Nordic Go Dojo

I wrote this bachelor’s thesis originally in Finnish in October 2013 and posted its English version on gooften.net on November 1, 2013. Its language has been slightly modified and improved for this re-posting on nordicgodojo.eu.

Abstract

Expertise is a topical subject in our ever-changing society. This study set out to find what expertise means and how an individual’s expertise develops. As the study of expertise comprises a large field of study, literary sources were based especially in The Cambridge Handbook of Expertise and Expert Performance.

In literature, one is often faced with the estimation that one needs ten years of practice in order to become an expert at something. I take it as a research question to find out if it is possible to become an expert in a shorter time with the right means and structure the thesis around this question. I discuss the subject by approaching it from the general directions that expertise has previously been studied from, and by analysing the requirements that expertise sets on the human memory.

This thesis proposes that, based on its study material, expertise is developed through deliberate practice. The requirement of ten years of practice is a fairly accurate generalisation, but not a requirement in itself. The research material also leads one to conclude that the notion of ‘talent’ that comes up in everyday language doesn’t imply that some individuals are inherently more suited to become experts in given fields due to their cognitive qualities.

Table of contents

Chapter 1: Introduction
1.1 On expertise in general
1.1.1 Definition of expertise
1.1.2 Historical aspects in expertise
1.1.3 The future of expertise
1.2 The motive for this study
1.3 The ‘ten-year rule’
1.4 The structure of this thesis
3.1 ‘Individual differences in cognitive capacity’
3.2 ‘The development of constantly-utilised skills into expertise’
3.3 ‘Top-level performance facilitated by favourable learning circumstances’
3.4 ‘Reliably repeatable first-class performance within the domain of expertise’
3.5 Summary on the approaches to research on expertise
4.1 The importance of memory on expertise
4.2 Expertise and short-term memory
4.3 Expertise and long-term memory
4.4 The expert’s working memory
4.5 On the size limit of memory chunks
5.1 Measuring expertise in chess and go
5.2 On the chess and go skills required for expertise
5.3 On reaching the skills required for expertise in chess and go
6.1 The recipe for expertise
6.2 The future of expertise

Chapter 1: Introduction

1.1 On expertise in general

1.1.1 Definition of expertise

Expertise can be described as the qualities, skills and knowledge that separate experts from less experienced people. Depending on the field of study, there may be objective criteria that experts can be identified with: master chess players usually defeat regular amateur players, and professional musicians can play musical pieces in a more skilful, careful, and elaborate way than less skilled musicians. [13, p. 3].

In some fields of expertise, it may be difficult to distinguish experts from laypeople. Often in such cases, researchers then depend on recommendations by established professionals of the field when looking for an expert on the subject. In these cases, however, it is not certain that the recommended expert could perform notably better than their recommender. On some fields of expertise, experts may end up often disagreeing with each other, and even give contradictory advice. [13, p. 4]. Although it is fairly easy to define expertise, it seems to involve a social aspect, the importance of which depends on how easy or difficult it is to measure the expertise in a field quantitatively.

1.1.2 Historical aspects of expertise

Expertise is an important and pertinent topic in our modern society. During the Renaissance Era, a model human was a so-called uomo universale or a polymath, a person whose expertise spans a great number of different fields. Later, however, it was realised that even the most gifted and intelligent people alive never had the time to learn everything that they wanted to. Since then, the general trend has turned towards mastering just one or a few subjects. In a sense, this development during the last few centuries mirrors what has happened in different industries: while a traditional blacksmith might have crafted a single item from start to finish by himself, the same item would nowadays be created from modules in factories. The new procedure does not require one highly skilled individual who can perform every single process during the creation of an item, but instead relies on several different workers who know how to do just a few processes each in the whole procedure. In this way, expertise in domains of more limited scope is getting more important for the general efficiency of the modern society.

In the western civilisation, there has throughout the history been a general interest in the unparalleled knowledge and skills that experts acquire as a result of their training. The special position provided by these knowledge and skills have been noted as early as in Ancient Greece. Later, in the Middle Ages, artisans founded different guilds to defend themselves from competition, creating for themselves a monopoly for their products and services with certain quality standards. [13, p. 4]. The master-apprentice system that followed from this may be viewed as a precursor of the way with which people today strive for expertise; today, the experts of tomorrow study under current experts in different schools and universities.

1.1.3 The future of expertise

The existence of experts has a clear financial value to the society, which makes the upholding and increasing of their number important. Up to now, the development in schooling in civilised countries has related to providing free education for the willing. In the longer run, it also seems necessary to study what kind of education and teaching is the most effective, and which settings enable the development and mediation of expertise as effectively as possible. At the same time, the notion of the possibility of being able to become an expert at something depending on one’s internal characteristics and qualities is quickly getting outdated. Many studies have repeatedly proved that the most important factor in the creation of expertise is long periods of deliberate practice and the resulting experience [24, p. 11].

1.2 The motive for this study

Originally, the subject of this thesis was outlined as a general mapping of expertise as a phenomenon, using my personal skills and interest in the board game of go as an approach point. As the study of expertise is an enormous field of research, the sources for this study were decided to be based in The Cambridge Handbook of Expertise and Expert Performance. As I conducted my research, I found that a more specific research question would be useful; I took it upon myself, then, to endeavour to find out how it would be possible to most effectively ‘build’ a new master of go.

The reader may be interested in this study for two different reasons. If the reader is a player or researcher of skill-based games, this study may well give them new points of view. For players, the study may even give concrete advice on how one could develop their skills further, and on how one could conduct their training more effectively. For non-players, too, the study may have some merit, as it will describe through examples on a general level how one should strive for expertise.

