Difference between revisions of "Mate Distance Pruning"
GerdIsenberg (talk | contribs) (Created page with "'''Home * Search * Selectivity * Pruning * Mate Distance Pruning''' If a forced mate has been found, '''Mate Distance Pruning''' (MDP)...") |
GerdIsenberg (talk | contribs) m |
||
Line 42: | Line 42: | ||
! Depth | ! Depth | ||
! MDP Disabled | ! MDP Disabled | ||
− | (Score / Nodes) <ref>Nodes until [[Principal | + | (Score / Nodes) <ref>Nodes until [[Principal Variation|PV]] is found</ref> |
! MDP Enabled | ! MDP Enabled | ||
(Score / Nodes) | (Score / Nodes) |
Latest revision as of 10:57, 12 May 2018
Home * Search * Selectivity * Pruning * Mate Distance Pruning
If a forced mate has been found, Mate Distance Pruning (MDP) will cut trees and adjust bounds of lines where no shorter mate is possible. Mate Distance Pruning is a safe type of pruning, as it only cuts irrelevant nodes. It will not add much to a programs playing strength as this pruning only comes into effect when a position is already decided, still it helps with analysis and solving "Mate in n" problems.
Mating Value
The prerequisite for Mate Distance Pruning is that decided positions are properly scored. So a position where the opponent is mated is scored SCORE_MATE (a very high value), a position where the victory is one move away is worth less SCORE_MATE - 1, Mate in 2 plies is scored SCORE_MATE - 2 and so on. On the other side, a position where the side to move is mated in n plies is evaluated -SCORE_MATE + n. The difference between the actual value and SCORE_MATE is always the number of plies the mate is away from the root position.
Upper Bound
Assume we are in a winning position. Through another line we have already found a forced mate in n plies. Thus our alpha is SCORE_MATE - n. From this information we can take that should we be searching a line that already is equal to or longer than n, we can impossibly increase alpha even if another mate was found. This means we could safely prune that tree.
Furthermore we can lower the upper bound. If we are n plies away from the root, the opponent can impossibly be evaluated > SCORE_MATE - n. Thus beta can be set to this value should it be higher.
mating_value = SCORE_MATE - RootDistance; if (mating_value < beta) { beta = mating_value; if (alpha >= mating_value) return mating_value; }
Lower Bound
Contrary to the pruning if we are in a winning position, scores can also be adjusted and trees pruned if we are in a loosing position.
mating_value = -SCORE_MATE + RootDistance; if (mating_value > alpha) { alpha = mating_value; if (beta <= mating_value) return mating_value; }
In Practice
Using the Chess Engine Glass 1.0 the treesize of a search without Mate Distance Pruning is compared to one with this feature enabled.
♔ ♖ ♚ |
8/7K/8/8/8/8/R7/7k w - - 0 1
Depth | MDP Disabled
(Score / Nodes) [1] |
MDP Enabled
(Score / Nodes) |
---|---|---|
1 | 1041 / 3 | 1041 / 3 |
2 | 1041 / 23 | 1041 / 23 |
3 | 1051 / 103 | 1051 / 103 |
4 | 1051 / 551 | 1051 / 551 |
5 | 1055 / 2151 | 1055 / 2151 |
6 | 1051 / 7638 | 1051 / 7638 |
7 | 1055 / 24353 | 1055 / 24353 |
8 | 1065 / 62692 | 1065 / 62692 |
9 | 1065 / 139241 | 1065 / 139241 |
10 | 1085 / 253338 | 1085 / 253338 |
11 | 1085 / 409101 | 1085 / 409101 |
12 | 1105 / 610285 | 1105 / 610285 |
13 | 1105 / 873959 | 1105 / 873959 |
14 | 1105 / 1417047 | 1105 / 1417047 |
15 | 1105 / 1907578 | 1105 / 1907578 |
16 | M8 / 2846864 | M8 / 2547293 |
17 | M8 / 3715208 | M8 / 2819164 |
18 | M8 / 4842666 | M8 / 3092833 |
19 | M8 / 6181827 | M8 / 3368068 |
20 | M8 / 7774698 | M8 / 3643489 |
From this table one can see that the treesize is equal for the first 15 plies, but then the Version with Mate Distance Pruning enabled found the exact mating line about 10% smaller. Furthermore the risk of treesize explosion for further iterations is greatly decreased. In this example the treesize of the Version with this type of Pruning enabled is less than half as big.
Forum Posts
- Search questions by Sven Schüle, Winboard Forum, July 17, 2007» Fail-Soft
- Discussion about Mate Distance Pruning by Don Dailey, CCC, March 14, 2009
- mate distance pruning problems and static null move pruning by Pierre Bokma, CCC, December 04, 2011 » Reverse Futility Pruning