocalising, ...)
Rich: I do the latter, except in very long ladders - ever since during one game, by the time I'd reached the other side of the board (with a couple of twists) I'd forgotten which colour was in the ladder.
Reuven - I always visualise the moves :S - But In monotonic stuff like simple ladders i don't bother, i just trace the chased path, without caring for the rest, till I think I should re-evaluate..
Reuven - Has anybody here tried one-color go? Or playing by memory? I did the later one, however I didn't get too far.. ^^* - As for the first one, I don't havea board.. Anyway, my point is - How do you visualise it withotu colors? Or do you imagine them anyway? I mean it sounds reasolnable as it's a rather simple division.. (Any ideas as to how badly does it "hurt" ones reading ability?)
B (2k): I am not sure that I get your question "visualise it without colors?" ... finally you will need to remember the colors, so either one has to visualise some stones as black/white. Or maybe it is simpler to remember it as sequences? I used to play blind go (9x9) regular. For me it was much easier to play blind infront of an empty board, rather than 'having the board in my head'. (My reading was very poor and i was actually happy just to remember the moves.) Another question: Should you actually do your reading on the board or in you head? (i.e. while looking somewhere else...) I think I read somewhere that pros do not correct/move off-centered stones, since it shouldn't distract your reading anyway!?
dnerra: How about How Does Reading Speed Correlate to Rank?? I think reading speed is pretty important. For the record, it took me about 2-3 seconds to read out the above example situation. So if it came up in a game, this would meant that I see immediately that I can play it if necessary, and then go on and think about something more important (e.g. whether escaping towards b might be better, keeping the capture in reserve). I would never feel bothered to double check the capturing sequence, and I would notice without any conscious effort when a white move makes the sequence impossible.
Friends of mine organize a speed-oriented problem-solving competition, and the correlation of results to the rank is,pretty good.
Calvin: I think reading ability does correlate to rank, or at least is a prerequisite for achieving a certain rank; however, like most skills one's perception of how one reads is often far off. Some people deny that they read but I think what happens is that they read so effortlessly that they don't notice that they are reading. It's more like Bill's "seeing." This is too subjective to be measurable. A better question is: does one's ability to solve tsumego correlate to rank? If you ignore the tsumego that reduce to just playing at the vital point (which presumably can be solved by pure shape recognition) and instead consider those difficult enough to require sequences of at least a few moves, then I would say the process used to solve them can be called reading. It doesn't matter if you do it with your brain or your basal ganglia or your pancreas, or if you are aware that you are doing it; it's still reading. I've yet to hear of a dan-level player complaining of an inability to solve 15k-level tsumego. The reverse -- the 15k who easily solves dan-level problems but is frustrated by his/her relatively low rank, is perhaps more common but still not typical. Usually such a player moves up quickly onces gaps in instruction are addressed. There are other factors in playing go, after all. Also, dnerra's comment about speed is important, especially in timed games. Speed reading is also correlated to rank.
Jonathan Cano : excellent point Calvin! I agree with your assessment whole heartedly. As more anecdotal evidence regarding speed, many top pros (e.g. Kobayashi Koichi) are renowned for their ability to solve extremely difficult tsumego problems on sight.
Obviously the less time you spend on answering a tsumego question the more time you have to consider higher level strategy.
ilan: My reading has improved with my rank, but that is because my general level of confusion has decreased. My reading errors mostly occur in aspects of the game that I don't understand and where I am overwhelmed by confusion, e.g., the opening and early middle game where intuition is most important. In general, the later in the game, the better my reading. I don't believe my endgame reading has improved much from 10K to 3K.
see : How to study on diagrams