How strong do ya have to be before studying pro games?

[Anzu]

I don't get any of their moves! Should I really be investing time into studying pro games? I know a dan-level player who told me it's a waste of time at my level.

Said, "I myself don't understand much when I look over the games of top players, you shouldn't bother at your level, waste of time if you ask me".

[Bill Spight]

30 Kyu

When should babies start listening to adult conversations?

[RobertJasiek]

If you can get something from pro games, it does not hurt. To really get something from uncommented games, you must be ca. 2k EGF or stronger. Modern games are harder because the opening is more flexible and short thinking times lead to more mistakes, most of which you cannot identify easily yet; such can be hard for strong amateur dans.

You might learn from commented pro games, provided the commentary is written for your level. However, IMO, you can learn faster and more if you learn from textbooks rather than commented pro games. Nevertheless, trying to follow an occasional pro game can add to excitement, if such meets your taste.

[xed_over]

its never too early to start learning from pro games.

[Pippen]

If machine learning teaches us something it is: Try to get to 15k, so you know the rules and basic stuff. Then go over 500 pro games (just memo them) and you'll be a danplayer, 5.000 and you'll be High-Dan, the rest depends on your talent to read. In between play games on a Go-Server, no reviews necessary. Very simple. The only question: Is the game worth all these hours of studying? If yes, then start!

[RobertJasiek]

Wait - AlphaGo used 30,000,000 games.

[skydyr]

There are two ways at least to go over pro games. Playing out each move, thinking about it in depth, considering deep variations, etc. does require a certain amount of preexisting skill.

However, you can also just play over the game in a lighter fashion to train your subconscious. There's nothing intrinsic that tells you what a normal move looks like, so you need to expose it to as many good moves as possible to let it create its own idea of a normal move. You will likely also find yourself being surprised a lot: "why didn't he respond to that hane on the first line, when it's so big?" or some other question where the move seems obvious to you but wasn't played. Commentary can help with understanding this, but it's also good to know that what you thought was right isn't the only option and may be outright wrong.

[dfan]

Wait - AlphaGo used 30,000,000 games.

To be precise, AlphaGo used 30,000,000 positions.

[RobertJasiek]

1 position per game, so 30,000,000 games.

[dfan]

Their paper says

We trained the policy network p_sigma to classify positions according to expert moves played in the KGS data set. This data set contains 29.4 million positions from 160,000 games played by KGS 6 to 9 dan human players.

They did use 30 million positions from 30 million different games during the Reinforcement Learning phase, but those were games that AlphaGo played against itself.

[topazg]

I don't get any of their moves! Should I really be investing time into studying pro games? I know a dan-level player who told me it's a waste of time at my level.

Said, "I myself don't understand much when I look over the games of top players, you shouldn't bother at your level, waste of time if you ask me".

This is a hobby. Do you enjoy it? Then do it, it's never too early. Do you find it incredibly boring? Then don't do it