Tsumego Speed And Accuracy Table
Dieter: I picked this idea from Matthew Macfadyen who has started a table like this at https://web.archive.org/web/20130927122627/http://www.jklmn.demon.co.uk/gradprob.html. Tamsin's idea of force feeding prompted me to take it up here at Sensei's Library.
Patrick Traill We actually also have the page Problem Book Grades (a slightly more general title), which overlaps with this page (especially Get Strong at Tesuji). Should we perhaps merge these two pages? (I would be prepared to do it.)
If you have gone through (part of) the following tsumego collections, then add your name and data to the entry. Add an entry if the collection hasn't been mentioned yet.
1. Korean Problem Academy at GoBase, level 1.
Name: Lynx. Rank: 4k. Number of problems: 100. Run number: ?. Seconds per problem 1.95. Accuracy: 100
2. Korean Problem Academy at GoBase, level 2.
Name: Dieter. Rank: 2D. Number of problems: 25. Run number: 1. Seconds per problem: 8.4. Accuracy: 92%.
Name: Lynx. Rank: 4k. Number of problems: 30. Run number: 1. Seconds per problem 3.66. Accuracy: 100
Name: Lynx. Rank: 4k. Number of problems: 73. Run number: 2. Seconds per problem 3.9041. Accuracy: 100
Name: RBerenguel. Rank: 15k?. Number of problems: 50. Run number: 1. Seconds per problem 14.4. Accuracy: 82
3. Korean Problem Academy at GoBase, level 3.
Name: Dieter. Rank: 2D. Number of problems: 25. Run number: 1. Seconds per problem: 18.4. Accuracy: 84%.
Name: Dieter. Rank: 2D. Number of problems: 25. Run number: 2. Seconds per problem: 9.6. Accuracy: 92%.
Name: Lynx. Rank: 4k. Number of problems: 25. Run number: ? (have done it before). Seconds per problem 3.44. Accuracy: 96%.
Name: Lynx. Rank: 2d. Number of problems: 50. Run number: 1. Seconds per problem 5.4. Accuracy: 96%
4. Korean Problem Academy at GoBase, level 4.
Name: Dieter. Rank: 2D. Number of problems: 100. Run number: 1. Seconds per problem: 33.6. Accuracy: 89%.
Name: Dieter. Rank: 2D. Number of problems: 50. Run number: 2. Seconds per problem: 8.4. Accuracy: 96%.
Name: Lynx. Rank: 4k. Number of problems: 25. Run number: ? (have done it before). Seconds per problem 4. Accuracy: 100% Name: Lynx. Rank: 4k. Number of problems: 96. Run number: ? (have done it before). Seconds per problem 6. Accuracy: 96%
5. Beginner exercises at Sensei's Library
Name: Dieter. Rank: 2D. Data irrelevant due to replacement of some of the harder ones.
6. Maeda's graded Go problems volume 2
Name: Dieter. Rank: 2D. Number of problems: 25. Run number: 1. Seconds per problem: 74.4. Accuracy: 84%.
is this the same book as Life and Death Intermediate Level Problems ??
7. Weiqi Introductory Problem Collection (ISBN 7-5326-0661-9)
Name: Dieter. Rank: 2D. Number of problems: 50. Run number: 1. Seconds per problem: 14.4. Accuracy: 98%.
Name: Dieter. Rank: 2D (void). Number of problems: 100. Run Number: 2 (after a long break). Seconds per problem: 18.6. Acccuracy: 96%
Name: Dieter. Rank: 2D. Number of problems: 200. Run number: 1. Seconds per problem: 10.7. Accuracy: 99.5%.
Name: Dieter. Rank: 2D. Number of problems: 200. Run number: 2. Seconds per problem: 7.4. Accuracy: 99.5%.
Actually, the numbers are not entirely fair. Many of the problems in this book feature in the abovementioned Chinese work, which I am studying simultaneously.
|Name||Grade||Time Taken||No. Wrong||Time per problem|
|M. Macfadyen||6 Dan (European)||125min||2%||14s|
|S. Goss||2 Dan (European)||303min||10%||34s|
|D. Keeble||1 Dan (European)||26%|
|M. Dohme||1 Kyu (European)||155min||22%||17s|
(Note, this large collection (534 problems) is a very good candidate for the "do lots of easy problems fast" exercise)
Dieter: Time elapsed rather than used. Some collections are on the web others in books, so the time needed to browse through them may differ. Some people have slower connections than other people. Which makes speed hardly measurable, but it can give an idea. Also - and that is what Matthew used it for - the table gives an idea of how different collections compare to each other, and there one player's elapsed time can be a useful fact.
