I've never understood the point of tournaments played with fair handicap. Instead of testing who's the strongest player, you're just testing who's got the most inaccurate rank. (I'm fine with handicaps for social games. I don't mind too much either way.)

For smaller tournaments, I think you do need some sort of handicapping, because you'll sometimes get pairings such as 5 kyu versus 12 kyu, and an even game wouldn't be very interesting for either player. It's less of an issue when you have hundreds of participants. I think handicap = rank difference minus three (even games if you're within 3 ranks) could work well, but I've never managed to persuade an Australian tournament director to try it. I think some UK tournaments use handicap = McMahon score difference minus one, so handicaps in later rounds may depend on results of earlier games?

Many McMahon tournaments, at least in Europe, do this. The problem they are trying to solve is that in small tournaments there are usually gaps in the McMahon scores (or few players with the same score) and after a few rounds it is inevitable that players need to be paired across McMahon score and then that players with very different McMahon scores are playing each other. I think H-1 and H-2 are common, I'm not sure but maybe H-2 is more common than H-1. In case of H-2, if a player with a MM score of m needs to be paired with someone with score of m - 2 it is still an even game but but if the player needs to be paired with someone with even less MM score there will be handicap.

This kind of handicap should be a rare occurrence if the tournament has a reasonable size, except for the lowest ranked players, and I think it is most of the time only applied if one of the players has a rank lower than some preset level (15 kyu?). Assuming it is a rare occurrence it might not be that useful to use large handicap reductions, I'd expect that the number of handicap games would fall off very sharply when increasing the handicap reduction.

I have wondered if a better solution to this specific problem would be to group players in equal sized groups by rating and assign the initial McMahon score based on this ranking. Then simply ensure that each group is large enough to be able avoid pairings with wildly different McMahon score.

Last time I inspected the tournament results from US Go congress (very long time ago) it appeared to me that it was played in groups based on rank. That is it was like separate tournaments for each rank band, but maybe this was my misunderstanding or it could be that some congresses were like this? That system seems like another way to test who's rank is most modest. Maybe it is a problem that is hard to solve once you decide to seed players based on rank.

The US Open is McMahon. I think the handicap has been rank difference minus two in the past (I see a 2018 game I played (and won!) as White despite being a 4k facing a 2k), which is probably good enough, although there still ended up being some matchups at the end that were like 4k-1k, which I think deserve to be even, especially if the 4k is having such a great tournament! I've never seen McMahon difference used for handicap.

I'm still not pleased in general about the fact that a handicap of H may indicate a real rating difference of anything from H-0.999 to H+0.999 (though this could be addressed with a little work), but once you're adding a bunch of stones anyway it may not matter so much.

One slightly odd thing (it seemed to me) this year was that the distance between the bands doubled halfway through the tournament (I think). As a 4k I was playing 3ks with two fewer wins with me, rather than one. I can see the reasoning but it was slightly disheartening to see the stronger players moved further away from me. I don't know if this is common McMahon practice. (As I said, I got six good games in the end, so it worked out fine.)

One nice thing about keeping formal track of my tsumego book results is that I now have a relatively objective way of measuring the difficulty of various collections. I tend to say things like "I think that book is somewhere between volumes 2 and 3 of Graded Go Problems for Beginners" but now I have actual numbers to back that up.

I keep all this information in a Google Sheet at https://docs.google.com/spreadsheets/d/ ... NvF8TOP0c/ but I'll paste the current data here for ease of viewing (the original does have a little more information about the books). The date is there for reference in case I improve (or decline) over time. I already have two trips through 1001 Life and Death Problems in there. I'll add to the sheet as I finish more books, and I may update the copy here as well. The books I'm currently working on will probably end up in the lowest quarter of this list; I think I could use a bigger challenge than the 90+% books for now.

Yes, the Mateusz Surma books (You Won't Get Dumber While Thinking) really are that hard.

Maybe you should keep track of average time spent per problem as well? Presumably you will get more problems correct if you go slower, or fewer problems correct if you go faster.

I’ve tried. It’s too much work to run a stopwatch on every problem, and it’s annoying when I leave a problem and come back to it, and most importantly it encourages me to make decisions too fast. Hopefully accuracy is enough information to be useful.

My reading was feeling a little soggy so I took another trip through 1001 Life & Death Problems and the results were unexpected:

What happened?

Here are some components of performing well in tsumego:

• Visualization (navigating the tree of variations accurately). This has always been a big issue for me, and since I started playing again a couple of years ago I've been really concentrating on it. I wouldn't have thought this had gotten much better over the last year but I have been doing harder problems that require me to keep track of more things, so maybe my muscles have improved without me realizing it.
  
