Hypothesis: IQ and Chess

Copeland wrote (photographic topic):

Kasparov in fact has an IQ of 135.  The major German magazine Der Spiegel had Kasparov tested at its own expense.  This is not a "low" IQ at all; indeed it is in the top 1% of the population.  Further, I believe that this level of general intelligence (roughly:  symbolic reasoning fluency and capacity) is more than sufficient to enable performance at the very highest levels of chess (and not just in Kasparov's case).  At lower levels of chess, IQ is more likely to be the limiting factor. 

My "Sufficient Intelligence" Hypothesis:

Below grandmaster level and/or an IQ of 130, chess play is more likely to be limited by general intelligence (as measured by IQ) than by other talents, and ranking may well follow the Levitt formula.  Above that level, IQ becomes less important than other talents and the experience and education of the chess player.

Support for the Hypothesis:

What is known about the predictive validity of IQ in a number of other domains suggests that IQ is very effective at predicting success at the lower and intermediate levels of an occupational or educational domain, where it often is the single best predictor of success or performance.  At higher ("master") levels of performance in professional and technical fields, as well as in graduate education, IQ still matters, but not as much.  Other talents, as well as learned expertise and accumulated knowledge,  become more powerful predictors at the top levels of achievement.

I believe that chess follows this pattern, and indeed, may be the single best example of it.  Thus Kasparov ascended to the highest level of chess with general intelligence that, while very superior, is not exceedingly rare: (a big-city high school should be able to fill a small classroom with students having IQs of 135 or more).  By contrast, Kasparov's world-class chess performance is exceedingly rare -- less than 1 in 1 million. 

Early in Kasparov's chess playing history, his high IQ (general intelligence) enabled him to quickly learn the game and accumulate expertise in depth.  As his experience increased, Kasparov probably used more specialized cognitive talents such as situational memory, in addition to an increasingly large fund of retained expertise, to continue his development as a chess master.

My Hypothesis, Restated:

1.  Very superior general intelligence is necessary, but not sufficient, to ascend to the highest levels of chess mastery.

2.  The highest levels of chess mastery require high intelligence, exceptional specialized cognitive talents -- most probably including situational memory -- and an extremely large fund of chess expertise and knowledge, acquired over a period of time through intense study and practice.

Interesting. a few years ago a study at University of Pretoria showed that children who were taught to play chess actually achieved an increase in IQ levels after a time period, I can't remember how long. they also became better at maths.

The more you do IQ-tests, the better you get at them, so which one do we go by? The 90-or-so I got when I was 13 and did a test on the internet, or the 130 I get now, 9 years later but also 15 IQ tests later?

IQ does not measure your generall intelligence, and it has been debunked that it does a number of times already.

"My Hypothesis, Restated:

1.  Very superior general intelligence is necessary, but not sufficient, to ascend to the highest levels of chess mastery.

2.  The highest levels of chess mastery require high intelligence, exceptional specialized cognitive talents -- most probably including situational memory -- and an extremely large fund of chess expertise and knowledge, acquired over a period of time through intense study and practice."

Sorry, but I don't think there is much substance to this. You say you need to be intelligent, have lots of chess knowledge and chess expertise from intense study and practice.... Isn't that a given? (eg To be a good sprinter you need to run fast.)

I'd be looking at how IQ tests somewhat overlap with chess rating as there are many questions about pattern recognition, which I find is a key aspect of chess. This may account for the link between high IQ and GM chess ratings..?

Nachos,

You say that my hypothesis -- that chess mastery is predicated on high IQ and significant expertise  -- is equivalent to: 

(NachosA.):  "To be a good sprinter you need to run fast".

First, observe that NachosA. is a logical fallacy -- it is a tautology that's always true.  If I understand you correctly, you are arguing that my hypothesis is "given" in its statement of itself, and so is also a tautological fallacy.  Forgive me, Nachos, but your reading of my post is incorrect.   Consider the following argument:

1.  This is my hypothesis, in concise form:

(ContrablueA.):  "Chess mastery requires high general intelligence,  voluminous acquired expertise,  and rare specialized talents"

2.  This is my hypothesis, restated in terms of sprinting:

(ContrablueB.):  "To be a good sprinter you need long legs, intensive training, and athletic talent"

(ContrablueC.) ContrablueA = ContrablueB (logical equivalent).

3.  But Nachos, you've argued that:

(NachosA.):  "To be a good sprinter you need to run fast".

is logically equivalent to:

(ContrablueB.):  "To be a good sprinter you need long legs, intensive training, and athletic talent"

So

(NachosB.)  NachosA = ContrablueB (logical equivalent)

But NachosB is, prima facie, false.  "To be a good sprinter you need to run fast" is a logical fallacy, and not at all equivalent to "to be a good sprinter you need long legs, intensive training, and athletic talent".  Thus your critique of my hypothesis is false. 

