Life In 19x19

Hi,

Dan Heisman, who is a good chess teacher for beginners/intermediates, suggests that if you want to improve you should be playing chess games where you have at least 30-45 minutes of thinking time, that blitz games don't really help you improve.

However, in most of the Go improvment internet posts etc that I have read, everyone is advocating playing a lot of faster games and also doing a lot of problems that never take longer than just a couple of minutes maximum. Even the "Go of Ten" insei guy was playing some large number of 30-minute sudden-death games in Class E.

A game I posted for review sort of illustrated this problem to me, as many mistakes that I'm making in the game are because I'm not taking the time to read everything out to my horizon for being able to read things out because of lack of time, so I end up making an instinctual call about whether this is a reading situation or not. This has led to, for example, not seeing a rather obvious point about a corner pattern that appears in at least 50% of my games, because I just play those moves my rote, or letting a group of stones die because I don't slow down to ensure that they are safe.

I make these kinds of errors in fast chess and the answer that I would get is obvious--play longer time controls. But the prevailing wisdom in Go seems to be different, and I'm not sure why.

I think the idea behind it is not reinventing the wheel everytime (e.g. reading everything out) but recognizing patterns. And for you to recognize patterns you have to seem them a lot and often. You will never have enough time to read everything out in a serious game.

Ortho wrote:...in most of the Go improvment internet posts etc that I have read, everyone is advocating playing a lot of faster games and also doing a lot of problems that never take longer than just a couple of minutes maximum. ...

Really? The advice seems pretty mixed to me, and for the simple reason that both are important. Playing lots of fast games and doing easy tsumego helps raise your awareness for what kind problems can arise, and what situations you tend to have difficulty with. Playing slow games and puzzling over harder problems gives you the chance to question your instincts and learn to find better moves. If however you only go down one route, you're bound to miss out on the advantages of the other.

I would say that playing many fast games is good until you are at least single digit, then you need to focus more on reading and understanding what you intuitively learned during your double digit fast learning.

I played a few thousand igowin 9x9 games while my computer was compiling and that got me to around 12k so.. Heh. I just wish it was equally easy to go from 1k to 5d.

Ortho wrote:However, in most of the Go improvment internet posts etc that I have read, everyone is advocating playing a lot of faster games

Not everyone. I prefer slow games. At the US Open this year, the time control
was 90 minutes initial time, plus byoyomi -- I was very happy.

But I also agree with what coderboy said: for beginners, fast games are OK.
But later, you can slow down.

For me the important thing is to review the game afterwards, wether you play blitz or with some slower time-settings (personally I prefer at least 15 minutes maintime and 5*30sec Byo-Yomi).
Of course you will have to take a more general approach when reviewing blitz games because they tend to have a lot of misguided moves in them.

After the game, just go over it again and ask yourself e.g., was this the biggest point on the board (respectively, was it ugent?)? Did I chose the right direction here (to attack, approach, expend...)? Where did I lost the game? Was it because of tactics or because of my strategy (or lack of the latter)?

No one knows the answer.

I also think both kind of games are important.

The fast ones for pattern recognition and routine.
The slower ones for reading practice and/or a possibility to try to come up with new ideas instead of repeating the same mistakes over and over again.

I also tend to feel way more involved in long games - so they increase my motivation.

Fast games to me are mostly just practice, while playing long games (even without any time limit at all) is "for real".

Interestingly, I was leafing through a chess book just yesterday, which recommended that beginners play lots of blitz games so that they become comfortable playing under time pressure.

Anyway, there are a few different considerations.

(1) Using time effectively. Playing 8s/move is excellent use of time in a blitz game, and awful use of time in a slow game. If you want to play quickly, that's fine, but use the right time settings; if you play at blitz speeds with slow time settings that just shows you're afraid of the clock.