1.3 The ‘ten-year rule’

When studying expertise, one is often faced with the so-called ten-year rule. It is an oft-repeated estimation that if one wants to gain the knowledge and skills required to become an expert at something, it will take them at least ten years to do so [24, p. 12]. During my research, I strived to find out if one could reach the level of expertise faster than this, and if so, with what means. As it happens, exceptions to the ten-year rule exist in large numbers.

In the context of go, players can be divided into amateurs, who are the usual everyday players, and professionals, who are players certified by the Japanese, Chinese, Korean, Taiwanese or the us Go Association. A professional player, in a sense, has a secure job position at their national go association. As very few amateur players reach the playing level of professionals, it seems justified to call professional go players experts of the game. Of the current Japanese professional players, for example Fujisawa Rina passed the professional exam at the age of 11 years and six months, which means that at the time she became an ‘expert’, she had probably only had six years of time to study the game seriously [20].

1.4 The structure of this thesis

The purpose of this study is to find what expertise means and how it is born, primarily basing the scientific sources in The Cambridge Handbook of Expertise and Expert Performance. These results are further applied in finding out what makes somebody a master of chess or the Chinese board game of go. In the light of the results, I will contemplate whether it is possible to create a general guideline on how one could become an expert at something, and if it is possible to do so in less than ten years’ of time.

In Chapter 2, I provide a short summary on the nature of chess and go. In Chapter 3, I present and review those approaches to the research on expertise which I found important to chess and go. In Chapter 4, I venture to find out what expertise requires from the human memory and what happens in the human brain during the development of expertise. In Chapter 5, I sketch out what an expert on chess or go should know and be able to do, and in light of chapters 3 and 4, assess the quickest path to expertise in said games. Finally, in Chapter 6, I give a summary on my findings in the expertise on chess or go on a more general level.

Chapter 2: Chess and go

Chess is nowadays a well-known game throughout the world, with a great number of players in the western countries. Generally, as well as in literature, chess is often profiled as a game played by especially smart people. A single game of chess can be imagined to be an abstraction of a medieval battle between two armies.

Compared to chess, go is at the time of writing a relatively unknown game in the western countries, although its visibility has gradually increased by the 21st century [11]. Similar to chess, go is a skill-based board game with no luck element. Also, similar to chess, the rules of go are quick to learn, but the process of becoming a master takes years and years of time. If a game of chess represents a battle between two armies, a game of go can be thought of as a full-scale war between two forces, consisting of numerous smaller and larger battles in different parts of the game board. What is more, in go, it is possible to win the war even if every battle is lost. In the context of this study, I presume that chess and go are very similar games in terms of expertise, even though chess skills don’t directly correlate with go skills [21].

More recently, go has received additional news coverage due to the fact that stronger go players still invariably defeat go-playing ai (at the time of writing in 2013). In chess, this has not been the case for well over ten years. Burmeister [5, pp. iii–iv] evaluates that the brute-force calculating methods used by chess ai do not work in go because a go player not only has to estimate the territories staked out by the players, but also has to see which stones on the board are dead and which are alive. Additionally, the larger game board and the resulting vaster number of different game trees greatly slows down the ai’s calculations. According to Burmeister, the game of go provides for a good field for ai research in planning, problem solving, pattern recognition, and predicting the opponent’s actions. Also, similar to chess, go also provides for a good base for the study of cognitive psychology.

Basic rules of go and general strategy is presented in appendix a.

Chapter 3: Pertinent approaches to research on expertise in chess and go

3.1 ‘Individual differences in cognitive capacity’

Galton [16] noticed that on many domains of expertise in Great Britain, a significant portion of experts came from a small number of different families, far more often than would be likely by pure chance. From this observation, Galton proposed that inherent characteristics of individuals affect the development of knowledge and skills. Later in the 20th century, however, it was proved my psychometric studies that there were no conclusive proof to Galton’s hypothesis, for the skills of experts were limited to the experts’ own domains. In 1996 Ericsson and Lehman noted that:

1. measures of cognitive capacity cannot be implemented to predict if an individual will become an expert in some domain (1),
2. experts’ knowledge and skills are limited to their own domain, with surprisingly few benefits outside of said domain, and
3. that systematic differences between experts and less skilled individuals practically always relate to characteristics that the experts develop during their prolonged training. [13, p. 10].

In everyday language, there still exists a notion that some individuals are simply more gifted than others on certain domains, and that it is more or less predetermined on which domain an individual could become an expert. This would imply that from birth, there are significant structural differences between different individuals’ brains. This subject has intrigued researchers for a long time, and especially in recent history, different methods for measuring an individual’s intelligence have been developed. Cattell’s [8] g-factor is one of the better-known of these. In his study, Carroll [7] divided the concept of intelligence further into so-called fluid intelligence, Gf, and crystallised intelligence, Gc. Fluid intelligence stands for an individual’s general power of reasoning and crystallised intelligence for knowledge that an individual has internalised for use in problem-solving. It is worth noting that crystallised intelligence is strongly culture-bound.

Laboratory experiments have shown that while a fluidly intelligent person may at first learn new things more quickly, fluid intelligence in itself is not a reliable indicator of the level of skill that an individual will reach [13, p. 32]. In other words, the biology of an individual’s brains has a limited importance in the development of expertise. It is more important, for example, how interested the individual is in their domain, or how much time the individual ends up using practising.

