Monday, October 29, 2007

How long to spend on postmortems?

I've been wondering if I should spend more time on postmortem analysis.

I like to do my postmortem analysis the same day as my game, but am usually sort of tired, so my lazy postmortem consists of:
1. Go through once making annotations about thoughts during and after the game. I don't do much thinking here, just describing what I actually thought, and any obvious errors or good moves made. (10 minutes max)
2. Check up on the opening to see where I went off book. (5 minutes)
3. Fire up Fritz and go back over the game to check it against my annotations, especially to find tactics and other missed opportunities. I also use it to check how I did in positions on which I used a lot of time. (10 minutes)
4. Post the pgn with annotations at chess.com and link to it here (5-10 minutes).

I wonder if, perhaps, instead of playing more, it would benefit me to study my games more. That would mean postponing the postmortem to the day after a slow game. One fun thing I'm thinking about doing is finding miniatures in my database for games with the same opening. That would take some time, but would be great for finding attacking ideas.

Is there any received wisdom on the best way to do postmortem, and how long to take?

22 Comments:

Blogger Dean said...

What I usually do after a playing an OTB league game in a weekday evening is just manually enter the moves into Fritz. Then I leave fritz running overnight in Full Analysis mode with say 2mins per move. This sometimes gives some nice variations that can be missed just going through move by move. Then the next day I'll do the analysis, concentrating on the parts where fritz has put a variation (e.g. where it disagress by more than a third of a pawn). I find this better than looking at every single move in detail and losing the bigger picture.

It will be interesting to hear how other people analyse their games. What I'm starting to do now is add more commentary of what I was thinking.

10/30/2007 01:58:00 AM  
Anonymous Samuraipawn said...

I always analyse my games without Fritz to start with. I try to do it as soon as possible after the game when my thoughts of the game is still fresh.

If it's a game where I did A LOT of faults, whether I've won the game or not, I try to pick the three most important faults. In some games one fault lead to another and so on, and to find every single fault can then be too much work.

Then Fritz can have a go at it. I publish it at Chess.com and if I'm lucky I'll get some comments. Then I go through it again with the comments in mind. You tend to find a lot more stuff when heavy weighters' such as Tempo, Likesforests and Loomis have a look at it.

What time to spend on your game is hard to answer. My last game lasted 15 moves and I haven't spent more than 15 minutes on that one. A Swedish master who was mentioned in a book by Jesper Hall, said that it took him at least 8 hours to go through a whole game properly! I usually spend maybe an hour max, including the post-post mortem when people have commented on it.

One fun thing I'm thinking about doing is finding miniatures in my database for games with the same opening.

That sounds like a great idea. A reoccurring tip that you read about is to use a database to find tactics, strategical concepts/manouvres, miniatures and even endgames that "come from" your type of opening/opening structure.

10/30/2007 04:35:00 AM  
Blogger Loomis said...

I think the amount of time will vary largely from player to player and from game to game. Some games you play won't require as much time as others.

I think the purpose of doing post-mortems is to identify the flaws in our game so that we know what to work on. As class players there are lots of flaws -- I could spend forever figuring out why nearly every move I made was not best. Instead, I want to figure out what are my biggest mistakes, what can I fix, how can I change how I work to play better.

Often I will simply think about a game, no board, no variations, just wondering what went wrong. This helps identify critical positions. Then I wonder about either playing differently at those positions, or whether I could have avoided them. Eventually I write down some thoughts and alternative variations -- I try to make them the kinds of things people can respond to if I'm going to post the game in hopes of helpful analysis. Of course, I do some double check with Fritz.

The whole process takes me a day to a few days depending on how obsessed I am about that particular game. This may be longer than I really need, but I enjoy it.

10/30/2007 09:54:00 AM  
Blogger Grandpatzer said...

First: if you can, do the postmortem with your opponent, esp. if they're higher rated.