(2) Playing just for fun versus for review. If you enjoy playing blitz games, there's nothing wrong with them, but then if you ask someone to review the game with you, you'll be spending vastly more time on the review than was originally spent on the game itself. And you'll get that uncomfortable feeling where the reviewer can point out better sequences on every single move, and you think to yourself, or say out loud if you're feeling defensive, Oh yeah, I could have seen that, I just didn't have time to read because it was blitz...

(3) Applying what you know. If I've been studying something, I want to get get it right in the next game I play. It's frustrating to, say, study L&D positions and then bungle them in a game because I was breezing through them faster than I would if I saw them in a book.

It's important to review your games. I don't think that the speed at which you play is very important. If you pick out the mistakes in your play, well, this is when you will improve.

One big difference between chess and go you might want to keep in mind is the importance of pruning in your reading. Let's assume a good goal for a beginner is to consider their move, and all ways an opponent might respond with their next move (a 2 ply search if you will).

In chess, you will have about 30-50 possible moves in a middle-game position and your opponent will have 30-50 responses. If you can prune off even just 2/3 of all possible plays as not helpful or unreasonable, you end up with a search space of 100-250 move/response pairs you have to evaluate...a large, but manageable number if you have decent time controls (many combinations may take less than a second to evaluate, and any sequence with a "check" will dramatically reduce the number of legal plays available). In go you realistically must prune 95% or more of the moves at any point just to get yourself into that reasonable number (At move 60, you have ~300 possible moves), and your opponent ~300 possible responses). If you wish to evaluate deeper into a sequence, obviously you must do even more pruning, and what's more it must be done at every stage.

Many people emphasize the importance of building pattern recognition because without it even figuring out where to begin reading a sequence may be challenging (and potentially a poor way to spend your time). It is extremely helpful to be able to immediately recognize a handful of candidate moves and responses from shape memory, etc to use as a starting point. Once you have a feel for selecting moves, then spending more time evaluating them (and even finding when your instincts are wrong!) will be time better spent.

I'm a chess expert (USCF ~2150) but only 2k Go player (I've been playing just for a year and a bit) so take this with an appropriately-sized grain of salt.

It totally baffles me that someone can claim that fast games are better for improving than slow games. When you play a slow game and review it, you can literally go back a week later and point at the things you learned: In this common position, X is a strong tesuji. I undervalued the potential when he had a framework that looked like this; he can get lots of points if he makes these two moves. I played move A to enclose the corner, but when he has stones like this, move B is more secure, since it forestalls the potential endgame move C. There is zero magic, just lots of things to learn. If you learn them, then, by definition, you improve. In contrast, if you play a game quickly and don't review it, you don't pick up on those things. So how can you improve?

I appreciate that fast games might make you better at time management or improve your concentration. But they can't teach you those things about the game, unless you go back and spend as much time looking for them afterward as you would have spent playing a slow game in the first place. So where else would you learn them?

The "lose 100 games" advice is unbelievably weird. I can't imagine how this advice can really be how someone learned Go; more likely you play 100 games and you're still 20k. When you take a college course, do you try to learn the material by quickly skimming 10 textbooks but not doing any exercises?

cata wrote:I'm a chess expert (USCF ~2150) but only 2k Go player (I've been playing just for a year and a bit) so take this with an appropriately-sized grain of salt.

It totally baffles me that someone can claim that fast games are better for improving than slow games. When you play a slow game and review it, you can literally go back a week later and point at the things you learned: In this common position, X is a strong tesuji. I undervalued the potential when he had a framework that looked like this; he can get lots of points if he makes these two moves. I played move A to enclose the corner, but when he has stones like this, move B is more secure, since it forestalls the potential endgame move C. There is zero magic, just lots of things to learn. If you learn them, then, by definition, you improve. In contrast, if you play a game quickly and don't review it, you don't pick up on those things. So how can you improve?

I appreciate that fast games might make you better at time management or improve your concentration. But they can't teach you those things about the game, unless you go back and spend as much time looking for them afterward as you would have spent playing a slow game in the first place. So where else would you learn them?

This is an unjustified assertion.