3.2 ‘The development of constantly-utilised skills into expertise’

Ericsson found [13, p. 11] that already some 19th century studies proposed that reaching an especially high level of skill is a natural consequence of broader experience in a domain. Ericsson refers to a research by Bryan and Harter in 1899, which argued that ten years’ worth of experience is necessary for an individual to become an expert telegrapher. This observation is consistent with what I wrote in Chapter 1.3, but from the context of Ericsson’s text, it looks like Bryan and Harter did not take into account the quality of practice; that, by being ‘exposed’ to a ten years’ worth of experience, everybody could become an expert. Later studies have consistently showed that when an individual reaches an acceptable level of skill for themselves, additional experience will then yield marginal utility [13, p. 691]. It is for example easy to find amateur go players who have played tens of thousands of games but still play at an intermediate level – and on the other hand, there also exist fast-improving players, who may be called skilful after they have only played a thousand games. It is clear that mere time expended in a domain is a bad measure of skill, as the utility of time is not always the same.

Historically, the general performance level of humans has been on a steady rise. Roger Bacon argued in the 13th century, that by using the studying methods of that time, it would have been impossible to become an expert in mathematics in less than 30 years of time. Today, pretty much the same knowledge is taught to students by the end of high school. The reason to this enormous difference in the time required to learn is the fact that the study material is nowadays easier to access and better structured for learning. Even in domains that have seen fewer changes in training methods, such as playing the piano, experts of today are able to do things previously thought impossible: expert pianists of today are able to play pieces that were thought impossible to even the best musicians of the 19th century. In the field of running, it was previously thought impossible to run a mile in less than four minutes. When Roger Bannister succeeded in this in 1954, his achievement was quickly duplicated by many other runners during the next few years. This quick change was attributable to changes in practice methods. [13, p. 690].

In the Berlin music academy, weekly tabs on time usage of violin practitioners have been kept. One study remarked that, although practitioners spend roughly equal time for practice weekly, differences in skill were created. It was noticed that the best violinists used more time weekly for activities that were deliberately engineered to better an individual’s playing of the violin. This kind of activity is nowadays referred to as deliberate practice. [13, p. 691]. Training methods are often engineered by the practitioner’s mentor, who aims to improve certain aspects in the practitioner’s performance.

In Japanese go schools, as well, deliberate practice is used. Common training consists of playing training games, their post-analysis, reading theory books, rote learning of played professional games, and solving go problems. Of these, the two last are not very obvious methods of training, but they are given a considerably big weight. Rote learning of played professional games is thought to develop a player’s intuition: it increases the number of familiar shapes and sequences contained in a player’s memory, the shapes and sequences then becoming available to call out from memory when needed. Therefore, rote learning is effective at improving a player’s global intuition of the go board, while go problems – meaning game positions in which a player is to make a group of go stones to live or to die in a restricted space – improve a player’s intuition for local situations. The teacher of a go school often fine-tunes the learning material, depending on the student’s phase of learning.

3.3 ‘Top-level performance facilitated by favourable learning circumstances’

In the second half of the 20th century, expertise was studied as top-level performance facilitated by favourable learning circumstances. Bloom [3] and his colleagues interviewed 120 internationally successful people on their parents, teachers, and trainers. At the start of the project, the researchers presumed that the researched individuals would ‘have been identified as extremely gifted from the start, and that they would have been given special teaching and care’ [3, p. 520]. Later, however, Bloom told an interviewer: ‘We were looking for exceptional children, but found exceptional circumstances’ [6]. Sosniak [26] listed the main different types of favourable learning circumstances as the chance to learn, practical exercises, and an encouraging social background.

Optimal circumstances often support deliberate practice, which according to Chapter 3.2 promotes the development of expertise. It is also possible that this positive effect of favourable learning circumstances on expertise played a role in the previously-presented Galton’s observation about a small number of Great Britain’s families providing for a large number of experts. Even if said families did not have a significant genetic advantage compared to other families, it is possible that the interests of parents were passed on to their children. On the other hand, well-educated or rich parents will also more likely end up supporting their offspring at their interests, both spiritually and financially.

3.4 ‘Reliably repeatable first-class performance within the domain of expertise’

It is difficult to study scientifically which factors led to success in a single record performance or success story. Ericsson and Smith [15] proposed, that in order to study expertise with laboratory-level precision, it is necessary to find an exercise that captures the essence of a domain of expertise. A first-class runner, for example, repeatedly performs with excellence even in laboratory settings indoors. De Groot [23], on the other hand, noticed that the skill at finding the best move in a given chess board position is the best correlate for a player’s chess rating and success in chess tournaments. As per my assumption in Chapter 2 about chess and go being games of similar nature, De Groot’s observation should apply for go as well.

The possibility of doing laboratory research is useful for research on expertise, as it also enables the evaluating of the level of performance of less-skilful individuals. Thus, it becomes possible to follow an individual’s progress, and by changing different controlled variables during the research, it is further possible to study which practice methods are effective. Some domains of expertise, such as chess and go, are well fit for laboratory research, while for example soccer, due to its large space requirements, is clearly less well fit. Recalling back to Chapter 1.1, it was by laboratory researches such as these that it was found that fame-based experts do not always perform better than their recommenders [13, p. 13].

3.5 Summary on the approaches to research on experise

The approaches presented above make up but a fraction of all the research that has been conducted on expertise up to date. As we can see from the presented approaches, expertise research is already a large field that can be approached from many, largely varying points of view. The four approaches presented above provide for what I saw as important points of view for the research on expertise in chess and go, and research based on these approaches will be further applied in Chapter 5 of this thesis.

Chapter 4: Expertise and memory

4.1 The importance of memory on expertise

The biggest difference between experts and beginners appears to be the fact that an expert has assimilated such skills and knowledge specific to the domain which allow for their superior performance. Thus, the substance of expertise seems to be tied to human memory and the biology of human brains. The content presented in Chapter 3 could already be implemented to give recommendations on how to effectively become an expert in a given domain, but in order to provide for a more comprehensive understanding on the subject, I felt it important to also study the changes that the development of expertise creates in the human brain.

