Temperature?

[Mef]

The kikashi only cools the board, just like every other move, since the player removes that point from the available moves. Playing the kikashi removes available points from play. Essentially the big points will get played and the net some of the kikashi + big points is that there are fewer points on the board than playing the big points only.

Suppose there is a two point kikashi and five point play after which there are only one point plays. The two point kikashi serves to cool the board just like any other play, however it is urgent.

If it is a two point kikashi, then prior to your move, the temperature will be 2 points (difference in your play vs. their play). After you play your move but before your opponent responds, the temperature is now going to be the difference between responding to the kikashi (returning to status quo) and letting you follow up on the kikashi (presumably worth much more than two points, if your move truly merits a response). The result is that the temperature on the board is now much higher, and after your opponent responds it cools off.

[tapir]

Most of the time temperature is just a metaphor for urgency (hot, cold, cooling, heating up, in this thread: crystallize) and it is more mystifying than helpful to read much into it. If you count, talking about miai value/size of play works fine without metaphors.

[SmoothOper]

The kikashi only cools the board, just like every other move, since the player removes that point from the available moves. Playing the kikashi removes available points from play. Essentially the big points will get played and the net some of the kikashi + big points is that there are fewer points on the board than playing the big points only.

Suppose there is a two point kikashi and five point play after which there are only one point plays. The two point kikashi serves to cool the board just like any other play, however it is urgent.

Game temperature always cools and ambient temperature remains the same:

http://senseis.xmp.net/?GameTemperature

Average value of gote play.

http://senseis.xmp.net/?TemperatureCGT

The bottom line is before and after kikashi, the largest point on the board would still be five points. I think many beginners don't quite understand this, and play inside frame works just to get a response, but who am I to correct that behavior.

[tapir]

Game temperature always cools and ambient temperature remains the same:

http://senseis.xmp.net/?GameTemperature

"It is often useful to think of a game temperature that never rises during the game. Even if a hot battle erupts, as a rule it heats up a local region, while the ambient temperature remains the same."

You are aware that game temperature in the first sentence is synonymous to ambient temperature in the second. Imho this is a good example of the mystifying nature of these concepts when taken too serious. There was a perfectly clear summary above, but some posts later suddenly concepts are multiplying and everyone is confused.

[Bill Spight]

Game temperature always cools and ambient temperature remains the same:

http://senseis.xmp.net/?GameTemperature

"It is often useful to think of a game temperature that never rises during the game. Even if a hot battle erupts, as a rule it heats up a local region, while the ambient temperature remains the same."

You are aware that game temperature in the first sentence is synonymous to ambient temperature in the second. Imho this is a good example of the mystifying nature of these concepts when taken too serious. There was a perfectly clear summary above, but some posts later suddenly concepts are multiplying and everyone is confused.

A quick history of temperature in go:

Temperature started as a technical term in combinatorial game theory. It applies to games, which in go are independent regions of play (or the whole board). It corresponds to miai value in go, even though miai value applies to plays. A gote or reverse sente in a region gains on average at most the temperature of the region.

In the 1990s temperature started to be used by amateur go players on reg.games.go in reference to the whole board (whole board temperature) or to the rest of the board aside from the region of play under consideration (ambient temperature). In those terms you can say that a play is likely to be played with sente (at least for one play) if it leaves the local temperature higher than the ambient temperature.

Later yours truly used the idea of the temperature of the environment to define the temperature of combinatorial games. This enables us to evaluate positions with multiple kos or superkos.

Now, since temperature in go is a regular word, like a regular word it is not surprising that it is somewhat ambiguous. The same thing is true for many go terms that have a long history, such as sente and gote.

As for the idea of a game temperature that never rises, Berlekamp showed that if you start play in a game at a certain temperature and the temperature goes up, it is in some cases possible to err by making the play with the highest miai value. You can avoid such errors by keeping in mind the starting temperature. Similarly, if a local region heats up, the ambient temperature does not change (except perhaps in regard to ko, but kos break independence). It is generally good to think about sequences that raise the temperature as temporary and local.

[Bill Spight]

Recently there has been some discussion about temperature in go games. I've vaguely heard of the term before (http://senseis.xmp.net/?Temperature), and hearing about it reminds me of Berlekamp and Wolfe's book.

