x
Chess - Play & Learn

Chess.com

FREE - in Win Phone Store

# Silman's Math?

• #1

In it he talks about how a way to become good at chess you have to absorb patterns by going over lots of master games. He says that in order become an IM or GM, you need to look at 100,000 games. If the average time per move is 10 seconds (including setting up the board, etc) and the average moves per game is 40, then that's over 22,000 hours of just going over master games. If you did this 4 hours per day, ever single day of the year, then it would take you over 15 years to reach 100,000 games. Can any IMs or GMs confirm that this sounds reasonable? or was Silman perhaps exagerating?

I'm curious because his article was the first time I'd ever read about becoming good by playing quickly through lots of master games just to absorb the patterns. At first it sounded too easy to be true but then when I crunched the numbers it seemed crazy. I'm trying to determine whether setting up a program to play quickly through master games would be better than doing tactical problems, going slowly through master games, or other types of chess study.

• #2

a lot of what silman says is good but like all of them you have to filter what works for you i have played chess for 20 years and gone over A LOT of GM games probly not that many but i dont see how going over them fast will do you a darn bit of good seems like takeing your time understanding the moves in a game seems better

• #3
qrayons a écrit :

He says that in order become an IM or GM, you need to look at 100,000 games.

Can any IMs or GMs confirm that this sounds reasonable? or was Silman perhaps exagerating?

By 100 000 games I think he means 99 000 bullet games on chess.com + 900 blitz with friends + real work on 100 WC games.

• #4

Doing it fast just gives you patterns like he says.

A game I recently looked at went to a B+N vs B+B endgame with equal material, but an open position and the bishops ended up winning.  Yeah I knew they were strong, but now I have that pattern in my head (what the pawn structure was, and that B+B gives good chances vs B+N in it).

GMs have (as Silman says) "zillions" of these patterns for reference when they calculate.

But yeah, 100,000 seems like an exaggeration.  Clearly these teenage GMs didn't look at that many.  But maybe Silman realizes our brains (unlike the prodigies) forget half of what we see, so for the common man he's recommending an extra strong dose

Of course going over them slowly is good too.  Silman also recommends this.  He doesn't say how many, but IIRC in HTRYC he mentions having a few notebooks full of your own analysis of GM games you went over slowly is a good idea.

• #5

I think he meant 10 000.

• #6
hicetnunc wrote:

I think he meant 10 000.

I think he meant A LOT of grandmaster games

• #7

You don't need to spend 10 seconds per move if you're going through tons of games. You'll see the same openings over and over, so you'll fly through those. Even in the middle and endgames, if you're just going for quicky pattern recognition, 1-3 seconds per move is enough.

• #8

I don't get it. If you don't understand GM games how are you meant to improve by looking at them. I guess they'd have to be annotated, but I hate looking at variations, they seem kind of pointless half the time. Seriously though, 50% of the moves I see in GM games are quiet moves that I wouldn't be able to tell apart from a 900 player move (I'm not saying they play at a 900 level, just that they look pointless at first sight).

• #9
Fromper wrote:

You don't need to spend 10 seconds per move if you're going through tons of games. You'll see the same openings over and over, so you'll fly through those. Even in the middle and endgames, if you're just going for quicky pattern recognition, 1-3 seconds per move is enough.

quiet wrong, spend extra time in the opening and try to figure out why the moves in the opening are played. For example why this move instead of this right away...

• #10
jetfighter13 wrote:
Fromper wrote:

You don't need to spend 10 seconds per move if you're going through tons of games. You'll see the same openings over and over, so you'll fly through those. Even in the middle and endgames, if you're just going for quicky pattern recognition, 1-3 seconds per move is enough.

quiet wrong, spend extra time in the opening and try to figure out why the moves in the opening are played. For example why this move instead of this right away...

If you're studying games in detail, yes.

If you're studying lots of games just to go through a ton of them and get a "feel" for how GMs play, then no.

These are two different types of master game study. This thread is about the second type - learning by osmosis just by going through a ton of master games very quickly. That's what I was talking about.

If you want to spend time on the individual moves, then you're talking about the first type, which should be done with books of well annotated games with lots of verbal description. Don't bother with the variations unless you have a specific question as you're going through a game.

• #11

watch Kingcrusher on youtube, he has some very good content and has plenty of games explained.

• #12
sammynouri wrote:

I don't get it. If you don't understand GM games how are you meant to improve by looking at them. I guess they'd have to be annotated, but I hate looking at variations, they seem kind of pointless half the time. Seriously though, 50% of the moves I see in GM games are quiet moves that I wouldn't be able to tell apart from a 900 player move (I'm not saying they play at a 900 level, just that they look pointless at first sight).

Just for basic patterns like I mentioned before.  Mini lessons like, oh, in this structure this attack is effective.  Or in that opening white can play for a central break.  Or such and such an endgame is a win or a draw.

The goal isn't to actually understand why.  As you said you can't really do that on your own.  But lets say in your next game you have a chance to go into an endgame (as I mention in my first post in this topic).  And hey, if you can get that certain pawn structure you know it gives you winning chances.  So you play exd (or something), the forced recapture achieves the structure, and you go into the endgame with a rook trade.

Or you're on the weak end and your opponent threatens to do it, so you avoid the trade.  Or you avoid the structure.  etc.