4.2 Expertise and short-term memory

One of the best-known phenomena in the context of expertise is that as the amount of practice and skills accumulate, an individual starts to cognitively organise available information in larger units [17]. This phenomenon has been widely documented, and it was first found in the study of chess [10]. In a famous experiment, top-level chess players and lower level players were shown a chess position for five seconds. After the five seconds, the player was to reconstruct the game position just seen. The top players were able to remember the place of almost every piece, while the less-skilled players could only reconstruct about five pieces. Even a person who is unable to play chess could quite easily memorise the position of five chess pieces for a short time.

The original, now fairly common explanation for the top players’ better performance in the aforementioned experiment has to do with a phenomenon referred to as chunking. With experience, a top player essentially builds a large database of differently-sized patterns of game pieces in their memory—these are called memory chunks. A top player will analyse a board situation by utilising these chunks, evaluating how they could interact, while a beginner player is only able to take into account several pieces at a time. When looking at a game situation, a top player will immediately notice many familiar units of game pieces. When reconstructing a game position, a beginner might recall the locations of six or seven different game pieces, which is nearing the limit of human short-term memory, while a top player may recall six to seven memory chunks, which could easily translate to 23–25 pieces. [24, p. 9]. In chess and go, chunking skills are based on the fact that both games include some fairly standard game piece patterns and sequences, found optimal for both sides, which often see play. When one gathers experience at a game, the amount of learned game patterns and sequences likewise increases. For example, the first 30 moves of a 250-move game of go could well be played according to a relatively famous opening pattern, judged even by professional players.

The existence of memory chunks could be proved by another, simple test. This time, top-level and lower-level chess players were shown a chess position for five seconds, but with randomly placed chess pieces. Now, the top-level players were not able to effectively utilise their memory chunks, and could only recall the locations of a number of pieces similar to the lower level players. [13, p. 50].

These phenomena, first found in studies of chess skills, were replicated many times over with different tests, both with chess and in different domains such as bridge, go, and electronics. In most studies, it was exactly the size of memory chunks that became the key difference between beginners and experts. Both beginners and experts have similar limits in their short-term memory usage, but for example with beginner chess players this translates to only about five chess pieces, while with an expert player this could translate to up to 25 pieces. [13, p. 50].

More effective usage of short-term memory alone is not enough to make for a master player; it has been estimated that an expert of chess or go should know some 10,000–100,000 different game patterns [22, p. 5]. Compared to this, five memory chunks stored in short-term memory has a marginal weight. It is probable that a skilful player’s short-term memory chunking skills are developed at the same time when they build a larger database of different game patterns in their long-term memory.

4.3 Expertise and long-term memory

‘Knowledge’ was defined as the key component in chess expertise as early as in the late 19th century. Since then, many studies have been conducted on how information stored in a player’s memory turns into playing skills. Points of interest have been, for example, the memorisation of static positions, the memorisation of moves and move sequences, and the general structure and content of long-term memory. [13, p. 526].

In the context of expertise in chess and go, it is especially interesting that master players choose their moves from a very small number of options. ai players have to evaluate hundreds or thousands of options, but for a human master player, the evaluation of a dozen or so promising-looking options is most often enough. This is a consequence of the game patterns and move sequences that a master player has memorised. The number of alternatives that a master player goes through was studied by following the eye movements of master players. [13, pp. 526–527]. Due to their extensive experience, a master player is able to compare seen game positions into previously-seen, already familiar positions, and can then pick a line of play that was previously found to be good, saving time in the process. Thanks to this quality, master players can for example play many games simultaneously, using little to no time for thinking their moves, and still perform with excellence. Compared to lower level players, master players are also able to analyse the game board in larger pieces, thanks to their more developed chunking skills and larger database of memory chunks.

When an individual develops their expertise, their knowledge and skills in the domain of expertise naturally grow, but they are also organised more effectively in relation to each other. The information system that results gives an expert a strong basis for choosing, organising, presenting, manipulating, and interpreting information. The more effective short-term memory that has developed alongside the development of expertise further supports the expert’s information-based reasoning. The combination of these skills gives an expert effective tools for understanding the deeper structure of a problem situation. Like this, an expert is more able to understand the substantial interactions that comprise a problem situation than laypeople, and can come up with many effective solution formulae to the problem. [13, pp. 598–599].

As was described in Chapter 3.1, an expert uses both fluid and crystallised intelligence when thinking of a solution to a problem. A beginner is only able to use their fluid intelligence because they lack any fundamental experience from the domain. As such, a beginner can only think of solutions to such aspects of a problem situation that are instantly seen. In chess and go, a beginner might be able to solve a very local situation correctly, but in terms of a larger picture it might still be the wrong solution. Further, through the combination of fluid and crystallised intelligence, an expert is able to contemplate and analyse promising-looking directions of development even in problem situations new to the expert. [13, p. 599].

4.4 The expert’s working memory

There seems to be a kind of memory related to expertise that cannot be directly compared to the usual short and long-term memories. Many researchers have conducted experiments on blind chess with an arrangement that the master player never gets to see the game board, and therefore has to base their playing on memory alone. As the master player may have to calculate through sequences 30 or 40 moves deep, it would seem like their brain is exceeding the theoretical maximum capacity of the human short-term memory. Ericsson and Kintsch [14] concluded from their research that the type of memory that these experts used in the experiment has to differ from general short-term memory, and proposed to call it the ‘long-term working memory’ (distinguishing it from long-term memory and short-term memory, the latter of which is sometimes called working memory) or the ‘expert’s working memory’. Based on research, this expert’s working memory is located in the same part of the brain that also processes the short-term memory. [13, pp. 599–600]. An interesting part of the expert’s working memory is that it is not very prone to disturbances; for example an experienced driver of a car is able to give first-class performance even when conversing with their passengers.

