If game Handol played White 1, Black 2, White 3. Win percentage dropped from 29 to 14 after Black 2 (not immediately but after some thinking). And after 3 it’s clearly lost for white, no other options.
But what if W 1, B 2 and then A. Or simply Black B? And from Lee’s analysis white’s move could also be K3.
I left some variations from video streams from after the game analysis by Lee and commentators. Please look at them.
I can’t believe it! He won!!! There is blood in the water, and I’m switching my vote to Lee Sedol. Count the minutes that are left before the end Handol… Lee is all that is (Hu)MAN!!!
Unfortunately, we do not quite enough information about the strength of those programs. Actually this is why this match was interesting for me. To see if the bot really can win with handicap.
And in the guessing of the outcome I was mislead by the info of the top 5 Koreans being defeated by Handol. I somehow assumed that was 2 handicap games. Actually was even games. And since master already defeated every top player in those fast online games two years ago, repeating the feat at even games seemed redundant, and that is why I assumed that these games were at handicap. And even so, Handol is not the strongest. Actually how much stronger are the other ones?
And some more, The bot seemed cheated at time. The official timer was for the stones placed on goban. But the handler used some tome to relay. It was not very much, maybe three to five minutes in total, and there was plenty of time left, but who knows, those few minutes maybe counted on reducing the play-outs. Or maybe not, maybe the internal clock of the bot was according to the moves played on the laptop, and in the end maybe the official clock would have been adjusted.
Anyway, we really need more matches like this to see if bot can really give handicap. Since those predictions improved for the bot, it seems that is able to recover the difference, the ladder reading could been botched in an even game too. Anyway, it still had a lot to recover, so maybe 2 stones makes the game even. 3 for sure are too many, and confirms what the pros always said being able to win with 3 to Go God. And even there are better bots, seems that they are just fractionally stronger. Strong enough to win every time at even games, but not so strong to give 3 stones. We know that at kyu level a rank is a stone, at dan a stone is for two ranks. So for sure at that level, o stone is equivalent of three or four ranks. So to win with 3 the bot should be maybe 20 dan.
I wondered whether the whole setup of this long netting sequence was deliberate, to try to create playout-intensive sequences that the bot would not read to the end… imagine having the skill to do something like that on purpose, if this is what it was!!
In an interview developers said that they didn’t expect this outcome because they tested Handol in handicap games against famous professionals and the conclusion was that program has no problems winning giving two stones against komi.
That’s funny. I haven’t seen that, but I did get blocked from posting one of Yeonwoo’s videos a couple of weeks ago because it would be the third post in a row from me
Means all the people who voted for a 3-0 streak for Lee are also knocked out, leaving only the (sensible?) 50% of people who either voted 2-1 or 1-2… we shall see in 2 days time!!!
that’s funny
