Copyright (c) 2013 John L. Jerz

The Art of Planning in Chess by Neil McDonald
Home
A Proposed Heuristic for a Computer Chess Program (John L. Jerz)
Problem Solving and the Gathering of Diagnostic Information (John L. Jerz)
A Concept of Strategy (John L. Jerz)
Books/Articles I am Reading
Quotes from References of Interest
Satire/ Play
Viva La Vida
Quotes on Thinking
Quotes on Planning
Quotes on Strategy
Quotes Concerning Problem Solving
Computer Chess
Chess Analysis
Early Computers/ New Computers
Problem Solving/ Creativity
Game Theory
Favorite Links
About Me
Additional Notes
The Case for Using Probabilistic Knowledge in a Computer Chess Program (John L. Jerz)
Resilience in Man and Machine

The Art of Planning in Chess by Neil McDonald

Neil McDonald quotes Lasker's claim that you can improve your game of chess by replaying the games of the masters. You can likely improve even more by replaying master-level games that are commented by chess teachers.
 
p.5"The most intelligent inspection of any number of fine paintings will not make the observer a painter... The intelligent perusal of fine games cannot fail to make the reader a better player and a better judge of the play of others. - Emanuel Lasker"
 
The concept of planning in a game of chess (for McDonald) involves getting your pieces to work together as a group.
 
p.6"Planning is about getting your pieces working together in a group, so that their overall strength is greater than the sum of the parts."
 
Lasker has taught not to begin an attack until the 'energy' of your pieces is superior to the 'resistance' capacity of your opponent. What if we came up with a way to estimate the energy of our pieces, as well as the resistance of our opponent? What if we used this concept in our evaluation function of our computer chess program?
 
p.18"World Champion Emmanuel Lasker warned us at the end of the 19th century not to begin an attack until we have a tangible advantage. It is a question of how much energy is attacking how much resistance."
 
Creating outposts for your pieces in your enemy's camp should be the first object of your attack, according to Lasker.
 
p.21"Let it be the first object of your attack to create strong points as near your opponent's camp as possible, and occupy them with pieces which have from there a large field of action - Emanuel Lasker."
 
Here we have McDonald's definition of a space advantage.
 
p.25"A space advantage means more potential scope for the pieces, and therefore a more promising middlegame."
 
Knights belong in the center because this piece is most effective at short range.
 
p.26"As a rule, a knight desires a stable post in the centre, as it is a short range piece"
 
Here we examine a game where McDonald claims that white's doubled, isolated pawns are not a liability. A traditional chess program might penalize white for the doubled pawns. McDonald argues that the central location of the pawns, the fact that they are not easily attacked, and their ability to deny mobility to black's pieces, gives white compensation.
 
p.34"A quick word about pawn structure. The white pawns on d4 and d5 [see diagram below] are doubled and isolated - according to positional lore, they are therefore terribly weak. Not so: the pawn on d5 provides vital support for White's next move [25.Qe6] , and, as we have seen so far, has hindered the black knight [on f7, formerly on d8] by keeping it out of the c6 and e6 squares. It has also restrained the black pawn on c7, keeping it a backward pawn. The pawn on d4 also has an important role in stopping Black from playing Ne5, blocking the e-file... The pawns aren't weak, statically speaking, as they are difficult to attack and because White has the initiative. Dynamically speaking, they are strong as they dominate key centre squares."

GK24nf7.jpg
Grischuck-Kamsky, 2005, after 24...Nf7

Richard Reti defines the foundation of positional play as a knowledge of combination. What if we put reasonably accurate estimates of a piece's ability to participate in combinations in the evaluation function? We might get a chess program that plays a better positional game of chess.
 
p.46"A knowledge of combination is the foundation of positional play.
            Richard Reti"
 
Tactics eventually come into play in a game of positional chess - just not right now. The pieces function like a coiled spring - energy is stored for release at a later time.
 
p.46"You will notice that in every game [presented in chapter 2], a tactical oversight of various degrees of seriousness allowed the victor to achieve his aim. Although this is primarily a book about strategy... watch out for tactics!"
 
Having 'a plan' is a good idea, but continuously modifying your plan to fit the situation on the board is an even better idea.
 
p.63"A plan needs to be continuously moderated or even changed completely according to circumstances. These days at least, you hardly ever see a game in which a player kept to the one and same plan throughout the middle-game. Even if the basic intention behind his schemes was the same... a multitude of frequently changing mini-plans was being employed."
 