4.5 On the size limit of memory chunks

Since expertise is based on the effective storage and organisation of information by memory chunks in the human brain, I found it interesting to inspect the neurobiology of memory chunks a little bit further. I came up with the following questions:

1. Is there a theoretical maximum to the size of memory chunks?
2. Does the maximum size of memory chunks depend on the individual? and
3. If the maximum size mentioned in question 1 does exist, what is the size limit based on?

One hypothesis relating to short-term memory is that it could only contain a specific, predetermined sum of concrete things. Brener [4] refuted this hypothesis in an experiment, in which it was shown that test subjects were able to remember more numbers than four-lettered words. A second hypothesis would be that the short-term memory could contain a constant amount of information in bits, as if the human brain worked similar to a computer. In a research conducted by Baddeley et al. [2], it was found that test subjects were able to remember easily-pronounceable words better than those difficult to pronounce, without dependence on the actual information content represented by the word. This would imply that the ‘bit restriction’ in the hypothesis cannot relate to the word’s textual form. Based on the research presented in Chapter 4.1, one may argue that the maximum capacity of the short-term memory is some constant number of memory chunks. In this case, the problem becomes that the size of a memory chunk is extremely difficult to estimate or nail down by research. Schweicker and Boruff [25], similarly to Baddeley et al., argued that the upper limit of the information contained in the short-term memory is directly related to how quickly the memorised content is possible to articulate verbally.

The ease-of-articulation theory is supported by many, still-used memorising tricks. An often suggested way to memorise things quickly is for example the visualisation of an apartment inside one’s mind, adding imagined visual cues inside the rooms to remind of the thing which they represent. When a list of memorised things has to be called to mind at a later point, it is easier and quicker to go through an imagined stroll through the visualised apartment than to remember the words for the memorised things by themselves. Already the Roman orator Cicero told a story of the Greek poet Simonides, who used a similar memorisation technique [13, p. 539]. The Japanese ‘expert of recalling numbers’, Ishihara, could recall 2,400 numbers in order with a notably small error margin. He used a speciality of the Japanese language, with which it is possible to connect a number to one or more syllables found in the Japanese language. This way, he converted the numbers to words in his mind, and then connected the words to imagined visualisations of 400 physical places that he knew. [13, p. 541].

When a beginner gets hands-on experience in a domain, it becomes easier and faster for them to memorise related information. For example the word combination ‘the stove is hot’ is naturally more difficult for humans to remember than the imagined feeling of touching a hot stove. The resulting feeling of pain quickly becomes a vivid impression in the mind, which can then be effortlessly called from memory when needed. The internalisation of such impressions accelerates the use of an individual’s fluid memory and chunking skills, and further increases the maximal information content stored in memory chunks.

Chapter 5: Expertise in chess and go

5.1 Measuring expertise in chess and go

It is hard to measure the level of expertise of a pianist because, out of two skilful players, it is hard to say which is better, even when closely inspecting their play. As games, chess and go give for an easier framework for measuring an individual’s expertise, because a single game of chess or go usually leads in a single winner. While a single game can be prone to random changes in a player’s performance, for instance due to human psychology, when measuring winning percentages after a large number of played games it becomes possible to say with relative confidence which of the players is more skilful.

In chess and go, some rating point system is often used, the primary utility of which is to make it easier for players to find opponents of a suitable level. Often, the used rating system is based on the estimation of winning percentages; for example, it might have been defined that a player of 1,000 rating points will defeat a player of 9,00 rating points in 75% of their games. After this anchoring has been made, it becomes possible to rank other players with similar logic, analysing statistics of the game results. When a player plays in a tournament, their rating points will change depending on their results by some formula—wins resulting in gained rating points and losses resulting in lost rating points. Of course, most rating point systems are generally somewhat more complex than this. [12].

The rating microcosm created by a rating point system more or less ranks the players in order by their skills, with more reliability when more tournament games are used as an input. The ranking system in itself cannot define an ‘expert’, but depending on the circumstances, it could be justified to for example call the best 1% of the players as experts. As, according to Chapter 1.1, a social aspect is involved in expertise, it isn’t unheard of that in a geographically secluded area, a local player who might only belong to the upper middle class worldwide might be called an expert.

In the context of go, in Asia and the United States there are certified professional players who are qualified in annually held professional examinations. Because the players participating in the examinations on average come from the very top spectrum of amateur players, and because very few players end up qualifying to become professionals, it is justified to call professional go players experts of the game. Worldwide, there are about a thousand professional players, over 400 of which are Japanese [19]. Professional players generally play in different tournaments from amateur players, and they are ranked by playing skill with a different system from amateurs. Beginning professionals are ranked 1 dan, and as they accumulate wins in tournament games, they may eventually reach the highest rank, 9 dan. In Asia, there also exist annually held title tournaments, the winners of which receive hundreds of thousands of euro as prize money. Title holders are commonly thought of as the most skilful current players.

There are several notably interesting features relating to expertise in chess and go. It is not uncommon at all to find a board position that will strongly divide the opinions of top players. Especially in the case of go, however, this can be understandable when taking into account the relatively large size of the game board; also, as there is no one clear way to winning the game, many considerably different game strategies can turn out to be effective. On the other hand, it is also easy to find game situations in which the vast majority of professional players would pick exactly one ‘correct move’, which however might not be found by the majority of amateur players. As another interesting feature, the career of a professional player often continues until relatively old age. For example in Japan, it is not rare for young professional, who has been recently performing exceedingly well in tournaments, to go to an older, less-well-performing professional’s school to study. In this fashion, when looking for a master player among top professionals, it may still be necessary to trust in the recommendations and evaluations of the top players themselves.

