Igo Hatsuyōron 120 and
Go-Playing Programs (2009 - 2011)
Some Typical Results
|
Here we are in the F-path ( | |||||||
| Komi | A | B | C | D | E | Y | X |
The table indicates, for these candidate points, and for different Komi, the percentage of games, which were won for Black, after a start at this point, by Many Faces. Please understand negative values of the Komi (Many Faces allows integers only) as additional prisoners given to White beforehand, resulting in an initial disadvantage for Black. The thick line in the table separates the estimated victories from the estimated losses. At the point of change, the corresponding Komi can be interpreted as an estimate of the final result by the program, from White's point of view. Here, a jigo results with six, or seven, prisoners given to White, so the result of the game - without Komi given - can be assumed to be at least six points for Black. It is clearly visible that the program put more points on its shortlist, when it believed, following its own evaluation, that is was behind. The bold and larger font indicates the move, which Many Faces would play next, as Black.
In this range of Komi, there typically also arise two types of moves that are not always useful for further analysis, and therefore must be examined more closely by humans . First, there are moves, which are not correct, like,
The move at |
| -11 | 53 | |||||||
| -8 | 52 | 51 | ||||||
| -5 | 49 | |||||||
| -2 | 46 | |||||||
| 1.5 | 43 | 42 | ||||||
| 4.5 | 41 | 40 | 41 | 40 | 41 | |||
| 7.5 | 40 | 39 | 38 | 40 | 39 | |||
| 10.5 | 38 | 37 | 37 | 38 | 37 | |||
|
After manually inserting the move sequence from | ||||||
| Komi | B | F | G | H | W | V |
The 50% line in the table is lower now. White has apparently lost some points in the upper right corner, compared with what had previously been calculated by the program.
The choice of the moves
Here again, with the computer "feeling" that it is lagging behind, it considers some incorrect moves (like |
| -11 | 42 | 42 | 41 | 42 | 42 | ||
| -8 | 44 | 44 | 44 | 45 | 45 | ||
| -5 | 46 | 48 | 46 | ||||
| -2 | 49 | 49 | 48 | 49 | |||
| 1.5 | 53 | 52 | |||||
| 4.5 | 57 | ||||||
| 7.5 | 59 | ||||||
| 10.5 | 61 | 59 | |||||
|
For | ||||||||
| Komi | J | L | K | F | M | G | H | U-R |
... the program recommends a move at With a larger "perceived" gap we can again see the tendency to play "desperation moves", like |
| -11 | 59 | ||||||||
| -8 | 57 | ||||||||
| -5 | 55 | ||||||||
| -2 | 52 | ||||||||
| 1.5 | 47 | ||||||||
| 4.5 | 43 | 44 | |||||||
| 7.5 | 41 | 40 | 40 | 40 | 40 | ||||
| 10.5 | 38 | 39 | 38 | 38 | 38 | 38 | 38 | 38 | |
|
With |
|||||
| Komi | G | F | K | J | V |
If the program "feels" ahead, it plays the safety-first move of In any case, some of the new candidates again gave us the desire to initiate further investigations. |
| -11 | 41 | 40 | 41 | |||
| -8 | 43 | 42 | 43 | |||
| -5 | 45 | 44 | 44 | 45 | ||
| -2 | 48 | 48 | 48 | 48 | ||
| 1.5 | 53 | |||||
| 4.5 | 57 | |||||
| 7.5 | 59 | |||||
| 10.5 | 62 | |||||

434
/
), White has just played at
. Marked in the diagram are the moves, which were displayed as candidates for the next move,
, by Many Faces.
and
, here. On the other hand, some moves are displayed that have the fundamental character of forcing moves, like
,
,
, and
, here. Only in very rare cases, as here with
ensures two eyes for the Black group in the top right, and is correct.
to fix the shape in the upper right corner (Many Faces would not immediately play at
), we will continue with the move of
(and the proposal for
) were the first indication of a certain miai relationship between the upper left corner and the left side.
) and simple forcing moves (like
, and
, which here include a potential subsequent loss).
, ...
, which we had not previously recognized as being promising. However, later extensive analysis showed us the value of this move.
, and
.
, the game seems to have reached a critical stage, because the evaluation of the program, with varying Komi, is not as uniform as before.