Dieter: An interesting hypothesis, after having done several collections, is that I apparently do not allow myself to drop below a certain accuracy level. While doing the Maeda vol 2 collection, I missed three problems early on and immediately slowed down my solving speed. So, the way the problems are presented has an effect (for me) on speed. Whether all solutions are together at the back or after every page, affects my solving behaviour.
Tamsin: While I think this will be a very useful page, may I sound a cautionary note? I believe one should emphasise repetitive study, not speed. I do not keep a tally of my time taken and success/failure ratio (it's enough to think about solving the problems), although normally it takes me about 30 minutes to run through a set of 150 and I have now moved up from about 80% accuracy at level 3 of the Korean problems to 100%.
What I do do is this: if I find a problem that I do not fully understand, I pause over it and convince myself why the failure lines fail and the success line succeeds. I am not satisfied to "solve" a problem merely by being able to recognise the vital point; I have to find out why it is vital. This is really my major point about repetitive study: most of us are good at solving problems and seeing tactics in a shallow way, i.e., being able to see strong moves, but we are also lacking in deep understanding of the underlying themes. The hardest part of many problems is not finding the answer itself, but in knowing how to complete the sequence, particularly against the strongest resistances. I believe that only when one has gone over an idea over and over until it is deeply understood rather than shallowly known has one truly studied it. This, incidentally, seems to be the reason why the original Korean problem books contain no answers - you have to convince yourself completely why your answer is correct.
Harpreet: Accuracy is, of course, most important but speed tells you something about how well you have learned the material. Weaker players can solve some of the same problems that strong players can solve but much more slowly. This means that often certain tesujis or ideas or not well learned and are then harder to build upon in more complex problems. If you look at the MacFadyen page about problem answering you'll see that accuracy for the very easy problems is uniformly quite high but speed varies quite a bit. I am certain it is diagnostic of skill. That's why repetition of similar themes is so important in studying problems.
Dieter: Completely agree with you Tamsin. There is more than one criticism thinkable for the approach this page seems to encourage. Still, the page gives me an incentive to study tsumego and it may do so for others.
Tamsin: An interesting thing has come up for me: my accuracy now, after nearly two weeks repetition, is total but my speed is much slower. I am finding myself questioning the solutions and trying to "break" the problems by finding alternatives, as it is not interesting merely to click on the right point for the 14th time. The more I look at the problems, the more sidelines and queries I see: what appeared to be straightforward themes and shapes can hide more complex patterns.
Dronak: A little question. Do those stats for seconds per problem and accuracy percentage get automatically calculated if you go through the entire set of problems? I did the first 10 in level 1 (25k-15k) because I'm new and did OK; I didn't get them all right on the first try though. But at that point I had to register an account. I didn't see any obvious way to get stats while I was doing the problems. So do they come at the very end automatically or did you keep track and calculate them manually?
LoP: The last time I worked through the Korean problems there was no automatic statistics generation at the end. You'll have to stop the time yourself. For offline problem collections you can use GoGrinder. It keeps track of your success rate and generates statistics.
Malweth: I'll be adding my data soon (including the Chinese version of 1000 L+D Problems which I'm pretty sure I have). Should books with sections be broken up by section (I'm thinking of the Chinese ver of "1000 L+D" from Nihon Kiin which has 3 levels included). Additionally, I've gone through the first two levels of the Baduk Academy on gobase, but not recently... will my data be skew because I've done them before - even though the last time was 9 months ago?
Dronak: Thanks for the information, LoP. I just downloaded the GoGrinder program and checked it out. I see that it does keep track of the average time and percentage correct stats. I downloaded a number of the collections they had links to so that I can practice them offline when I want to. Is there any easy way to download the Korean Problem Academy problems from GoBase though? I didn't see one, but it looks like the stats given here could have come from playing them through GoGrinder. I'd like to be able to do the same, to have a record of my stats, if possible.
LoP No it's not possible to download the whole collection from gobase.org
 RBerenguel: My first run in the last year and a half, I think, I looked at them about two years ago. And I connect to the internet by modem, my seconds per problem is lower than that (about 9 or 10), my accuracy is really that bad anyway.