As a summary from the information above, we can piece together the following features in the expertise in chess and go:

1. Crystallised intelligence is vital in becoming a skilled player.
2. Crystallised intelligence alone is not enough for finding the best way of playing in all game situations.
3. Game performance is influenced by random factors in singular performances; more reliability in the observations can be reached by analysing a larger sample of game results.
4. Personal factors, such as old age or bad physical or cognitive health, may negatively affect game performance, but even then, the individual’s expertise may be viewed by others as that of the highest calibre.
5. While an individual’s playing skills may be measured quantitatively, the social aspect of expertise never disappears completely.

5.2 On the chess and go skills required for expertise

The process of learning to play chess or go can be well paralleled to that of learning a new language. In order to be able to fluently communicate with a language, an individual has to be able to create coherent sentences. In order to create sentences, an individual has to have an adequate vocabulary, and for that, he has to learn the writing characters (for example the alphabet). One could think that the alphabet of a language parallels the simplest operations based on the rules of chess and go, such as the way that chess pieces move around with on the board. Words can be paralleled to basic tactics and sentences to strategies: tactics consist of smaller operations, and strategies consist of tactics. A player skilled in the game has to know the different operations used in the game, and to be able to compose effective tactics and strategies out of them.

An initiate player starts their learning from operations, after which they move on to tactics and strategies; first starting from simple tactics, then continuing to simple strategies consisting of simple tactics, then to more complicated tactics, and so forth. Also referring to Chapter 4.4, this kind of an iteration is necessary to get the game mechanics rooted in the player’s long-term memory, in more effective form than a simple rote learned phrase. For example, it is not uncommon for a player to several times test out a new move in a particular local situation on the game board; if the move always yields a bad result, the player would soon develop a strong negative association to the move, which would likely result in them abolishing it as an option.

Operations, tactics, and strategies form the knowledge base of chess and go skills. Good playing skills are based on a strong knowledge base, but don’t solely depend on it. Other essential qualities are the player’s general fluid intelligence and power of reasoning, the player’s short-term memory capacity, the speed with which the player can retrieve content from their long-term memory, visual processing and pattern recognition skills, and the player’s ability to concentrate. [13, p. 590]. It is notable that while the required qualities are largely dependent on the individual’s cognitive characteristics, for example the individual’s diet and general well-being have a strong impact on their level of performance [27, p. 6].

5.3 On reaching the skills required for expertise in chess and go

As in most domains, with chess and go, too, an individual has to invest a lot of time and effort in order to reach a level of expertise. The approximate required time of ten years, suggested in Chapter 1.3, seems to be more or less of the correct magnitude, when looking at how much time most professional players had to use studying before they turned professional [19].

Deliberate practice is generally accepted and reliably proven to be one of the most efficient means to learn new skills [13, p. 601]. As was mentioned in Chapter 3, simple additional experience may not help at all at learning new skills or at improving one’s playing level. In deliberate practice, it is essential that the practice be done in a concentrated mindset, that it be regulated and monitored, and that involve immediate feedback from the individual’s performance. The feedback may come from a third-party evaluator, or the practiser may provide it for himself, for example by comparing their performance to expert performance in a similar situation. In deliberate practice, an individual is given goals and aims that are above his current performance level, and while the individual strives for these, failures and lower-level performance may result on a shorter time span. Individuals striving for expertise may, indeed, see these failures as opportunities to improve. [13, p. 601]. In other words, the crucial factor becomes that the individual actively and constantly strives to improve, and does not end up merely ‘accumulating more experience’ [9, p. 163]. Quantity doesn’t make up for quality at deliberate practice, either: it has been noted that even when looking at very different domains of expertise, experts usually train for only 4–5 hours a day on average. Too much practice may, depending on the domain, lead to practice injuries or so-called ‘burnout’. [13, p. 699].

Currently, it cannot be said with certainty what kind of deliberate practice is the most efficient, and how quickly one could theoretically become an expert in chess or go [9, p. 163]. The research reviewed by this thesis clearly leads one to believe, however, that the key lies in deliberate practice, one way or another. This is in a way also proven by the fact that more than 99% of all the professional go players (or players of similar skill) in the world are Asian, even though the number of Western players is by no means small. The greater number of Asian professional go players is partly caused by the fact that go has a longer history in Asia, and that apart from the newly-founded us professional go players’ initiative, all professional go associations have also been located in Asia [18]. Today, too, Japan, China, Korea and Taiwan have a vastly superior number of serious go schools than Western countries.

Research leads one to believe that it is possible to reach expertise without outside help. In this case, however, the individual has to make sure by themselves that they have the right tools for practice. In terms of chess or go, this means the possibility to play games, to review them (preferably with the opponent or other third parties), and access to different theory books, games played by current and former experts, and chess and go problems. Even when all these factors are accounted for, the individual has to be careful about the relative contents of their practice. Getting instruction from a current expert is definitely the easier solution for the student, as by so doing the student delegates the preparation of study material to the expert, and can themselves fully concentrate on the actual practice. With this procedure, the student has the additional advantage of having a third party who will actively monitor their improvement and give additional guidance and coaching if necessary.

Chapter 6: Summary

6.1 The recipe for expertise

The ten-year rule presented at the beginning of this thesis is a fitting, if not perfectly accurate description of what is required in order to become an expert. The fact is that that the process of developing the skills required of an expert eats up a big amount of practice time, which has also been estimated at around 10,000 hours [13, p. 692]. Just 10,000 hours of practice in itself doesn’t guarantee expertise, however; the used time also has to be efficient.