But I've never studied this topic of "temperature" much, as I figured it was mostly good for theoretical study and endgame in computer programs.

Seeing the discussion on the forum, though, it sounds like a lot of people understand temperature better than I do, which leads me to these questions:
* Is temperature practically useful when you're playing games?
* Do you consider temperature even during the fuseki?
* If temperature is practically useful in games, what makes it superior to other ways of playing?
* Any useful tips about temperature that will help me understand it better?

In an independent region of the board there will normally be some gote and reverse sente plays (or sequences of play) that gain something. The most that any of these plays or sequences gains is the (local) temperature of that region. Once that play or sequence is played, normally the local temperature drops.

A play may raise the local temperature, initiating a sequence of plays that end in a lower temperature than the original temperature. If the number of plays in the sequence is even, the sequence is sente, as is the original play. If it is odd, the sequence is gote.

If a sequence of play returns to the original temperature, that sequence is ambiguous between sente and gote.

Even when regions of the board are not strictly independent, the concept of temperature may be useful. A few centuries ago it was common for a game to start with a 3-4 play, followed by a 5-3 approach, followed by a pincer. Over time, pros realized that the approach was not sente, and stopped making the pincer. Then they realized that the first play was not sente, either, and started playing more often in an open corner. The original 3-4 play lowers the local temperature in the corner.

An enclosure lowers the local temperature in the corner, but it may raise the temperature of the side that the enclosure faces. If the opponent then plays a wedge on the side, that lowers the temperature of the side.

Generally as a local position is played out, its local temperature drops. That fact may be useful in the opening in finding opportunities to play elsewhere.

[tapir]

After bugging you, Bill, with question about this for years, I am fairly sure I understand what is meant.

* temperature (common usage) = miai value of largest play
* local temperature = miai value of largest play in local area
* ambient temperature = miai value of largest play outside the local area

Ambient temperature can be seen as a baseline to which temperature in the common use of the term returns after sequences are finished and below which it doesn't fall. Local temperature (while a sequence is played) corresponds to the peaks in temperature in the common use of the term. I believe these concepts can be useful at times.

But:

In practice, it is far from sure that I can distinguish the local area from the rest of the board or that there is only a single sequence in play at one point. Something as basic as responding to a first line hane with a first line hane elsewhere knocks the distinction out (and this is endgame play where the parts of the board are as independent of each other as it gets during a game).

For common usage I don't see any benefits of saying "settling the group cooled down this area and other parts of the board are hotter now" instead of "after settling the group playing elsewhere is larger" or "this move heated up this area over the ambient temperature and there is an even number of moves in the sequence" instead of "this sequence can be played in sente now". Basically people are rephrasing their insights in a metaphoric language, which works for them, but leaves a visible question mark in the faces of bystanders.

[Bill Spight]

After bugging you, Bill, with question about this for years, I am fairly sure I understand what is meant.

* temperature (common usage) = miai value of largest play
* local temperature = miai value of largest play in local area
* ambient temperature = miai value of largest play outside the local area

Ambient temperature can be seen as a baseline to which temperature in the common use of the term returns after sequences are finished and below which it doesn't fall. Local temperature (while a sequence is played) corresponds to the peaks in temperature in the common use of the term. I believe these concepts can be useful at times.

But:

In practice, it is far from sure that I can distinguish the local area from the rest of the board or that there is only a single sequence in play at one point. Something as basic as responding to a first line hane with a first line hane elsewhere knocks the distinction out (and this is endgame play where the parts of the board are as independent of each other as it gets during a game).

For common usage I don't see any benefits of saying "settling the group cooled down this area and other parts of the board are hotter now" instead of "after settling the group playing elsewhere is larger" or "this move heated up this area over the ambient temperature and there is an even number of moves in the sequence" instead of "this sequence can be played in sente now". Basically people are rephrasing their insights in a metaphoric language, which works for them, but leaves a visible question mark in the faces of bystanders.

I was surprised when people on rec.games.go started talking about temperature. At the start it was only about global temperature, but spread to local temperature and raising it in sente, and so on. Whether they were using it metaphorically, I don't know. However, I did notice that some go concepts could be described in terms of temperature, as adopted by go players, not as a CGT term. That is why I wrote the SL Temperature page in the first place. Speaking for myself, I do not think of it as a metaphor, but as a concept.