QED (or should I say, checkmate?)

Regards,

CB

I don't think anyone argues with your simplified hypothesis. In fact, I'd say it's generally accepted as necessary and sufficient to achieve a Master rating.

What people do have contention with is that IQ alone does not a chess master make. There is a huge body of knowledge that needs to be learned by any aspiring candidate that raw intelligence can't reasonably overcome given the time limit imposed on the game.(Presumably it's possible for an individual to have such a natural affinity for the game that they can find the best move without learning theory, but it's a far stretch to say they can do this at a pace of 3 minutes per move on standard time controls.)

Rooperi,

Thanks for your post.

I was not familiar with the Pretoria study, but it is consistent with recent, and very exciting, research on the training of working memory.  This research suggests that working memory, as well as performance IQ (which depends on working memory speed as well as capacity), can be improved through a very specific form of training called 2-back sequence recall.  The improvement in performance IQ, interestingly, was observed not just directly via increased memory test scores, but through improved accuracy and speed in mental arithmetic and perceptual reasoning tasks! 

Cognitive scientists find this "cross fertilization" effect to be very compelling evidence that genuine improvement in performance is taking place here.  This cannot be explained away as the result of subjects having become "test-smart", which is what I believe Tricklev was describing (more on this later).  Rather, people who trained using 2-back exercises were becoming more intelligent, at least temporarily.

I'm tempted to, and I therefore will, throw out another hypothesis:  that training (with documented improvement) in 2-back recall will be accompanied by improvment in chess playing performance, especially in timed games, if compared to untrained controls.  Woudn't that be interesting?  Training on something quite unrelated to chess, per se, should produce improvement in chess ability!  I would be pleased to be the study bookkeeper/statistician, if we could get volunteers to train and play!

In the meantime, Rooperi (and any chess colleagues who'd like to try this!), you can download this EXCELLENT, acadmically correct 2-back trainer, which is open source software, here:

http://brainworkshop.sourceforge.net/

It is completely free, and in spite of that, is probably the best tool in the genre.

Let me know what you think!

Regards,

CB

Tricklev - You said a number of rather juicy, thought provoking things that I'd like to join you on here.  I quote:

"The more you do IQ-tests, the better you get at them, so which one do we go by? The 90-or-so I got when I was 13 and did a test on the internet, or the 130 I get now, 9 years later but also 15 IQ tests later?

This post will deal with your above question.

The phenomenon of "test smartness", which is what is observed when people take lots of IQ tests, is well known and understood by cognitive scientists.  Test smartness is merely a practice effect that produces erroneously high IQ readings.  Now, this does not have anything to do with the validity of IQ tests -- that is a separate issue entirely.  This has to do with proper scientific and professional testing protocol.

You probably remember or have observed that teachers, psychologists, and other test proctors have taken considerable pains to collect all test booklets, forms, and so forth after administering IQ tests (or pre-employment thinking tests, military recruitment tests, school ability tests, and college entrance exams, all of which are proxy IQ tests).  This test security is important because it is KNOWN that practice with IQ tests significantly increases IQs, but without increasing intelligence!  Thus, practice produces TESTING ERRORS of MEASUREMENT. 

I'm sorry to say this, but your IQ of 130 is highly suspect because of test smartness (the good news is that your 90 IQ score is also suspect, not because of your age -- some IQ tests are just fine when given to people from 5 to 75 years old -- but because the test you took was on the internet).  That's not to say that a qualified psychometrist or psychologist would be unable evaluate your IQ accurately, because probably, they could.  But this would probably require administration of a test you hadn't taken before (or a test not very similar to tests you've taken before).

It may go without saying that qualified test administrators always must identify when the candidate previously took the same or a similar test.  If this is known, there are statistical formulae, supplied by the test publishers, that can be used to properly discount the raw scores on the test and thereby correct measurement errors arising from test-smartness.

This leads me to the subject of online IQ tests.  All online IQ tests, taken in unsupervised environments, are invalid, and usually for multiple and diverse reasons.  The lack of supervision, of course, ensures cheating as well as inability to control and correct test-retest test smartness.  Then there are the INHERENT weaknesses in the online item content.  Having just been given a standard, individual IQ test (the WAIS-III) by a psychologist (this was done because I'm more than twice your age of 22, and my doctors want to have a baseline for keeping an eye on my cerebral health), I can tell you in broad terms about some of the glaring problems in the content of online IQ tests:

1.  Many online IQ tests have test items that appear to have been taken out of Mensa puzzle books.  These items are designed to be challenging and entertaining, but are often very problematic when used in professional intelligence tests.  If too few people can answer them, due to difficulty, they should just be removed from the test for economic reasons.  Many online test items are much more difficult than genuine IQ test items, while other items appear trivially easy and a waste of most test-takers' time.

