Why do bots - even Alphago - sometimes go on "tilt"?
The answer is given here:
http://lcipm.blogspot.com.au/2016/09/th ... -tilt.html
The Reason for Tilt
-
Pio2001
- Lives in gote
- Posts: 418
- Joined: Mon Feb 16, 2015 12:13 pm
- Rank: kgs 5 kyu
- GD Posts: 0
- KGS: Pio2001
- Has thanked: 9 times
- Been thanked: 83 times
Re: The Reason for Tilt
Hi,
Something is strange in the explanation. The percentage of winning probability has nothing to do with the possible answers, has it ?
After the first stone is played, there are about 26 correct moves out of 360 intersections. Does it mean that the winning probability for black in an equal game is (360-26)/360 = 93 % ??
In the tree search algorithm, the AI only takes into account the value of the strongest answer to estimate the value of the considered move. The fact that some weak answers exist doesn't make a move good.
Something is strange in the explanation. The percentage of winning probability has nothing to do with the possible answers, has it ?
After the first stone is played, there are about 26 correct moves out of 360 intersections. Does it mean that the winning probability for black in an equal game is (360-26)/360 = 93 % ??
In the tree search algorithm, the AI only takes into account the value of the strongest answer to estimate the value of the considered move. The fact that some weak answers exist doesn't make a move good.
-
djhbrown
- Lives in gote
- Posts: 392
- Joined: Tue Sep 15, 2015 5:00 pm
- Rank: NR
- GD Posts: 0
- Has thanked: 23 times
- Been thanked: 43 times
Re: The Reason for Tilt
the probability of a move winning is independent of the probabilities of other possible moves winning.Pio2001 wrote:Hi,
Something is strange in the explanation.
Last edited by djhbrown on Tue May 02, 2017 12:42 am, edited 1 time in total.
-
Kirby
- Honinbo
- Posts: 9553
- Joined: Wed Feb 24, 2010 6:04 pm
- GD Posts: 0
- KGS: Kirby
- Tygem: 커비라고해
- Has thanked: 1583 times
- Been thanked: 1707 times
Re: The Reason for Tilt
What a cordial replydjhbrown wrote:the explanation looks strange to you because you have misread it and muddled it up in your mind.Pio2001 wrote:Hi,
Something is strange in the explanation.
be immersed
-
sparky314
- Lives with ko
- Posts: 189
- Joined: Wed Mar 18, 2015 1:40 pm
- Rank: KGS 3 kyu
- GD Posts: 0
- Location: Chicago, IL
- Has thanked: 159 times
- Been thanked: 36 times
Re: The Reason for Tilt
FYI, the policy network prevents AlphaGo from "always being on tilt."
The article just looks like a reason to promote a paper.
The article just looks like a reason to promote a paper.
-
djhbrown
- Lives in gote
- Posts: 392
- Joined: Tue Sep 15, 2015 5:00 pm
- Rank: NR
- GD Posts: 0
- Has thanked: 23 times
- Been thanked: 43 times
Re: The Reason for Tilt
Feedback is always helpful, as it enables an author to discover where something he thought was as plain as day can instead be cloudy.
Last edited by djhbrown on Tue May 02, 2017 12:43 am, edited 1 time in total.
-
sparky314
- Lives with ko
- Posts: 189
- Joined: Wed Mar 18, 2015 1:40 pm
- Rank: KGS 3 kyu
- GD Posts: 0
- Location: Chicago, IL
- Has thanked: 159 times
- Been thanked: 36 times
Re: The Reason for Tilt
If you want to promote a paper, than promote the paper on its own merits.
If you want to discuss whether AlphaGo "always plays on tilt," than do that.
Are they related? No. They are not.
I won't continue this discussion past this point, since it's apparent you're not interested in a discussion, given your previous responses.
If you want to discuss whether AlphaGo "always plays on tilt," than do that.
Are they related? No. They are not.
Glass houses.djhbrown wrote: But of course, it is Quixotic of me to imagine that my words could have any effect other than to trigger hypothalmic reflex reactions in stone-throwing faithful acolytes of the new world order of Go.
I won't continue this discussion past this point, since it's apparent you're not interested in a discussion, given your previous responses.