Here's what I'm trying to do with my long time control games (which I need to play more of...I may have to check out long time control events on ICC). First analyze the game without fritz. Make your own notes, and record pertinent variations as best as you can visualize them. Focus on the positions where you were in the most difficulty. Try to figure out the best move for both sides at each turn. THEN feed it into fritz, and check your annotations and your own variations. For a long rated game, it should take more than a few minutes to analyze...it could take longer than the original game. I suspect the stronger a player you are, the more time you spend analyzing your own games. I'd imagine Kasparov has spent weeks studying some of his.

If, in the cold light of day, with the gift of 20/20 hindsight, and under no time pressure, your annotations are still crapped on by Fritz, and you can fathom why, these highlight weaknesses that you can correct.

Overnight Fritz analysis can be useful, but I like to go through my games using my own brain and with Fritz on infinite analysis mode. You get a clearer picture of what's going on in the position if you're feeling your way through it in centaur mode, I think. Also, sometimes if you nudge Fritz a move or two down a line it will radically change its assessment. Fritz commonly has problems with such things as "0.00 =" for a possible move repetition when the assessment is radically different if the repetition isn't forced. Another annoying thing is that it'll calculate "mate in 12" and then after you make the move decide "mate in 6" or other such weirdness.

Re: openings: Try not to just see where the opening "left book" but that you understand the moves up to that point, and understand the correctness or incorrectness of the first few non-book moves. Also, make sure you understand the main line had the play stayed in book.In your tactical gambits, this will tend to be just tactical matters. In others, there can be some work involved. If you have Marin's "Beating the Open Games", you can see the level of understanding you should strive for. For example, as Black I'm seeing a lot of four knights/guioco pianoish stuff that doesn't quite follow book but has certain recurring themes of winning the two bishops, deliberately enticing your opponent to take your "good" bishop, and timing of such moves as ...a6, ...d6 and ...h6. It's insufficient to just look up where the game left book (it's usually my opponent, with no outright refutation) but to try to understand the main lines as well. Then you can often understand why your opponent's move is "wrong". For example, if White normally plays ...c3 to avoid losing his bishop to ...Na5, and he doesn't play it, then that suggests it's right to play ...Na5 and get the bishop pair. Or if a certain move is played to prevent ...d5, and it's not played, that suggests that ...d5 is a good move.

10/30/2007 10:42:00 AM  
Blogger Blue Devil Knight said...

Thanks everyone for the comments. It seems we all agree that it is good to go solo before firing up Fritz.

It is also clear I'm not spending enough time on my slow games. I think I'm getting past the stage where just doing a quick blundercheck is sufficient: fewer of my games are being fully decided by basic tactics. I hope I can overcome the inertia and get off my ass.

GP: your approach sounds very interesting, something I will have to consider and I will surely steal some bits. I love the idea of using the off-book to understand book. I'm also thinking about using ChessDB to find miniatures from the said opening, to get a better sense for the attacking possibilities.

ICC is great for slowish games (well, 40 30...not super slow, but enough to get a couple of thinks in there. When I try to go to slower time controls it gets a bit harder to find people to play. There are a few leagues that play slower games, such as the T45/45 which I love).

10/30/2007 11:10:00 AM  
Blogger katar said...

FWIW, back in the day i played my club game 7-10pm, group-analyzed with my buddies (rated 1600-2100) for maybe 15-30 minutes (per game) at a late-night taco joint, then i went home and did a computer analysis and annotation that same night for another half hour or so, paying big attention to the jumps in the "Score Graph" feature of SCID/ChessDB.

my friend rated ~2190 could "sense" the critical juncture of the game-- the decision that dictated the nature of the resulting game. it is key to have a heightened awareness of such positions, and to use more clock to get it right. this is something that a computer could never tell you. my 2190-rated friend also grilled us 1700-rated weaklings on the reasons for our moves by asking lots of questions. (another thing a computer can't do)

finally, i believe it is useful/underrated practice to periodically review old annotations. every few months, critically evaluate the game notes and make corrections, add question marks, etc.. in this way you can improve as an annotator and continue to become more objective.

10/30/2007 12:51:00 PM  
Blogger Blue Devil Knight said...

