Life In 19x19

While I think this result is both interesting and maybe a bit obvious I'm not sure you can generalize from it to how people react to winning and losing at board games. Money represents the resources you have to have to continue to live comfortably, and anyone who stops caring about money is going to wind up homeless in short order. On the other hand you can lose a lot at go without it hurting you much. So, even if humans have evolved to be really risk-averse, if humans have also evolved to be really flexible it's not hard to imagine that a human might gain more pleasure from things incidental to winning and losing at Go than he suffers pain from losing, even if he loses more than he wins.

Kirby wrote:The argument seems to depend on the idea that komi is equal to 1/2 a stone. I suppose this makes sense: in an even game, black gets a 1 stone bonus to start the game. It's fair to give white an extra half stone of compensation because white gets sente for move two.

The argument that makes "komi = half a stone" crystal clear to me is this: Say that Black begins an even game by passing. What komi should he receive if we want to make it a 50-50 fight again? Obviously it should be the same komi that White receives in a normal game, because the situations are identical (empty board, opponent to move). These two situations (Black to play on an empty board, White to play on an empty board) differ by one move (Black's pass), and by two komi (to go from one to the other we take away White's komi, and then give an equal number of points to Black). So two fair komi = one move.

dfan wrote:

Kirby wrote:The argument seems to depend on the idea that komi is equal to 1/2 a stone. I suppose this makes sense: in an even game, black gets a 1 stone bonus to start the game. It's fair to give white an extra half stone of compensation because white gets sente for move two.

The argument that makes "komi = half a stone" crystal clear to me is this: Say that Black begins an even game by passing. What komi should he receive if we want to make it a 50-50 fight again? Obviously it should be the same komi that White receives in a normal game, because the situations are identical (empty board, opponent to move). These two situations (Black to play on an empty board, White to play on an empty board) differ by one move (Black's pass), and by two komi (to go from one to the other we take away White's komi, and then give an equal number of points to Black). So two fair komi = one move.

This seems (and probably is) very elegant. However, I am having some problems with accepting Black's pass as a move.

SpongeBob wrote:

dfan wrote:The argument that makes "komi = half a stone" crystal clear to me is this: Say that Black begins an even game by passing. What komi should he receive if we want to make it a 50-50 fight again? Obviously it should be the same komi that White receives in a normal game, because the situations are identical (empty board, opponent to move). These two situations (Black to play on an empty board, White to play on an empty board) differ by one move (Black's pass), and by two komi (to go from one to the other we take away White's komi, and then give an equal number of points to Black). So two fair komi = one move.

This seems (and probably is) very elegant. However, I am having some problems with accepting Black's pass as a move.

It is almost tautological if you think about it like this: the value of a move is how much you gain by playing instead of passing.

Sorry to derail the actual topic of the thread...

dfan wrote:

Kirby wrote:The argument seems to depend on the idea that komi is equal to 1/2 a stone. I suppose this makes sense: in an even game, black gets a 1 stone bonus to start the game. It's fair to give white an extra half stone of compensation because white gets sente for move two.

The argument that makes "komi = half a stone" crystal clear to me is this: Say that Black begins an even game by passing. What komi should he receive if we want to make it a 50-50 fight again? Obviously it should be the same komi that White receives in a normal game, because the situations are identical (empty board, opponent to move). These two situations (Black to play on an empty board, White to play on an empty board) differ by one move (Black's pass), and by two komi (to go from one to the other we take away White's komi, and then give an equal number of points to Black). So two fair komi = one move.

Yes, this is basically what I was thinking, but instead of "pass", I thought of it as just adding points and seeing what was fair compensation.

However, I still suspect that the "value of a move" is not constant as the game progresses, since intersections are taken.

Of course, this is true by definition. In general the value of a move slowly decreases, except for occasional large spikes.

Regarding the thread title:

Losing feels horrible - true. But winning is sweet and very satisfying. I think winning is what keeps me playing Go.

(Yeah ... not the whole truth ... a little bit of improvement is also needed)

palapiku wrote:From http://en.wikipedia.org/wiki/Loss_aversion :

"Some studies suggest that losses are twice as powerful, psychologically, as gains."

So if you lose and win equally often, such as with a stable rank on KGS, the experience will be mostly awful. Why do people keep playing this depressing game? I have heard suggestions that you should just play for the enjoyment of playing and not worry about winning or losing. Can anyone honestly say they have achieved that?

Despite fear of becoming a broken record...
My gut tells me a fallacy is nesting here, though it does seem hard to pin down, doesn't it?

Would loss aversion actually rather be suggesting

Go players consider even wins at least twice as valuable as even losses

On the "correct handicap adjustment" front, isn't the only useful way to talk about that observational/experimental? (I was existingly under the impression that eg. IGS handicaps were more appropriate, I guess I forgot to mention it).

Loons wrote:My gut tells me a fallacy is nesting here, though it does seem hard to pin down, doesn't it?