• Pattern matching. I do feel like I have gotten better at saying "This is clearly the first point to try", or "Look at all the lack of outside liberties, can I just get a golden chicken?". I have said before that the problems in this book are quite artificial, which thwarted my old pattern matching repertoire of things like L groups, but I feel like my life-and-death internal vocabulary (as opposed to just brute-force looking at every move) has expanded.
  
• Carefulness. Most of these problems are really about carefulness in the end, due to their small size; you should in principle be able to quickly falsify any incorrect try. Maybe I'm being more careful than before? I certainly know by now that the live-in-1 problems are trickier than they look. It's also possible that as my pattern-matching improves, I now see the correct move before I have the chance to miss the refutation to the incorrect move.
  
• Familiarity. This is at least my fourth time through the book. I hardly remember any individual problems (there was one that made me say "Oh, this gave me fits last time"... and I got it wrong again) but just doing these over and over has to have had some effect.
  
• Grading criteria. Am I being less harsh when I grade myself than in the past? Maybe a little... My confidence in saying "Well, I didn't look at White 4 explicitly but I looked at the idea in other lines and it was clear that it couldn't be the most testing move" may have gone up a little. This component is probably not zero but it can't account for that much.

Anyway, since I haven't been very confident in my actual play recently, this was a nice boost, showing that I have qualitatively improved at something.

Meanwhile, I keep playing in Yunguseng Dojang and BenKyo League and my results there have been good - for what it's worth, over the last year my YD rating has gone up about half a rank to a number In-seong estimates at about 1k AGA (I'm not sure I believe that) and my BKL rank, which is loosely tied to OGS, is a strong 2k. Fox games have regressed somewhat, perhaps due to mostly playing 20-second games in an attempt to get more games in, and I bounce between 3d and 2d. But I'm really trying not to worry about rank; I just want to play well and feel happy about how I'm playing.

I'm back from the 2024 US Go Congress in Portland, Oregon, and it was a lot of fun as always. I went 5-1 again (really 4-1 plus a forfeit ), so with my -3.15 AGA rating coming in I must be guaranteed to hit AGA 2k (if I don't, I give up on AGA ratings entirely). Of course I'm happy about that but then I look back at my early posts in this thread which talk about me feeling that I was basically 2k already... well, it's nice to "officially prove it" finally. And regardless of my perceived level then or my real level now, I definitely feel like I'm playing stronger Go (of course, this does not always translate into actual results).

My games were more stable than last year, which is also nice. In three of my wins I outplayed my opponent in the opening, kept on playing good moves in the middlegame, and didn't give it away in the endgame (always with some hiccups, of course). The other win was much more chaotic but it's nice to have not lost my head in the chaos and I think I'll learn a lot from going over it. In the sole loss, I outplayed my opponent in the opening, successfully dealt with the middlegame chaos he introduced, and then threw it away (my compensation after a big fight was a flower ko that I just forgot about). That's okay, I played well for a while and painful lessons are the most effective ones.

Part of my Go Congress ritual is to speedrun Graded Go Problems for Beginners during my downtime (apologies to those who are averse to soulless numbers!) and I can see that my accuracy has improved since last year, which is another nice quantitative measure of some sort of improvement.

All that aside, I swear that I still don't have rank goals! For one thing, AGA ranks are very noisy so a lot of your movement there is at the mercy of who you happen to get paired against. But more importantly, although I still would like to improve qualitatively (I like improving in general!) it is more about just being happy with my own games and feeling that I can play a full game that makes sense. That's something I think I am still capable of getting better at no matter what happens in individual games.

The other main interesting Go thing at the Congress was a series of three lectures that Michael Chen (= Zhaonian Chen = zchenmike) 1p gave on what he considered to be the principles of high-level Go. They came at some essential concepts from a different angle than I'm used to, and whether or not I'm able to apply them myself, it was really interesting to view the game in a different way from the usual. I took a lot of notes so I'll try to summarize his ideas in some other posts at some point.

The US Open and Congress in general were run smoothly and effectively, so props to the organizers for that. I particularly appreciated that there was an explicit crackdown on noise in the tournament hall after a lot of complaints during the first two rounds.

I see that he's planning to give the same lectures at the European Go Congress, so I won't spoil them here, at least for now, and will instead encourage people who can to attend them live. He's a very engaging presenter and I wasn't the only attendee who found the concepts interesting.