Converting a winning advantage in Go

In a recent stream, Michael Chen 1p made some interesting remarks about how to play when you’re ahead. He went through a professional game in which one player got an early lead, but instead of consolidating and simplifying he seized an opportunity to press his advantage and go for more, even though it introduced more complexity. Michael said that a lot of strong players have this attitude, and he thinks it’s because the risk of blowing your lead by going for more is offset by the risk that your advantage isn’t as big as you thought.

I thought this was interesting, because it’s the opposite of how strong chess players think. If you blunder away a bishop or a knight to a professional chess player, they will take every opportunity to simplify the game by trading away pieces and consolidating their position.

My question is: which philosophy is better for amateur go players?

  • As a beginner I was taught to fight for every point. I think this was good advice, because most people don’t develop positional judgement until well into the SDK’s.
  • In the SDK / low dan range, the advice is often to assess the position and play solidly if you’re ahead or aggressively if you’re behind. This doesn’t necessarily mean “avoid complications even if you an expand your lead”, but that’s often how I play in practice.

What do others think? What do you do when you have an advantage?


I think that strategy of tuning aggressive vs. safe play according to being ahead/behind is a simple strategy that can/should be done at all levels (if winning is the goal). What isn’t simple at all for lower levels is judging if you are ahead or behind (except in pretty extreme cases).


In one of the books of Ishi’s Elementary Go Series (forgot which one) this is perfectly phrased as “A rich man shouldn’t pick a quarrel.” Don’t start a fight if there is much to lose.

Winning a game that already seems / is decided is one of the most difficult things in go. If you play too passive, you lose. And if you play too aggressive, you also lose.


I wouldn’t play safe before move 150 or so. At my level it’s not uncommon to lose a game despite a 20-point lead at the start of endgame.

1 Like

In my thinking this would mean that the game wasn’t decided yet. For me, to feel the game is decided (with me winning) needs the following criteria:

  • we must be in late middle game at least
  • my counting must indicate that I win with at least 10 points if I judge all endgame situations extremly pessimistically

That means I can play passively.

Of course your definition of ‘seems decided’ may differ.

1 Like

Depends on on the size of the lead, ive sometimes gotten a small lead early on but then blundered that late in the game, so i wouldnt “play safe” unless my lead is large enough. Maybe if im sure that im ahead by at least 20 points i could start playing safe?

I ask myself “if I lose this game, will I regret this move?” to try to avoid me doing silly things.


I take this in a bit of different perspective in which i feel like it is not so good to ask too much from your stones. Like if there is an inherent mechanism nearby waiting to crumble your advantage if you push it too far.


Indeed it is easier said than done to properly execute this recommendation of playing safe when ahead. It depends on so many factors.

How solidly can you afford to play while still winning by points? How confident are you in the accuracy of your score evaluation? How confident are you in spotting aji in your own positions? How confident are you in your endgame skills?

That being said:

Some years ago, InSeong Hwang 8d gave me personalised advice to (instead of always trying to win during the middle game) take less risk when I feel I’m leading (or even just having an even game), and work on my evaluation and endgame skills to back up that change in my playing style.

Me trying to follow his advice seems to have led to me losing fewer games against weaker players, but also winning fewer games against stronger players. So it seems my results are more stable now, but not necessarily better.


I think a move that feels risky is often a good one to learn about, and improvising and experimenting are generally important for learning.

Since learning is my high level goal (rather than winning a specific game), I take as many of these risks as I can stomach, and then review after the game. I don’t assume the move was “good” or “bad” based on winning or losing the local battle (or the game); better to review with a stronger player (and/or KataGo).

I ask myself “if I don’t learn what this move means, will I regret it?” to avoid reinforcing blindspots.

EDIT: Note that sometimes what feels risky is to trust my count and play solidly to seal a victory. My general point is, I try to drop my ego about whether I win or lose the game, and instead focus on what I can learn.


My way diverge as i want to ask as less as possible from AIs or stronger players. When i play a move I’m not that much attracted by experimentation but instead by how my move fulfill all the fundamentals i can think of. And i put as much efforts as i can to read if it would work locally at least, like in a semeai. Even if i put all my energy to play what i think mastering, I will have still a lot to discover afterwards by these analysis from stronger to not go chosing slippery paths i can avoid.

Playing to win is still a guide on the road of progress beyond flattering the ego.

1 Like

To be clear, I do play to win, but it’s definitely second in priority to learning (well, I definitely try to make it second priority).

Playing to win is still a guide on the road of progress beyond flattering the ego.

I agree.

But to expand a bit, humans get much slower at learning the more the ego is involved, and the more they’re focused on “winning” instead of the process. One reason is that our ego causes us to avoid experimenting with the unknown or unfamiliar in case we end up looking silly. Toddlers are the ultimate learning machines, partly because they don’t yet have egos that get in the way.

A great book that covers this is The Perfect Wrong Note by William Westney… ostensibly about music, but I think interesting for all those curious about how learning works. Another great book is The Inner Game of Tennis by Timothy Gallwey, about far more than tennis. Both pretty fast reads.


If we have a huge lead like 30+ points, yea we can play it safe. We can avoid complications and secure easy victory. We can still fight but avoid aji heavy moves, we would play moves that are simple and easy to read.

If we only have a 10 points lead, game can go downhill if we play it safe because it’s too easy to turn the game around unless you’re top amateur or pro level.

1 Like

I think a better analogy from chess is “playing for a draw” (because draws in chess are not just possible, but quite common at higher levels). The cliché is that if it’s the last round of a tournament, you only need half a point to finish at the top, and you play for a draw, then you tend to play passively and lose the game. You should still play for a win but be prepared to compromise and accept a draw if the opportunity comes up.


In go, there’s at least a couple of different versions of “play safe”:

  • Endgame, they’re pushing into my territory, do I block or do I fall back and give away 2 points to keep it simple? I know I should block, and if they cut then I’m sure I can capture the cutting stone, but I’ll give away the 2 points anyway just in case I made a mistake in my reading.
  • Middlegame, do I invade their moyo deeply or do I reduce from the outside? Reducing seems to give me a 5 point lead, invading gets into a complicated situation that might give me a 20 point lead but could also turn into a 20 point loss. My instinct is that the invasion is good, but it’s way beyond my reading ability to be 100% sure, so I’ll play safe and reduce instead.

Don’t do the first one of those! That’s how you let your 10 point lead slip away bit by bit and turn into a loss. The second scenario I think is the right way to minimise your risk.


It’s not an either-or, because (above some fairly moderate level, like 1800 maybe) you can’t “pin the blame” for simplification on one side or the other. What happens is more like: the player who’s ahead plays both greedy and safe, i.e. they will post their pieces in the most active way possible, and then when “threatened” with a trade, they just let it happen, because they can.

I’m not sure if this translates to go at all or to what degree if it does.


1 Like

Also, a minor piece is a huge advantage in chess; I’m not sure what the Go equivalent would be, but I assume it would be massive

At least 20 points I would estimate.

1 Like

Such an ugly number: more than 1 stone and less than 2. /jk

A pawn handicap seems to be fairly similar to a handicap stone when measured by the effect on winrate at various levels of play [see Elo Win Probability Calculator]:

Elo points per pawn in chess:

Elo points per stone in go (blue line):

From those graphs, I’d estimate that a pawn is comparable to roughly 1.5 full handicap stones (at least in the range of intermediate to stronger amateurs).

So if a knight is worth 3 pawns, then a knight handicap in chess may be in the range of 5 handicap stones in go?

1 Like