Would loss aversion actually rather be suggesting

Go players consider even wins at least twice as valuable as even losses

How so?

Because if each even game of go is a 50-50 bet, (which we *know* people keep taking)

Loss aversion suggests people would only be taking it if the perceived gain was twice the perceived loss.


I think a lot of alternate explanations could be true as well, though.

I'm in the middle of "Thinking, Fast and Slow". He says that some groups display greatly reduced loss aversion. Poker players and stock traders, for example, IIRC. And experienced collectible card traders, but not inexperienced ones.

To measure your loss aversion, consider this bet:

We flip a coin. If you lose, you pay $100; if you win, you receive $X. How large does X have to be before you would voluntarily take the bet? (Repeat the exercise with $1, $10, $1000, and $100000 values to see how it scales.)

EDIT: Go vote in this poll before reading further, the hidden section may bias you: viewtopic.php?f=8&t=5407

daniel_the_smith wrote:I'm in the middle of "Thinking, Fast and Slow". He says that some groups display greatly reduced loss aversion. Poker players and stock traders, for example, IIRC. And experienced collectible card traders, but not inexperienced ones.

To measure your loss aversion, consider this bet:

We flip a coin. If you lose, you pay $100; if you win, you receive $X. How large does X have to be before you would voluntarily take the bet? (Repeat the exercise with $1, $10, $1000, and $100000 values to see how it scales.)

EDIT: Go vote in this poll before reading further, the hidden section may bias you: viewtopic.php?f=8&t=5407

Most people apparently require X be $200 to take this bet. I require much less, which surprised me as I thought I was more loss averse than average, not less...

All this to say, I suspect that experienced go players may be one of the groups having less than average loss aversion.

(hide a response to dts hide tag)

on the scaling question (would it change with 1,10000$ etc...):
The sum you bet is important: losing 100$ is not pleasant but you can live with it. I suspect people are much more conservative if the money loss can really hurt their life: would you accept to bet 200000 $ even with 75% chance of winning ? math says that you should. I won't because i can t risk my kids being homeless;
loss aversion in go is difficult to relate to money loss i think because do not really lose something material. You can lose at lot in a row without adeverse consequence (well except for the occasional wish to curl up in a ball and cry), as opposed to poker or trading


I have been trying to talk myself into playing the lottery even though i know that the expected return is negative: lets assume i burn 10$ a week during 20 years:
10400$ it seems like a lot but in 20 years its not a life changer. On the other hand if i win a million or more once then my life can change significantly (will i really be happier is another question);
on the other hand i still never bought a lottery ticket despite this nice reasonning.

1. We should be careful: math doesn't say to take the bet. Math might say the bet has positive expected value in terms of money.
2. The lottery reasoning is bad for another reason: people who win the lottery aren't any happier. (http://psycnet.apa.org/index.cfm?fa=buy ... -01001-001)

I would like to test that. Please, someone give me a large sum of money.

Two things.

First, loss aversion is inherently about frames that identify a "status quo" or "breaking even" as the expected or normal situation. How we evaluate a situation is judged by its distance from the status quo, but the metric is weighted different depending on whether the situation is better or worse than the status quo .

Example: Spending $20 on a cheapo go board might be a "loss" from the point of view of an accountant, but you weren't expecting to get the damn thing for free, so parting with that $20 doesn't count as a "loss" for the purposes of loss aversion. However, if you lose the board on the subway right after buying it, the extra $20 you would have to spend to buy a new one feels like a loss, and thus that $20 seems as expensive as ~$40 normally would.

So connecting a "loss" at a board game to a "loss" in psychological prospect theory is completely misguided. Yes, there are people called "sore losers" who really don't like losing games, but this phenomenon is completely different from loss aversion. In a properly handicapped game "win some, lose some" is the frame that shapes your expectations; a loss doesn't matter that much because in the status quo, you lose half your games. If you are on a losing streak, however, or if you get beaten by someone who is nominally much weaker than you, the extra pain this causes may be connected to loss aversion.

Second, there are two sorts of loss-aversion phenomena that get lumped together. One, which is often called risk aversion, is related to diminishing marginal utility. You don't really need the third sock, or the twentieth cookie, or the billionth dollar, or the seventh dead group, so you are relatively reluctant to put sock #2 at risk in order to get sock #3. So thus, someone might be reluctant to take a slightly favorable $100 bet because that would be gambling with the rent money, or something like that. But at the same time, they would prefer a sure $100 gain to a larger, less certain gain: whether it's a gain or a loss has nothing to do with it. The other is what people are normally referring to when they talk about loss aversion: that has nothing to do with whether having twice as much of something is twice as good, and everything to do with having a frame that establishes something as "normal".

In bets like the one Daniel mentions, you can tell the difference between them by comparing the sort of bet you'd be willing to take if it was a one-off bet with bets you'd be willing to take if you could bet 10, or 20, or 100 times. In the latter, you are able to frame the loss from any individual bet as occurring within the larger context of a series of profitable bets in which some losses are expected.