[RobertJasiek]

it is far from sure that I can distinguish the local area from the rest of the board

You can: choose a 'locale' (local set / region of intersections) of your choice! (Not randomly, but choose it as suitably as you can.) Next, for the given locale and the hypothesis that you have chosen it well, study what happens inside the locale and what happens outside the locale. If you cannot construct relevant sequences that are only inside, then you notice that you must have chosen your locale badly; alter it to fit it better! If, however, everything local and relevant happens within the locale, then you know to have chosen it well.

Within the suitably chosen locale, you get the local temperature. If you cannot identify it, then maybe because sequences are not local, that is, local = global. In many practical study applications, local <> global.

or that there is only a single sequence in play at one point.

Why do you care? What matters is whether all relevant sequences / value calculations fit into the / a suitably chosen locale.

For common usage

Temperature is not for "common usage" but is a concept for usage with some degree accuracy far above that of common usage.

I don't see any benefits of saying "settling the group cooled down this area and other parts of the board are hotter now" instead of "after settling the group playing elsewhere is larger"

The latter is a common usage conjecture. Expressing things by means of average move values or possibly temperatures has the potential of verifying or refuting such a conjecture.

or "this move heated up this area over the ambient temperature and there is an even number of moves in the sequence" instead of "this sequence can be played in sente now".

As before.

Basically people are rephrasing their insights in a metaphoric language, which works for them,

Does it? Can they prove that something can be played "in sente"? If they can, fine. If they can't, then more accurate tools such as local move values or temperatures might have a greater chance of allowing a proof.

but leaves a visible question mark in the faces of bystanders.

Sure. New (here: more accurate) concepts for them must be learnt before their question marks vanish.

I have made the experience though that CGT-like temperature talk is hopeless overkill in most practical cases. I prefer to speak of pragmatic compromise versions of more accurate concepts. Such that allow you to choose a locale and such that speak of average local move size instead of local temperature etc. Everybody can divide a difference of positional values by a difference of move numbers. Hardly anybody can apply formal CGT definitions during his games. Therefore I would not recommend the latter to bystanders when the former provides already more than enough accuracy.

[tapir]

it is far from sure that I can distinguish the local area from the rest of the board

You can: choose a 'locale' (local set / region of intersections) of your choice!

I gave the example of answering a first line hane by a first line hane elsewhere - a common play and explained on beginner pages. If I have to include all places, where such a play is possible, in the locale, the distinction between locale and rest of the board becomes dubious. Or during a ko fight, do you really want to include everywhere a ko threat can be played in the locale? Will there be left enough of the board to merit talking about ambient temperature?

Temperature is not for "common usage"

That is my point. I don't doubt that it can be a tool to theorize about positions, but it is hardly the most efficient way to express things.

I was toying around with some derived (and even more contrived) terms in private. I wanted a term for the difference in temperature that a sente move creates (how much it heats up the local area, and I was thinking about sth. like "heating value/Heizwert" ). The reasoning was that whether you can play sth. in sente at that point in the game depends much more on the size of the follow-up than whether it increases temperature over ambient temperature (it does, but that isn't enough). I was also thinking about preparation for a ko, where you play out some moves to increase the size of remaining threats or endgame techniques where seemingly pointless moves are played that turn a move into sente that would not have been otherwise. Later I realized that the whole "temperature" part of what I was thinking was embellishment, metaphoric and completely unnecessary.

[RobertJasiek]

answering a first line hane by a first line hane elsewhere - a common play [...] If I have to include all places, where such a play is possible, in the locale, the distinction between locale and rest of the board becomes dubious.

1) Yes, but...

2) you can set two locales or

3) set a combined locale around all those first line hanes.

Or during a ko fight, do you really want to include everywhere a ko threat can be played in the locale?

It depends on the chosen theoretical model. If your "model" is "read it out on the whole board", then locale is superfluous. If your model is the Unsettled Group Average method, then a locale is very useful.

it is hardly the most efficient way to express things.

It depends.

the whole "temperature" part of what I was thinking was embellishment, metaphoric and completely unnecessary.

Surely there are models for which temperature is unnecessary.