• #13

Silman is a great chess coach and I certainly am not qualified to question his statements but just to put my own two cents in, it seems to me that there is more value in going over annotated games and making an effort to understand why the masters actually made the moves they made. I have found personally that the most interesting way to do this is to watch the chess videos that are found on this website and other chess websites as well as youtube. I have also found move by move books and best games books to be helpful. Going through games without notes does show you patterns and may also be helpful.

• #14

Cross posted from the original article:

Ok, so I decided to run an experiment.

I loaded up HIARCS for iPad and opened up one of the master game databases (Fisher v Spassky) then set the replay to 5s per move and played through the database.

I found I can follow the game reasonably at that speed and each game takes around 6 minutes to play through.

So 10 games/hr, and at 5 hrs per day (50 game/day), 7 days/week (350games/wk) for six months (x26 weeks = 9100 games (I did that sum in my head, so feel free to check with a calculator someone).

That's remarkably close to the more commonly quoted 10,000 master games to reach a title (note: double that to get from FM to IM and double again to get to GM?)

Add on 45 mins a day of tactics training, 15 mins of thinking method / time management study and playing 2 slow games a week (5hrs each including post-mortem) and you get an additional 17 hrs per week on top of the 35hrs you are playing through games.

That's 10hrs play and 42hrs a week of chess study for 6 months. Alburt says in his chess training program that the Russian chess schools were training their students 10hrs per week (excl playing time) for 4.5yrs to get a title (IIRC, again feel free to check). That's 450x4.5 (how many weeks off did those Russian's get each year?) = 2000 hrs approx total for the Russians (4,000 including playing time). Double that to get from FM to IM and you have about 8,000 hours which again is close to the oft-quoted 10,000 hrs to master anything.

Baffo's training is a total of 35x26 = 1610 hrs, a little less than the Russians (again, I'm doing these in my head!).

All the numbers seem to point to similar numbers:

FM takes 5,000hrs (half study and half play).

IM takes 10,000hrs (half study and half play).

One could then suggest that to get from IM to GM would take the same effort again, a total of 20,000 hours. Lets assume that 10,000 hours of that was spent analysing master games at 10 games per hour, then Silman's idea of 100,000 games sound more realistic.

(As it should, who wants to argue with Silman about how to learn to become a master?)

Most experts says equal playing and study time is important, so Baffo's regimen seems light on playing time (!) as you would need 70hrs a week to equal just his study time with playing time.

• #15

10000 hrs = GM

• #16
RobbieCoull a écrit :

I loaded up HIARCS for iPad and opened up one of the master game databases (Fisher v Spassky) then set the replay to 5s per move and played through the database.

I found I can follow the game reasonably at that speed and each game takes around 6 minutes to play through.

All the numbers seem to point to similar numbers:

FM takes 5,000hrs (half study and half play).

IM takes 10,000hrs (half study and half play).

One could then suggest that to get from IM to GM would take the same effort again, a total of 20,000 hours. Lets assume that 10,000 hours of that was spent analysing master games at 10 games per hour, then Silman's idea of 100,000 games sound more realistic.

"Each game take around 6 minutes" ? And you think you could progress with that ?

When I study a game (a well annotated one, from world champions or top GMs), he takes me 5, 10, 20 and sometimes 50 hours to master it. You have to work on many 'key positions' on an single game, and you must take 30 minutes or 1 hour (or more) to solve (or just try to solve) 1 position.

5 000 or 10 000 hours of study/play to become a FM/IM/GM is probably ok, but 100 000 games is not realistic at all.

• #17

Try to work this annotated game (with just 10 'key positions' = 10 diagrams to solve) :

If you're not able to remember all the moves and 90 % of variations the next day, this work is useless, so take your time to solve the 10 diagrams.

• #18

LePontMirabeau, I am not a chess master but I am an experienced instructor in my field.

From an educationalist point of view, spending time studying master games in depth is very useful. However, you are likely to progress much more quickly in a pattern recognition subject like chess by studying a large number of games in very low detail.

This also holds true for tactics. Doing lots of tactics quickly is a much better use of time than studying a small number of tactics in detail.

Running through large numbers of patterns quickly is a system 1 training tool. ie: you will develop heuristics and subconsious pattern recognition skills.

Doing a small number of detailed analysis develops your system 2 thinking skills, which involve conscious thought.

Both are required for progression, but experts recommend 90% system 1 training and 10% system 2 training (or 80:20 or 70:30). I'm not aware of any experts recommending only system 2 training for chess.

• #19
RobbieCoull a écrit :

Running through large numbers of patterns quickly is a system 1 training tool. ie: you will develop heuristics and subconsious pattern recognition skills.

Doing a small number of detailed analysis develops your system 2 thinking skills, which involve conscious thought.

Both are required for progression, but experts recommend 90% system 1 training and 10% system 2 training (or 80:20 or 70:30). I'm not aware of any experts recommending only system 2 training for chess.

I don't know any serious trainer or GM who would recommand system 1 training. Could you name me one ? I've heard many time 'bullet kill your brain' from many top players. I love bullet games but the 'subconcious pattern recognition skills' you get with them (I understand that, I've play a lot of bullet) is just very bad for long games.

I know a lot of good trainers/ GMs who recommand only system 2, I could name many of them.

• #20

I'm interested to know the trainers who put a strong emphasis on system 2. Don't they recommend a heamthy dose of system 1 as well ?

Top
or

Online Now