Difference between revisions of "Aspiration Windows"

From Chessprogramming wiki
Jump to: navigation, search
(6 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
'''[[Main Page|Home]] * [[Search]] * [[Alpha-Beta]] * Aspiration Windows'''
 
'''[[Main Page|Home]] * [[Search]] * [[Alpha-Beta]] * Aspiration Windows'''
  
[[FILE:TheNon-EuclideanWindow.JPG|border|right|thumb|The Non-Euclidean Window <ref>[[Arts#GerhardHoehme|Gerhard Hoehme]] - The Non-Euclidean Window (1964), mixed technique / wood on canvas, [[Arts#ArtMuseumBochum|Art Museum]], [https://en.wikipedia.org/wiki/Bochum Bochum], [https://en.wikipedia.org/wiki/North_Rhine-Westphalia North Rhine-Westphalia], [https://en.wikipedia.org/wiki/Germany Germany], part of [[:Category:Industrial Heritag Trail|The Industrial Heritage Trail]] of the [https://en.wikipedia.org/wiki/Ruhr Ruhr area], Photo by [[Gerd Isenberg]], October 30, 2016</ref> ]]  
+
[[FILE:TheNon-EuclideanWindow.JPG|border|right|thumb|The Non-Euclidean Window <ref>[[:Category:Gerhard Hoehme|Gerhard Hoehme]] - The Non-Euclidean Window (1964), mixed technique / wood on canvas, [[:Category:Art Museum Bochum|Art Museum]], [https://en.wikipedia.org/wiki/Bochum Bochum], [https://en.wikipedia.org/wiki/North_Rhine-Westphalia North Rhine-Westphalia], [https://en.wikipedia.org/wiki/Germany Germany], Photo by [[Gerd Isenberg]], October 30, 2016</ref> ]]  
  
 
'''Aspiration windows''' are a way to reduce the [[Search Space|search space]] in an alpha-beta search. The technique is to use a guess of the expected value (usually from the last iteration in [[Iterative Deepening|iterative deepening]]), and use a [[Window|window]] around this as the alpha-beta [[Bound|bounds]]. Because the window is narrower, more beta cutoffs are achieved, and the search takes a shorter time. The drawback is that if the true score is outside this window, then a costly re-search must be made. Typical window sizes are 1/2 to 1/4 of a pawn on either side of the guess.  
 
'''Aspiration windows''' are a way to reduce the [[Search Space|search space]] in an alpha-beta search. The technique is to use a guess of the expected value (usually from the last iteration in [[Iterative Deepening|iterative deepening]]), and use a [[Window|window]] around this as the alpha-beta [[Bound|bounds]]. Because the window is narrower, more beta cutoffs are achieved, and the search takes a shorter time. The drawback is that if the true score is outside this window, then a costly re-search must be made. Typical window sizes are 1/2 to 1/4 of a pawn on either side of the guess.  
Line 14: Line 14:
 
instead. However, a fully fledged search is typically full of [[Search Instability|search instability]], and it will often happen that the above re-search will fail low! This is why it is best to only widen the bound that fails, and leave the other bound unchanged.
 
instead. However, a fully fledged search is typically full of [[Search Instability|search instability]], and it will often happen that the above re-search will fail low! This is why it is best to only widen the bound that fails, and leave the other bound unchanged.
  
Modern engines, like [[Robbolito]] or [[Stockfish]], start with a rather small aspiration window, and increase the bound that fails in an exponential fashion.
+
Modern engines, like [[RobboLito]] or [[Stockfish]], start with a rather small aspiration window, and increase the bound that fails in an exponential fashion.
  
 
=PVS and Aspiration=  
 
=PVS and Aspiration=  
Line 59: Line 59:
 
* [http://www.talkchess.com/forum/viewtopic.php?t=46837 Aspiration Windows] by [[Jerry Donald]], [[CCC]], January 11, 2013
 
* [http://www.talkchess.com/forum/viewtopic.php?t=46837 Aspiration Windows] by [[Jerry Donald]], [[CCC]], January 11, 2013
 
* [http://www.talkchess.com/forum/viewtopic.php?t=47996 Aspiration windows] by [[Marco Pampaloni]], [[CCC]], May 14, 2013
 
* [http://www.talkchess.com/forum/viewtopic.php?t=47996 Aspiration windows] by [[Marco Pampaloni]], [[CCC]], May 14, 2013
* [http://www.talkchess.com/forum/viewtopic.php?t=53972 Your experience with PVS + Aspiration window] by [[Fabio Gobbato]], [[CCC]], October 07, 2014 » [[Principal Variation Search]], [[PVS and aspiration]]
+
* [http://www.talkchess.com/forum/viewtopic.php?t=53972 Your experience with PVS + Aspiration window] by [[Fabio Gobbato]], [[CCC]], October 07, 2014 » [[Principal Variation Search]], [[PVS and Aspiration]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=54241 Solving a fail low situation at the root] by [[Alberto Sanjuan]], [[CCC]], November 03, 2014 » [[Fail-Low]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=54241 Solving a fail low situation at the root] by [[Alberto Sanjuan]], [[CCC]], November 03, 2014 » [[Fail-Low]]
 
==2015 ...==
 
==2015 ...==
 
* [http://www.talkchess.com/forum/viewtopic.php?t=57113 Parallel aspiration windows] by [[Giuseppe Cannella]], [[CCC]], July 29, 2015
 
* [http://www.talkchess.com/forum/viewtopic.php?t=57113 Parallel aspiration windows] by [[Giuseppe Cannella]], [[CCC]], July 29, 2015
 