Katar: actually revisiting old annotations, eh?! Wow. I wish I had friends here who played chess and liked to do postmortems!

10/30/2007 01:17:00 PM  
Blogger wang said...

I have been taking between 30-60 minutes on my postmortems. I don't use the computer for much other than saving it and checking a few key positions.

I typically do the postmortem the day after. I have one to do today, as I am studying for two hours today and won't play again until Wednesday.

What has helped me the most is scheduling postmortems into your study time. I have had to eliminate one game a week from my schedule because of this but I feel its the most important thing I can do to improve right now.

10/30/2007 04:34:00 PM  
Blogger wang said...

One more thing, as far as how to do it...

I look at the opening and see any critical variations, anyplace where me or my opponent veered out of book.

Then I look at the critical moments. Sometimes 2 or 3 of them in a game. I look at some sharp continuations I thought about playing and see how they would have turned out. I also pay particular attention to the "shoulder" parts of the game. When transitions were made from opening to middlegame, a change in pawn structure, or overall fell of the middlegame, and finally the transition into the endgame.

Then I look at the endgame and see if I was right to enter into that particular endgame or not, and see if my play was good, bad or just plain awful.

10/30/2007 04:43:00 PM  
Blogger Blue Devil Knight said...

Wang: slowing down for the critical moments is a good idea.

10/30/2007 07:12:00 PM  
Blogger Grandpatzer said...

BDK: "I wish I had friends here who played chess and liked to do postmortems!"

I'm hoping to make some at my new locale. If you can find like-minded people at a local club, that can help in all sorts of ways.

I noticed some decent players at our club playing through positions from an endgame book against each other, trying to find the solutions...I think it was Endgame Challenge! by Hall. If you find studying endgames as distasteful as taking your cod liver oil, that's one way to make it more fun (failing that, playing endgame positions against a computer helps as well).

One trait of good chess players is that, when they find an interesting position they gnaw at it till they think they've gotten at the truth...the more simplified the position, the greater the possibility of a concrete solution, and the more fascinated they are with it. Try to cultivate that mentality, and try to find friends that share it.

10/30/2007 11:08:00 PM  
Blogger ryan murray said...

bdk, if you're ever interested, i wouldn't mind driving out to your neck of north carolina and playing some chess. we could even do post-mortems. = ) i can't be too far away from you, since i'm in greensboro, nc.

not to take away from the theme of the comments. what've i've done in the past is take my games to a friend of mine who is a stronger player and we'll go over them together and he'll point out things i could have done better, where i could have mated the other player and so on. i also use a computer to go over games, but i've only got crafty and not fritz, so there's a little bit of program envy going on while i read these comments.

after reading these comments, i do have a lot of ideas for things i could do to improve any post mortems i do. thanks everyone!

(if you're interested in playing chess sometime, let me know. my email is ryan-murray@clearwire.net)

10/30/2007 11:24:00 PM  
Blogger happyhippo said...

One thing I do like is to use 2 different engines on a game I just played.

Typically, I would use Fritz/Rybka because they tend to balance each other more than purely using Fritz. However, when I find certain junctures where both Fritz/Rybka agree and I deviate is where I pay the most attention to.

Like others have mentioned, I normally start annotating on the very night my game ended so my mind is still fresh with the ideas/concepts about the position.

I pay particular attention to when there is a deviance of 0.5 or more because it usually highlights problems with my play which need work.

I also find it useful to annotate my opponent's comments into the game notes so I have a good idea of his thought processes from his POV.

All in all, I'll end up spending not more than 15min putting in comments into the game before the computer takes over and I let it run throughout the night and only look at the post mortem in the morning.

I find that I remember better doing it this way.

YMMV. Just my 2 cents.

10/31/2007 12:37:00 AM  
Anonymous Anonymous said...

Honestly BKK, I can't imagine you devoting any more time to chess. Your chess activities already read like an advertisement for a full time job - reading and studying, playing, analyzing your own games, lessons, maintaining your blog, etc. I would say it is a matter of prioritization. As long as you are studying something and enjoy it, you will continue to improve.