Even if it is completely true, other possibilities are:
- even if you learn less per game, you can play enough extra games that you ultimately learn more
- fast games teach different things, such as improving the ability to recognise shapes automatically

The "lose 100 games" advice is unbelievably weird. I can't imagine how this advice can really be how someone learned Go; more likely you play 100 games and you're still 20k. When you take a college course, do you try to learn the material by quickly skimming 10 textbooks but not doing any exercises?

I assume its clear that that proverb is semi-humorous. The idea, which I strongly believe is correct, is that for a beginner it is more valuable to play games fairly fast than to think extremely hard about every move and still get them wrong. Playing faster games, especially against an experienced teacher, quickly gives the student an idea of what sorts of shapes are normal, what moves are tricky, what moves they can't ignore, and so on. I've taught other people who think very hard about every move, but often this is just wasted effort that doesn't actually teach much compared to the rapid and varied knowledge that a series of teaching games can provide.

Obviously this is both my opinion, and dependent on circumstances. I do think the viewpoint that slow games are the only way to learn is extremely wrong in all but the loosest sense. However, it is likely less wrong as the player improves.

cata wrote:you can literally go back a week later and point at the things you learned: In this common position, X is a strong tesuji. I undervalued the potential when he had a framework that looked like this; he can get lots of points if he makes these two moves. I played move A to enclose the corner, but when he has stones like this, move B is more secure, since it forestalls the potential endgame move C. There is zero magic, just lots of things to learn. If you learn them, then, by definition, you improve.

If this was an accurate picture, then reading books, the L19 forums, or techniques on sensei's library would be as good as playing.

amnal wrote:Other possibilities are:
- even if you learn less per game, you can play enough extra games that you ultimately learn more
- fast games teach different things, such as improving the ability to recognise shapes automatically

OK. What is the mechanism? How can a fast game teach that? Here is how I improve my ability to recognize shapes automatically:

Step 1: I don't know a shape.

Step 2: I play a game and I think hard about a move and I pick the wrong shape, because I have no experience with the right shape, and I can't see deeply enough to know that it's right.

Step 3: When analyzing the game, I identify the fact that I wasn't satisfied with my move, or I find that it led to a bad result. I either dig in deeply and discover that the right shape would have worked better, or a stronger player points it out to me.

Step 4: I think about it for a few minutes until I think I understand why the right shape is right. I try to think of other times when this shape would be useful.

Step 5: Because of all this thinking and the context of my game, I stand a very good chance of noticing this shape the next time I have an opportunity to play it. Occasionally I remind myself of what happened in my game to reinforce my memory of it.

Step 6: Now I know the shape.

I think we can agree that if you follow those steps, you will wind up seeing lots of good shapes! There's no mystery step where a miracle happens.

If I were playing a fast game, step 2 doesn't work. I make mistakes because of time pressure, so it is more difficult to identify what errors were the result of time, and which errors were because I was genuinely confused. Steps 3 and 4 only work if you spend a lot of time in the review. If your future games are fast games, step 5 doesn't work very well, since you often won't spend enough time on your moves to say "Oh, two weeks ago my game looked like this. I remember the right idea was X. Why was it the right idea again? Does it apply here?" If you notice it at all, you'll just have to guess about that stuff.

So this process seems like it doesn't work well if you play fast games and it cannot work if you do not review.

What is your alternative process? I understand that sometimes you can just sort of get better without consciously trying to learn, which is fine as far as it goes, but that's not exactly a recipe you give to an adult who wants to spend time getting better efficiently.

hyperpape wrote:If this was an accurate picture, then reading books, the L19 forums, or techniques on sensei's library would be as good as playing.

Well, I think that for a lot of people, if they study books and spend as much time as they would spend on a slow game working through the book, it is as good as playing. That's at least true for chess players I've known. If you just skim a book, it probably isn't very useful.

Personally, I find games more useful specifically because my recall is better when I see something in the context of my game than in the context of a book. If I can remember the things I learn from a book, they're equally useful as the things I learn in a game, but I'm less likely to remember them.