McDonald tells the reader to look for a weakness in the enemy's pawn structure. If there is none, look for a way to create one. Failing that, try to take advantage of a pawn majority, or even a pawn minority.
 
p.77"Let's assume our opponent is sensible enough to develop his pieces to decent squares, fight for the centre with his pawns, and avoid doing anything reckless with his king. We will therefore need to soften him up gradually if we are going to get the win we deserve.
   This can best be done by targeting one or more weak points in his pawn structure. In what follows we shall look at various techniques for forcing and exploiting such weaknesses - or even persuading the opponent to create them for us!"
 
Each position is unique and it is the features of the position that determine who is better. It seems that you have to do more than consider the location of the pieces on the board in order to accurately evaluate the winning chances of a position in a game of chess.
 
p.86"Every position has to be evaluated according to its particular features."
 
Mobility that leads to pressure is a good thing - a piece being tied down to the defense of a weak piece is a bad thing.
 
p.88"Here we see the superiority of a bishop over a knight in an endgame where there are pawns that can be targeted on both sides of the board... whereas the black knight has been deprived of all its mobility by the need to defend b7."
 
Don't let your opponent play good moves - this also means not letting him play good positional moves. If you cannot recognize good positional moves for yourself, then you cannot recognize (and therefore prevent)  good positional moves by your opponent.
 
p.155"An essential component of any plan is preventing the opponent from playing the moves he wants to play."
 
Adjacent pawns have the potential to support each other by advancing. We would need to determine the pawn mobility and resolve any limiting factors in order to see if this is possible. 
 
p.163"In general, pawns are strongest when they are adjacent to each other in the centre, as on e4 and f4. The question of whether to advance one of them to the 5th rank is always of the greatest importance."
 
McDonald would judge pieces by what they can do, rather than where they sit on the board.
 
p.176"This retreat reminds us that the value of a piece is to be judged by how well it contributes to the health of the whole army, not its own individual worth."
 
Our evaluation function should reward the queen+knight combination (as opposed to the queen+bishop combination) for the ability to pressure a large number of squares.
 
p.177"The queen and knight are well known to be a formidable attacking force, as their different powers increases the chance of finding a key to unlock a defense."
 
Outposts for pieces are generally located on squares that cannot be threatened by pawns.
 
p.185"A so-called 'hole' is a weak point of this kind. It is a square in a player's pawn structure that can no longer be protected by one of his pawns. If the hole is on a centre square or another point of strategic interest, it can make a fine outpost for an enemy knight."
 
McDonald frowns on counting mobility squares by themselves - we need to look at the interaction(s) of our pieces with the other pieces. The following quote should be read by anyone who includes the concept of mobility in the evaluation function of a computer chess program.
 
p.189"As so often, we need to remind ourselves that it isn't the fortune of the individual piece that matters - it is how it functions with the other pieces. For this reason it is often completely meaningless to count the number of squares, in the centre or otherwise, that a piece controls or attacks."
 
The hardware-based chess computer Hydra defeated British Grandmaster (and the world's 7th best player) Michael Adams in a recent tournament. The traditional evaluation function that guided its choice of moves apparently had no understanding of what it was doing.
 
p.229"It is incredible that the Hydra computer can crush Michael Adams 5 1/2 - 1/2 in a match [28 June, 2005], and yet not have any understanding."
 
From the hyrdachess.com web page:
 
It runs on a 64CL and evaluates about 200,000,000 chess positions per second, roughly the same as the much older Deep Blue, but with several times more overall computing power.  Whilst FPGAs generally have a lower performance level than ASIC chips, Moore’s Law allows modern-day FPGAs to run about as fast as the older ASICs used for Deep Blue.  The engine is on average able to evaluate up to a depth of about 18 ply (9 moves by each player), deeper than Deep Blue, which only evaluated to about 12 ply on average.  HYDRA’s search uses alpha-beta pruning as well as null-move heuristics.  The extra search depth over Deep Blue is due to its use of more modern type B forward pruning techniques that are slightly less perfect but generally play better due to the greater search depth these techniques permit.

Under the stewardship of Dr. Chrilly Donninger, HYDRA R&D is currently developing a 64CL new version.

Enter supporting content here