As was presented in Chapters 1.1.3 and 3.1, there are no meaningful cognitive differences between normal individuals that would affect the possibility of becoming an expert at something. Based on the research that has been made, it is safe to assume that the amount of practice is by far the most important factor in the development of expertise. In this sense, the notion of ‘talent’ so often mentioned in everyday language is something of a misleading term. In a sense, though, it is possible to interpret as talent the fact that an individual would be interested in a domain and ready to spend a lot of time and effort while practising at it.

While expertise in one domain does not correlate with the individual’s skills in another domain, many domains of expertise have many similar characteristics in terms of their required practice. Referring to Chapter 5.3, just having access to the necessary information in a domain does not guarantee learning, but even with that, the trainee has to actively and constantly aim to improve their skills. Whichever means are the most optimal to study the necessary information content seems to depend on the domain of expertise, but in all cases it is important that the individual be able to effectively assimilate the content that is important in the domain. Becoming an expert by one’s own means is possible, but the more popular, and presumably easier way for the student, is to study under an accomplished current expert.

6.2 The future of expertise

The material assessed in this thesis leads one to believe that the development of expertise requires a big amount of deliberate practice. For the human society, it is better the faster it is possible to train new experts, which makes research in this field important in the future as well. For this, the quote from Roger Bacon that was presented in Chapter 3.2 brings up an interesting theme. The information content that required 30 years to learn in the 13th century is nowadays presented so that most high schoolers have learned it in 12 years. In other words, by developing the teaching methods, it is possible to considerably reduce the amount of time required for becoming an expert, and, like this, the standards of expertise may notably change by time.

When thinking of the future, it is not hard to see the system of higher education being renewed so as to enable a more effective development of expertise even during one’s period of education. Currently, university studies provide for students big amounts of information from a wide spectrum of domains, which however is not enough for the development of any kind of expertise. Like this, a university student will get to develop their expertise only after they have finished their studies and got employed. A more effective procedure for the society as a whole, and a theme that sparks a lot of discussion even today, is more effectively directed studies in relation to the domain that a student will end up getting employed in.

Appendix A: The basics of go, and how a skilful go player perceives the game board and decides their move

Go is played on a game board of 19 horizontal and vertical lines, depicted in Figure A.1. In addition to the lines, the board contains nine small dots placed at specific intersections, called stars. The stars have no game-mechanical function, and merely make it easier to perceive the board. One of the players has black, the other white game pieces, called stones. The players take turns placing a single stone at one of the board’s empty intersections until neither of the players wants to continue. The black player starts.

The aim of the game is, by using one’s stones, to surround a bigger number of empty intersections than the opponent: in other words, at the end of the game, to control over 50% of the game board’s area (including the stones a player has on the board and the empty intersections surrounded by the stones). Depicted in Figure A.2 is a simplified end position of a game played on a board of 9 horizontal and vertical lines. Both players have just passed their move turn, signalling that they no longer want to continue the game. Black has surrounded a territory of 31 empty intersections and white a territory of 30: we say that Black wins the game by one point.

If a stone gets surrounded from all sides by opposing stones, so that it no longer has empty intersections next to it, the stone is removed from the board and moved to the capturing player’s bowl of captured stones. A stone’s connection to an empty intersection is called a liberty. Stones of the same colour that are horizontally or vertically next to each other form a single unit: in order to capture a unit, one has to fill all the liberties of every stone of the unit with one’s stones. In Figure A.3a, triangles depict the liberties of three different black units. Depicted in Figure A.3b is how the game board would look like if white got to capture all the black stones. Each captured stone is worth one point at the end of the game. Theoretically, then, it could be possible to win the game simply by capturing the opponent’s stones, but this almost never happens.

The only limitation to the moves that a player makes is that a player may not place a stone in such a place where it would immediately become captured. In Figure A.4a, White could not play at ‘a’, because the stone would get immediately captured by Black. The intersection marked with ‘a’ is called an eye of the black unit. A little exceptionally, however, White could play at ‘b’, in which case the black stones would get removed from the board.

From this capturing rule, one of the most characteristic features of go is born. In Figure A.4b there are two black units that may never be captured. The unit on the left side has formed two separate eyes, in neither of which White can play; in order to capture the black stones, White would somehow have to be able to place two stones at one time, which is impossible. The unit on the upper side of the board only has one eye of four intersections, but at any given time, Black is able to play at either ‘a’ or ‘b’ in order to divide the large eye into two smaller eyes, and White is not able to prevent this – again, White would need two stones at one time.

The surrounding of territories consisting of empty intersections, and the life and death of groups of stones comprise the core of go strategy. In Figure A.5a is a game position from a game played by two professional players. When thinking of the next move, a player should contemplate how he could surround the biggest amount of additional territory with the least amount of stones – but at the same time the player has to assess which stones on the board are alive, which are dead, which are strong and which are weak. The strength and weakness of stones correlates with how easily a unit is able to form itself two eyes. The survival of every stone is not required for the winning of the game, however; effective sacrifice strategies are often employed.

In figure A.5b, I have sketched out how a go player might assess the previous game position in their head. The sectors marked with yellow quite surely end up as the white player’s territory, while the sectors marked with blue may end up as the black player’s territory (however less confidently than the yellow-marked sectors). At the centre of the board, in the sector marked with red there is currently a conflict, where two white stones have become seemingly surrounded by black stones. In the close future, these stones will probably attempt to flee the black oppression or try to form two eyes.