Bill

10/31/2007 08:55:00 AM  
Blogger Blue Devil Knight said...

Ryan: I think some people get together at a local coffee shop to play T and R nights. I went once and it was OK, but mostly just people playing, not much discussion of chess after. I'll try it again sometime. It is at Francesca's on Ninth Street. I think they start around 7PM.

Anon: since I finished the Circles it feels like I've been slacking!

Happy Hippo: two engine wars sounds like a really good idea for postmortem crosschecking.

10/31/2007 09:23:00 AM  
Blogger BlunderProne said...

I tend to first see where eitehr i went astray or my opponent went astray in teh opening. If it was my oppnent who went astray, I check to see if I made teh correct response. Understanding the position at that moment. I use Bookup and make entries in that to track the variations of the game.

I enter the rest of the game into Bookup adn export to a PGN for further analysis.

I try to identify a key position which was the turning point. Usually a few following the stray from the opening. I try to recognize on my own, the positional theme and goals now that the clock isn't running and even jot down any thoughts I had during the game.... THIS IS WHY ITS CRITICAL TO DO THIS ANALYSIS SHORTLY AFTER YOUR GAME. I don't always do this and regret even mroe when its a day or two after the game. insight can be gained from the thoughts.

For instance, about a year ago, I discovered I was chasing "ghosts" in my game ( false threats) and playing passively in poisitions that really required me to play mroe aggresive. Capturing the " I thought black's threat of capturing on b2 was strong" but after looking at how I still had the initiative and how "unreal" the threat was change my playing style.

But you know all that.

10/31/2007 11:43:00 AM  
Blogger Blue Devil Knight said...

BP: Now that my games aren't all blunderfests, I starting to need to develop more positional understanding, as opposed to simply understanding the tactics missed. It is simply amazing how many opening books simply dump variations without discussing pawn structure, general plans, and typical attacking lines. They often show GM games that are basically draws until mistakes on move 35 leads someone to resign because they just went a pawn down (except the 'Understanding the X' series, which is the big exception!).

I'm thinking I'll split things up into two days: first night just a more fast one where I write out what I was thinking so I don't forget (I often forget: my coach would say, what were you thinking here? and I'd be like I don't remember and then he'd see my next move and say Ohh, that was your plan and I was like Uhh, oh yeah I forgot.).

10/31/2007 01:34:00 PM  
Anonymous Anonymous said...

BDK - One more thing on analyzing your own games.

You might consider taking lessons to bump up your thoughts on your games to a master. The master can quickly identify what types of errors you are making, and help you focus your improvement.

I try to do that myself with my coach, as well as encourage my students to do the same with me. In particular, I try to get my students to think about the right things at the board. I don't care so much about the decision reached, but what elements were involved in a decision. For example, what was thought about that was irrelevant or wrong, what was not considered at all that should have been.

Bill

10/31/2007 03:29:00 PM  
Blogger Blue Devil Knight said...

Bill: my coach was very helpful. I'll go for a new one once I plateau. It's unfortunately expensive. He told me I didn't need him at the time (after we met a couple of months), that I should focus on tactics and playing, and these were things a coach wouldn't be all that helpful with.

10/31/2007 04:29:00 PM  
Blogger wormwood said...

I pretty much only look for the move where I lost the game, and figure out what (and when!) I should've done differently. maybe check out a couple of moves which I wondered if they could work. everything else I just skim through to see if fritz can point some clear, shallow tactics. I just ignore everything else it thinks. 2-5 minutes depending on the game.

blundered or won games I don't check at all.

10/31/2007 11:08:00 PM  
Blogger Blue Devil Knight said...

Worm: that's what I used to do, but I need more now. For one, checking over wins and draws is important in my arc right now, as I discussed a lot in recent posts. Perhaps for blitz that method is good though.

11/01/2007 12:42:00 AM  
Blogger transformation said...

im late to this rich thread. read all comments and all excellent. thank you BDK and thank you all. dk

11/01/2007 12:52:00 AM  

Post a Comment

<< Home