* [http://www.talkchess.com/forum/viewtopic.php?t=58542 Restarting iterative deepening] by [[Harm Geert Muller]], [[CCC]], December 09, 2015 » [[Fail-Low]], [[Iterative Deepening]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=58542 Restarting iterative deepening] by [[Harm Geert Muller]], [[CCC]], December 09, 2015 » [[Fail-Low]], [[Iterative Deepening]]
'''2016'''
 
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60285 Dynamic aspiration window] by [[Sergei Markoff|Sergei S. Markoff]], [[CCC]], May 26, 2016
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60285 Dynamic aspiration window] by [[Sergei Markoff|Sergei S. Markoff]], [[CCC]], May 26, 2016
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60402 Aspiration Windows on the root search -- Determining margin] by [[Andrew Grant]], [[CCC]], June 08, 2016
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60402 Aspiration Windows on the root search -- Determining margin] by [[Andrew Grant]], [[CCC]], June 08, 2016
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60916 Iterative Deepening Question] by [[David Cimbalista]], [[CCC]], July 23, 2016 » [[Iterative Deepening]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=60916 Iterative Deepening Question] by [[David Cimbalista]], [[CCC]], July 23, 2016 » [[Iterative Deepening]]
 
* [http://www.open-chess.org/viewtopic.php?f=5&t=2995 Aspiration window with TT question] by [[Sander Maassen vd Brink|sandermvdb]], [[Computer Chess Forums|OpenChess Forum]], August 01, 2016 » [[Transposition Table]]
 
* [http://www.open-chess.org/viewtopic.php?f=5&t=2995 Aspiration window with TT question] by [[Sander Maassen vd Brink|sandermvdb]], [[Computer Chess Forums|OpenChess Forum]], August 01, 2016 » [[Transposition Table]]
'''2016'''
+
'''2017 ...'''
 
* [http://www.talkchess.com/forum/viewtopic.php?t=63706 Conceptual question on aspiration windows] by Jacob Wilkins, [[CCC]], April 09, 2017
 
* [http://www.talkchess.com/forum/viewtopic.php?t=63706 Conceptual question on aspiration windows] by Jacob Wilkins, [[CCC]], April 09, 2017
 
* [http://www.talkchess.com/forum/viewtopic.php?t=64321 (I)ID and PV dropout] by [[Harm Geert Muller]], [[CCC]], June 17, 2017 » [[Fail-Low]], [[Internal Iterative Deepening]], [[Iterative Deepening]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=64321 (I)ID and PV dropout] by [[Harm Geert Muller]], [[CCC]], June 17, 2017 » [[Fail-Low]], [[Internal Iterative Deepening]], [[Iterative Deepening]]
 
* [http://www.talkchess.com/forum/viewtopic.php?t=65589 Aspiration window problem] by [[Sander Maassen vd Brink]], [[CCC]], October 30, 2017
 
* [http://www.talkchess.com/forum/viewtopic.php?t=65589 Aspiration window problem] by [[Sander Maassen vd Brink]], [[CCC]], October 30, 2017
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=68278&start=16 Re: Lazy SMP ideas] by [[Folkert van Heusden]], [[CCC]], October 03, 2018 » [[Lazy SMP]]
 +
* [http://www.talkchess.com/forum3/viewtopic.php?f=7&t=69627 Impact of Aspiration on tree size] by Frank Kopp, [[CCC]], January 16, 2019
  
 
=External Links=  
 
=External Links=  
Line 84: Line 85:
  
 
'''[[Alpha-Beta|Up one level]]'''
 
'''[[Alpha-Beta|Up one level]]'''
[[Category:Industrial Heritag Trail]]
+
[[Category:Gerhard Hoehme]]
 +
[[Category:Art Museum Bochum]]
 +
[[Category:Industrial Heritage Trail]]

Revision as of 09:41, 18 January 2019

Home * Search * Alpha-Beta * Aspiration Windows

The Non-Euclidean Window [1]

Aspiration windows are a way to reduce the search space in an alpha-beta search. The technique is to use a guess of the expected value (usually from the last iteration in iterative deepening), and use a window around this as the alpha-beta bounds. Because the window is narrower, more beta cutoffs are achieved, and the search takes a shorter time. The drawback is that if the true score is outside this window, then a costly re-search must be made. Typical window sizes are 1/2 to 1/4 of a pawn on either side of the guess.

Gradual Widening

Some programs, such as Crafty, also use a gradual widening on re-searches. For instance, if the window is, in pawns:

[g - 1/4, g + 1/4]

and the search fails high, the next search would be

[g - 1/4, g + 1]

It's important to note that the bound that didn't fail is unchanged. In a basic alpha-beta without search instability, one could have done the next search on

[g + 1/4, g + 1]

instead. However, a fully fledged search is typically full of search instability, and it will often happen that the above re-search will fail low! This is why it is best to only widen the bound that fails, and leave the other bound unchanged.

Modern engines, like RobboLito or Stockfish, start with a rather small aspiration window, and increase the bound that fails in an exponential fashion.

PVS and Aspiration

Using aspiration windows together with the principal variation search (PVS) causes some additional complications.

See also

Publications

Forum Posts

1995 ...

2000 ...

2005 ...

2010 ...

2015 ...

2017 ...

External Links

References

  1. Gerhard Hoehme - The Non-Euclidean Window (1964), mixed technique / wood on canvas, Art Museum, Bochum, North Rhine-Westphalia, Germany, Photo by Gerd Isenberg, October 30, 2016

Up one level