For a blog about Chess and Science, you might expect more articles about the intersection between the two. Indeed, I have occasionally written here about research articles that have involved chess (such as studying the physiology of players during a game, the Einstellung effect), even including my own efforts in exploring mutual information in chess. Usually, chess is used as a vehicle in science to study physiology, memory, or decision making in general. However, I think it is worth exploring a truly scientific approach to Chess.
In its essence, Science is about gaining useful knowledge in a systematic way to solve problems. Chess players engage in a similar activity all the time, if even subconsciously, studying games and reading literature to build a model of the game in their mind that can be applied to making decisions at the board. Likewise, scientific knowledge about the natural world has informed incredible advances in technology for a wide range of industries.
Below I suggest different ways in which the scientific method can be applied to different aspects of chess. I encourage any interested readers to take up the challenge of performing chess research, following the principles featured in the rest of the article. I intend on making this blog a vehicle for such research, and welcome submissions of original research to this blog (or at least posts that link to your analysis). Perhaps Science on the Squares can become the first real scientific journal of chessology (or chessonomics?)
Do you think that such a method will prove useful and yield insights? Is it too slow and laborious? Have you made any discoveries in Chessology? Please share your thoughts and comments below, or contact me directly.
Science as a Methodology
Steintz and his followers such as Tarrasch are often crediting with establishing the Modern or Scientific school of chess. In fact, it is difficult to find evidence that Steintz considered his approach scientific, or that it even used the scientific method properly or at all. The very idea of a scientific school of chess usually evokes thoughts of a systematic, principled and overly formulaic description of strategy. Indeed, Tarrasch was often criticized by the hyper-modernists as being too dogmatic about his beloved positional principles. However, I wish to make a clear distinction between a scientific style and an application of the scientific method to the study of the game.
Applying the scientific method to the game of chess can take various forms, but in all instances, should take pre-existing knowledge that encapsulates your working model of chess and use it to generate testable hypothesis. Here the goal would most often be to increase your knowledge or understanding of some aspect of chess, but not to generate dogmatic rules to stick by and not to try to prove some theorem on the game. Rather, it is a systematic way to increase and refine your understanding, which can help inform how you solve problems at the board.
Taking this approach in any walk of life (natural phenomenon and chess alike) through experimentation and testing of the hypothesis you may find that the data confirms your prediction. In this case, you should search for another hypothesis to test your model of chess, as confirming or supporting an idea you already have doesn't add as much value as refuting a misconception. Data that doesn't match your hypothesis should lead to one of several different actions:
An example of the scientific approach
(Click to enlarge the image)
1. You should review your experimental procedures. Did you actually test your hypothesis, and did so objectively?
2. You should reexaime your hypothesis. Did you formulate a hypothesis that had the appropriate scope (or was it too broad, too narrow), to really test your pre-existing ideas about chess.
3. If the above two points are satisfied (and this may take several experiments) but your hypotheses are still being refuted by the data, then you must examine the underlying ideas about the game you used to rationalize your hypothesis. This is the most important phase in the scientific method, when you change your actual thinking, rejecting old ideas that did not hold up to examination and experimentation.
There might be many ways to apply this method to chess. In all cases, your ideas about chess should be used to generate the hypothesis. Here are some examples that I can think of (what other ways and terms can you think of?).
Drawing from the field of bioinformatics, this is the way in which I would describe the study of chess in which statistics and opening 'theory' is analyzed to generate conclusions. Many chess players already do this, although most probably do not formulate an explicit hypothesis first. Experiments following this method might make predictions about the winning chances offered by a certain opening move, or which move is most popular. More creative questions could involve looking at which squares are most often used*, or to determine the value of material versus clock position in blitz games. I would personally like to see the later study repeated at longer time controls, and to quantify time and material imbalances in terms of differences in rating point performance. In other words, to not only equate seconds to pawns, but both to an quantifiable measure of increased winning chances. This type of study might yield important insights, although it is limited to pre-existing games and positions (numerous that they are) and their outcomes.
A more commonly used method of analysis, the evaluation of a position by a computer engine, is another tool that might be used to craft chess experiments. Instead of generating a hypothesis that predicts the frequency of a move in the database, questions could be asked about specific moves, plans, and predictions about the engine's evaluation can be made. Many chess players already do something to this effect when analyzing a game, or an opening line; use the computer to check their analysis and evaluate their ideas. I propose that this should be done in a more systematic way, especially with regards to formulating a prediction, and using a systematic and defined procedure. This will make it easier to share and compare results, akin to how scientific studies are communicated (ideally). I also suggest that, when performing analysis on a certain idea, such as testing the concept 'Knights on the Rim are Dim", experiments should be setup in a way that isolates a particular variable. Instead of simply cherry-picking known positions that have a Knight on the Rim (or doing a statistical analysis, which would be appropriate in the preceding section), the same position with different Knight placements should be analyzed. Here, the emphasis is not so much on determining what
Psychological and Physiological Analysis
The scientific approach to chess study has already been implemented when studying the chessplayers themselves. Ideas about pyschology and decision making have been tested (one example is the Einstellung effect in Chess), as well as analysis of the physiological state of the player. Since tournament chess is a game played against other people, such studies have real value to understand how your opponent is arriving at their decisions and even potentially how to spot tells that give you insight on what they are thinking. (in general). Extracting lessons that can be employed at the chessboard from these studies, however, is a difficult thing to do; despite writing about the subject, I often only think about the Einstellung effect when analyzing, not actually when playing (when it could prove useful and inform my decisions at the board.) This is a personal limitation, however, and does not negate the value such studies have.
Others take on Scientific Chess
This article would not be complete without mentioning others that have written about a scientific approach to chess, in one form or another. This is not a complete bibliography, and also isn't a strict description of my references (some of these writings don't directly address my argument), but rather are provided because they may prove to be, if nothing else, interesting reading.
A blog post at the wonderful Kenilworthian blog, authored by Michael Goeller. This
Science Secret of Grandmasters Revealed
A nature article that summarizes research showing how strong chess players have an objectivity and tendency to falsify their own hypothesis rivaling that of scientists. (Good news for myself, I suppose).
The First Scientific Theory of Chess
An article in two parts (See Part I or Part II) that talks about formulation of 'theories' in chess, and even attempts to provide statistical evidence to prove the popular and long-held belief that the game is drawn with perfect play. You can also find a response to this article, which has a long and complicated argument against scientific chess but an argument which I believe is ultimately flawed and incorrect. In a future post, I may offer a more developed rebuttal, but for now leave the reader to judge for themselves.
Training in Chess: A scientific approach
An excellent article by Fernand Gobet (whose research has been featured here before), which is also reviewed by the pawn to rook four blog. This is another article which also deserves more attention and another few blogs entries, but for now I can summarize this work by saying that is contains insights on how chess players learn the game and develop their skill (rather than insights on chess itself), insights which have been gleaned from experimental evidence.
*A note on this article: the author suggests that most players used d4 openings more, since that square is more frequently visited in the data set for those players. I would argue that this actually shows a greater amount of players using 1.e4, for it is in those openings that a lot of traffic at d4 can be seen. Consider the first few moves of the Open Sicilian: 1.e4 c5 2.Nc6 Nf3 3.d4 cxd4 4.Nxd4; despite this being an 1.e4 opening, there has been only 1 move to that square compared to two moves to d4, both the pawn and the Knight. Further exchanges are likely to increase that number further.