In Figures A.6a and A.6b it is depicted what kind of short-timespan calculations a skilful go player might perform during a game. In Figure A.6a it is identifiable that the black group of stones marked with ‘a’ are in danger of dying because they are surrounded by white stones and unable to form two eyes. In this position, however, Black has a way out, because the white stones surrounding the black stones are positioned poorly.

A skilful player would probably find the 17-move sequence presented in Figure A.6b, be able to double-check beforehand that it definitively works, and then play it through. As a result, the white stones in the centre of the figure get captured. If the white player were as skilful as Black, he would likely not play at 2 at all, because he could see that Black has already escaped.

In Figure A.7 we have a more interesting-looking end position of a game played on a 9×9 board. At ‘a’, one black stone has been removed and moved to White’s captured stones. Both players have one stone on the board that is surrounded by the opponent’s stones (White at ‘b’, Black at ‘c’), but both players feel that they are not able to save their stone (there is not enough space to form two eyes). If both players pass at their next move turn, the game will enter the scoring phase, in which stones accepted as not possible to save get removed from the board. If the players disagree, the situation is played out. In this game, Black has 30 surrounded intersections and one prisoner while White has 26 surrounded intersections and two prisoners. In other words, Black wins the game by three points.

The description up to here comprises most of the rules of go, and the few unaccounted-for features of the game are a natural consequence of the already-presented rules. A few exceptions in the rules that come up in the game were not discussed in this presentation, such as the prohibition of repeating an earlier game position.

Literature

[1] P. L. Ackerman. A theory of adult intellectual development: Personality, interests, and knowledge. Intelligence, 22:227–257, 1996.
[2] A. D. Baddeley, N. Thomson, and M. Buchanan. Word-length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14:575–589, 1975.
[3] B. S. Bloom. The role of gifts and markers in the development of talent. Exceptional Children, 48(6):510–521, 1982.
[4] R. Brener. An experimental investigation of memory span. Journal of Experimental Psychology, 26(5):467–482, 1940.
[5] J. M. Burmeister. Studies in Human and Computer Go: Assessing the Game of Go as a Research Domain for Cognitive Science. PhD thesis, The University of Queensland, School of Computer Science and Electrical Engineering and School of Psychology, 2000.
[6] B. Carlson. Exceptional conditions, not exceptional talent, produce high achievers. The University of Chicago Magazine, 78(1):18–19, 49, 1985.
[7] J. B. Carroll. Human cognitive abilities. Cambridge: Cambridge University Press, 1993.
[8] R. B. Cattell. Abilities: Their structure, growth, and action. Oxford: Houghton Mifflin, 1971.
[9] N. Charness. The Role of Deliberate Practice in Chess Expertise. Applied Cognitive Psychology, 19:151–165, 2005.
[10] W. G. Chase and H. A. Simon. Perception in chess. Cognitive Psychology, 4:55–81, 1973.
[11] China Daily. Way to go. http://usa.chinadaily.com.cn/weekly/2012-04/27/content_15162240.htm, April 2012. [online] [Accessed 18.11.2012].
[12] European Go Database. EGF rating system. http://www.europeangodatabase.eu/EGD/EGF_rating_system.php, 2012. [online] [Accessed 11.11.2012].
[13] K. A. Ericsson, N. Charness, P. J. Feltovich, and R. R. Hoffman. The Cambridge Handbook of Expertise and Expert Performance. Cambridge: Cambridge University Press, 2006.
[14] K. A. Ericsson and W. Kintsch. Long-term working memory. Psychological Review, 105:211–245, 1995.
[15] K. A. Ericsson and J. Smith. Prospects and limits in the empirical study of expertise: An introduction. Cambridge: Cambridge University Press, 1991.
[16] Sir F. Galton. Hereditary genius: an inquiry into its laws and consequences. London: Macmillan and co., 1869.
[17] R. Glaser and M. T. H. Chi. Overview. In The nature of expertise, pages xv–xxviii. Hillsdale, NJ: Erlbaum, 1988.
[18] The American Go Association. AGA Professional System. http://www.usgo.org/aga-professional-system, 2012. [online] [Accessed 18.11.2012].
[19] The Japanese Go Association Nihon Ki-in. List of professional players. http://www.nihonkiin.or.jp/player/index.html, 2012. [online] [Accessed 11.11.2012].
[20] The Japanese Go Association Nihon Ki-in. Fujisawa Rina. http://www.nihonkiin.or.jp/player/htm/ki000434.htm, 2012. [online] [Accessed 21.10.2012].
[21] F. Gobet and H. A. Simon. Recall of rapidly presented random chess positions is a function of skill. Memory Cognition, 24:493–503, 1996.
[22] F. Gobet and H. A. Simon. Expert chess memory: Revisiting the chunking hypothesis. Memory, 6(3):225–255, 1998.
[23] A. De Groot. Though and choice in chess. The Hague, the Netherlands: Mouton, 1978.
[24] H. B. Richman, F. G. Gobet, J. J. Staszewski, and H. A. Simon. Perceptual and Memory Processes in the Acquisition of Expert Performance: The EPAM Model. Complex Information Processing Working Paper, 528, 1995.
[25] R. Schweickert and B. Boruff. Short-Term Memory Capacity: Magic Number or Magic Spell? Journal of Experimental Psychology: Learning, Memory, and Cognition, 12(3):419–425, 1986.
[26] L. A. Sosniak. Developing talent: Time, task and context. In Handbook of gifted education, pages 247–253. New York: Allyn Bacon, 2003.
[27] R. Uauy and D. A. Dangour. Nutrition in Brain Development and Aging: Role of Essential Fatty Acids. Nutrition reviews, 2006.