PhilGoetz comments on Rationality Lessons in the Game of Go - Less Wrong

40 Post author: GreenRoot 21 August 2010 02:33PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (145)

You are viewing a single comment's thread. Show more comments above.

Comment author: PhilGoetz 24 August 2010 02:25:49PM 0 points [-]

Try redoing the calculation with geometric averaging: 300 moves, 150 of which are yours, suppose the final score is 80 to 70:

  • x^150 = 70, x = (exp (/ (log 70) 150)) = 1.028728
  • y^150 = 80, y= 1.029644
  • y / x = 1.00089
Comment author: lavalamp 24 August 2010 05:38:57PM 1 point [-]

I don't think that's an improvement. As I said in another comment just now, I think that in go having a small advantage does not make you more likely to gain additional advantages.

Comment author: PhilGoetz 26 August 2010 10:48:31PM *  0 points [-]

Then why does handicapping work? Giving someone 3 stones on star points at the start of a game will have a much larger impact than giving them 3 stones on star points at the end of the game.

Comment author: lavalamp 27 August 2010 04:44:24PM 0 points [-]

I finally saw your point-- moves are more valuable at the beginning of the game, mistakes come at a more or less constant rate, therefore the margin of victory shouldn't be divided up evenly into every move of the game. Yes.

I tried to put a blanket disclaimer in my post that started this thread ("There are some problems with averaging things like this which I probably don't need to point out to you all...") in the interest of brevity but perhaps that was a mistake.

There are problems with my calculation that yours does not solve. Namely, mistakes do not tend to be small and come at a constant rate. If I lose by 10 points it's entirely possible that I made a single 20 point mistake and my opponent made 10 single point mistakes. (well, for example only. In reality amateurs make a lot more mistakes than that)

That said, now that I understand why you suggested it, your calculation does represent the situation more accurately.

The escalate/accumulate/linear/exponential discussion threw me off, as did the fact that I was looking for an answer expressed in points (it's easier to visualize what that means), and the fact that I have seen this calculation done by stronger players than I am. Obviously an answer expressed in points can't be constant throughout the game, and I should have seen that.