In lieu of memorising games - an experiment

[John Fairbairn]

A hardy perennial in the garden of go is whether it's worth memorising pro games. I say no. I think reading game commentaries is better However, there is a sort of halfway house which I was going to mention on the appropriate thread, but after some rumination I decided that there may be an even better approach.

My starting point was GoScorer. This has been around for about 16 years now and is still one of the most commented features on the GoGoD CD. Two people mentioned it to me in Santa Barabara. It's simply a program for letting you guess the next move, keeping score on the way. In itself that wasn't original - I copied the idea from chess - but what was perhaps original was a system of hints where the next move might be. A couple of other people have since taken the same idea and added a different version of hinting (I think there's one in SmartGo Kifu), and maybe these are popular, too. I personally rather forgot about the program. In fact, I had trouble finding it on the CD recently when someone asked a question about it, but I do remember once, at the London Open, when T Mark set up a stall where people could try to win a prize by getting a prize, some people took it so seriously that we had to hurry them up to let other people have a go. FWIW, at the end of the event, scores correlated almost perfectly to grade, which may or may not be a talking point.

So, I was going to suggest on the other thread that people might want to try GoScorer. However, I was mulling over the commentary business, and this sharpened my old feeling against commentaries, that they describe the variations/thoughts after the move, and while that is obviously valuable, it's also a bit cock-eyed. Once you've seen the actual move it affects all the subsequent comments a little too powerfully, I find. Columbus's egg and all that.

I got to wondering if there was a way to combine GoScorer and commentaries with the comments in the "right" order, i.e preceding the move. The result is the file below. You won't need GoScorer for this, but you will need some explanation how to use this file.

First, it's a game you are very unlikely to have seen. It's not even in the GoGoD database yet.

Second, I have used a fairly detailed commentary by Sugiuchi, but I have re-written it drastically, for two reasons. One is to adapt to the new format here, and the other is to coincide with comments I have been making elsewhere on L19 recently about the importance of the neglected concept of bullying. Despite the rewrite, the comments are fully informed by the pro version, of course.

To use the file, first try to guess the next move at each point. You will find a brief comment which constitutes a hint. Once you go on to view the move played, you will often find a game comment rather than (or as well as) a hint. The game comments are in [ ]. They look back on the move just played in the usual way, but hints - which are unbracketed - look forward. Occasionally, though, a game comment may also imply a future hint. You may have to force yourself to remember all this, but I'm trying to make do with a standard reader that wasn't meant to be abused in this way. Most moves are hinted at in some way, but sometimes the hints carry over to a few moves in a sequence. The hints here take their cue from the game comments and in no way resemble the hints in GoScorer, or other programs.

You will have to keep score yourself here, and the nature of the new format makes the standard sgf reader even clunkier for variations, so I have just used labels, but if you do try this, I'd be interested in reaction as to whether this format really does work. In fact, it may be useful to others if you record your scores (1 point per move for a correct guess and 0 for a failure), as well as any comments you have, in a hidden post, maybe. I'd also be interested to know whether those who do memorise games find it easier to memorise this game once they have played through it first in scoring mode.

I also found the commentary quite interesting in its own right - some nice points about dealing with moyos.

[Joaz Banbeck]

I think that this idea has great possibilities. The idea of pre-comment is rather unexplored.

I would like to see it extended a bit. Whe we are talking about a move before it is played, we often speak of goals and motivations. I'd like to see that here.
For example, the text at move 5 - talking about move 6 - merely says 'pincer'. I don't learn anything from that. ( I might as well be back to memorising ) What I would like to see is an explanation of why white might want to pincer, something that would enable me to invoke similar reasoning on the next occasion when I should be playing a pincer.

[daal]

I think that this idea has great possibilities. The idea of pre-comment is rather unexplored.

I would like to see it extended a bit. Whe we are talking about a move before it is played, we often speak of goals and motivations. I'd like to see that here.
For example, the text at move 5 - talking about move 6 - merely says 'pincer'. I don't learn anything from that. ( I might as well be back to memorising ) What I would like to see is an explanation of why white might want to pincer, something that would enable me to invoke similar reasoning on the next occasion when I should be playiong a pincer.

Quickly approaching pro Malkovitch.

[daniel_the_smith]

Thank you for posting, that was very entertaining.

[BobC]

A bit similar to goscorer is this:

http://www.weiqiok.com/asp/GameGuess.as ... yDay=1&L=1

It allows you to play out large numbers of games.

I find it attractive to use - to the extent that I tried to dig up goscorer. I have a student project this year which aims to build a gameguess/goscorer type system.

Ant plans to advance goscorer?

[lorill]

Neat idea. I got 54 moves right, only a 34% success rate.
I was bad at local fights, abandonning them too early. I also missed most of the forcing moves, and a few joseki (the first pincer for example).

[mitsun]

I like the general idea, but think it would be better to select a limited number of "predict the next move" positions during the game. These should be positions where there is one clearly best move or sequence, but where spotting that move takes some knowledge or skill for an amateur. Hints are a nice idea when several good but different strategies are possible, or when timing of a good move is difficult to predict. For what it is worth, I scored 82 points, but a large percentage were joseki or forced sequences. Maybe one point should be awarded for getting all the moves in a forced sequence? The only success I am really proud of was move 59.

[Oroth]

I think something like this has promise. I actually gave up on self-scoring because I was doing atrociously - the game felt like quite a difficult one for me. One thing I noticed though was that if I read the hint first and then looked at the board I tended to start looking in the wrong place - I was becoming more focused on answering the hint and not playing the right move.

A dedicated program might work as follows:
You are given no hints and can guess anywhere.
If you don't know where to play you can reveal a written hint
If that fails to help you can also reveal a multiple choice selection to narrow down the possibilities.

[John Fairbairn]

A dedicated program might work as follows:
You are given no hints and can guess anywhere.
If you don't know where to play you can reveal a written hint
If that fails to help you can also reveal a multiple choice selection to narrow down the possibilities.

This is exactly what GoScorer does. The format here is far from ideal, though I'm puzzled why hints of the type given here would lead you to a different area of the board. Any ideas?

[Oroth]

It might have just been a case of not taking my time or tiredness. However, without thinking through the board first I ended up interpreting clues overly-literally. For instance hint on move 31: "Is White's focus now the lower side or the left side?" started me deliberating between plays at M3 and C11 - far off track. Following intutition first would lead me to consider the merits of blocking off D6 or extending from R3. All this is probably only reflective of personal brain flaws.

[Mef]

Fun exercise -- I tried playing it in the browser by setting the SGF viewer to play mode, set up a spreadsheet with 1 columns...Click my guess, if it's right (easy to tell by the variation tree) score one in the first column, if it's wrong, choose my second move before clicking back (that way when I click back and see the variations labelled it won't have been spoiled). If the second guess is right, score one in the second column. I tried playing through the first 100 moves this way (44 on first try, 19 on second--Guess I need some practice)...Not quite as good as GoScorer since there's no hints, etc, but still fun.