Re: Topics on influence

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

After I find the strength of each group, I radiate influence
from each stone.  The base influence is from a table, and
depends on the group strength.  Influence added at each
point is a function of the distance (table lookup), multiplied
by the base.

My current table of value vs distance is:

static float rtdist[] = {  /* falloff in influence with distance */
0.f, 1.f, .5f, .33333f, .24f, .16f, .1f, .067f, .0444f,
.03f, .02f, .0133f, .009f, .006f, .004f };

As you can see, the influence falls off a little bit faster than
1/distance.

I radiate influence away from each stone, but not through other
stones, and not through tight connections like one point jumps.

I use influence to detect territory, as below, and to tell if
a group is surrounded.  To see if a group can run, I look out
a few points away from the group at the friendly and enemy
influence, subtract the influence from the group itself, and 
see if there is more friendly or enemy, and if each is
increasing or decreasing as you move away from the group.

David



At 07:36 AM 3/31/99 -0800, Stuart Cracraft wrote:
>Can you provide the function you used to calculate those numbers
>as well as its method of use?
>
>At 07:16 AM 3/31/99 -0800, you wrote:
>>
>>Okay, let's say there is a corner like this:
>>
>>
>>    . . . . . . 
>>    . . . . . . 
>>    O O X X O .
>>    . X O O O . 
>>    . X X . . . 
>>
>>What would a common influence function calculate for the
>>above (per square)?
>>
>>With 3 tactically unsettled groups on the board the 
>>influence doesn't mean much.  My influence function
>>has high white influence everywhere, and black influence
>>very near 0 everywhere.  The white influence comes from
>>the 4 stone group.
>>
>>The territory estimates, in 50ths of a point, with positive
>>better for white are:
>>
>>6
>>11 8
>>17 13 8  6  9  8
>>26 19 15 12 13 13 8
>>39 30 29 20 21 19 13 8
>>0  0  0  0  11 27 17 12 7
>>0  0 -15-15 45 45 28 19 13 9  5  3  2  1  1
>>6  6  6  45 45 45 45 31 21 15 10 6  4  2  1
>>6  6  6  45 50 47 32 20 12 9  6  4  2  1
>>
>>I only have one influence function at the moment that I
>>use for territory estimation.  You can look at the
>>handtalk web page for a description of his influence
>>function that he uses for connectivity.
>>
>>David
>>
>>At 09:02 PM 3/30/99 -0800, you wrote:
>>>Could you give examples of each for the example given,
>>>with corner in lower-left.
>>>
>>>Looking for three sample influence functions for the
>>>three purposes you mentioned.
>>>
>>>At 08:21 PM 3/30/99 -0800, David Fotland wrote:
>>>>
>>>>
>>>>I can't tell which is the corner :) 
>>>>
>>>>Influence functions can be used to help determine
>>>>connectivity, help tell if a group is surrounded, and
>>>>help find secure territory.  I suspect a different
>>>>function is ideal for each purpose.
>>>>
>>>>David
>>>>
>>>>At 07:31 AM 3/30/99 -0800, Stuart Cracraft wrote:
>>>>>
>>>>>Okay, let's say there is a corner like this:
>>>>>
>>>>>
>>>>>    . . . . . . 
>>>>>    . . . . . . 
>>>>>    O O X X O .
>>>>>    . X O O O . 
>>>>>    . X X . . . 
>>>>>
>>>>>What would a common influence function calculate for the
>>>>>above (per square)?
>>>>>
>>>>>I am interested particularly in the influence function that
>>>>>calculates the likelihood of a particular stone surviving.
>>>>>
>>>>>--Stuart
>>>>>
>>>>>At 09:45 PM 3/29/99 -0800, you wrote:
>>>>>>
>>>>>>
>>>>>>I just use it tell if groups are surrounded, and to estimate
>>>>>>the territory count and thickness.  I don't make high level
>>>>>>plans.  Just take it one move at a time.
>>>>>>
>>>>>>David
>>>>>>
>>>>>>
>>>>>>At 12:34 PM 3/30/99, Darren Cook wrote:
>>>>>>>>I like best the idea that the influnce of each color measures the
>>>>>>>>probability that a stone of that color can be placed there and live.
>>>>>>>
>>>>>>>This is a nice complement to the idea that territory is a point where a
>>>>>>>stone of the opposite colour will be dead.
>>>>>>>
>>>>>>>Once you've found influence what do you do with it? A human might use
>>>it to
>>>>>>>construct high level plans, but are computer programs really upto that
>>>kind
>>>>>>>of thing?
>>>>>>>
>>>>>>>Darren
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>>
>>
>>
>>
>>
>
>