2.  A very popular IQ testing site, that gives candidates who achieve IQs of 125 (on their test, of course) membership in the "International High IQ Society", has test items that are just riddled with scientific flaws.  I took a look at their online test, and noticed several rather specialized items having to do with professional sports.  Such items are clearly biased against anyone who doesn't follow the sports about which the questions were derived, and would never appear in a professional test.  There are many other content biases and other problems.  You can research this matter further on the internet.

3.  Nearly all online IQ tests are timed.  While thinking speed is believed to be part of general intelligence, professional IQ tests isolate speeded test items into certain sections that focus on fluid intelligence, while the balance of the test contains items that are unspeeded (including matrix reasoning, vocabulary, etc).  The unspeeded items are sometimes referred to as "power tests" because such items can test accuracy and quality, as well as the ability reason with many moving parts.  This means that online IQ tests frequently omit power tests.  Chess players are very well aware of the distinction between timed and power mental tasks!  Any IQ test omitting power items will have sharply reduced predictive validity not only for chess mastery, but many other criteria that a well designed IQ test properly predicts.

4.  Online IQ tests are not properly normed and refined for criterion and internal validity.  These are technical, statistical flaws that generally reduce or eliminate the ability of online IQ tests to properly discriminate between test takers' abilities, or to compare a given test taker with a relevant population who is presumed to compete on the same test.  Professional IQ test publishers (not online IQ testing sites) spend millions of dollars (or Euros, etc) to refine the statistical validity of their tests. One flaw IMMEDIATELY apparent to a statistician (among other things, I am a statistician) who looks behind these online tests is that the score you get is compared only with scores obtained by others on the same online test, taken online. The problem?  No measures were taken to ensure that the group of people whose scores establish the test norms were a random representative sample from the population that is relevant to each test taker.  Boom!  The test is invalid!

Ok, that's about as much as I'd like to say about online IQ tests right now, except for this:  Online IQ tests associated with certain European Mensa groups appear to be much better in many respects (they still suffer from zero test security, and so are not accepted by third parties as valid).   So Tricklev, if you want an ESTIMATE of your IQ (this will be too high because of your test-smartness, but you can discount your score if you like), check out the Denmark, Belgium, Germany, and Spain Mensa sites for their online or take-home tests.

I will address your other statement about "debunking" of IQ in another post.

Regards,

CB

I've taken the WAIS-III and it definitely seems much more reliable than any sort of online test. I've been curious though about how the scale of different tests match up. I couldn't find any information, but I assume that the WAIS-III has a maximum score.

I would think the less comprehensive a test is, and thus the lower the maximum score that can be accurately represented, the more chance there is of representing scores as being lower than they actually are. Thus someone who gets say a 140 on an online test may get a 180 on the WAIS-III just because the online test couldn't accurately measure past 140.

Does anyone by chance know the highest possible IQ on a WAIS-III?

I,ve taken about five IQ tests from a year ago until to day (14-15) and on average am getting 130.  I doubt this is accurate, internet IQ tests are rarely accurate unless their the ones that you have to pay for.  Although, I took a free IQ test on Mensa and it said I was a genius which I am not.  The only sure way to test your IQ is by taking an officialy registered one.

Highest IQ:

I briefly researched this and according to Guiness "Marilyn Vos Santos" had the highest  IQ of all time, 228!  There has been much dispute however on the accuracy of this.

Hi Rubixcuber and RN9 --

Yes, I know some things about the WAIS-III because I asked my psychologist, or was told, and the information is fresh in my mind.

Interestingly, the MAXIMUM possible IQ score on the WAIS-III is 155.  (You can safely assume my score was less than or equal to 155 .  Other than that I don't know precisely what my IQ is -- my psych wouldn't tell me.  I was given percentiles, which I won't discuss here because how could you verify it anyway?  And why would anybody care?!?!?)

Ok, back to the WAIS-III.  For most purposes, 155 is adequate range for a test like the WAIS-III.  IQs at that level are very rare; an IQ of 115 is 1 in 7 people; 130 1 in 50; 145, about 1/1000 and 155, 1 in 50,000.  Of course, if a tester specializes in testing of high-IQ groups, such as chess players, 155 may not provide adequate range.  Another problem:  any score above 140 on the WAIS-III is likely to have some degree of underestimation of the "true" IQ because it is likely that the candiate achieved the ceiling score on one or more of the 13 subtests.  If this occurs, we don't know just how capable someone is on that particular measure, but we know he/she is more capable than our test was able to measure).

Ok, RN9 you mentioned Marilyn Vos Savant's 228 IQ.  In general, any IQ higher than 160 was achieved using an older method of scoring called ratio scoring.  Ratio IQs are NOT comparable to, for example, WAIS-III IQs, which are called "deviation scores".  Ratio IQs are invariably higher, and in many cases, ratio IQs in the 200s were revealed to be in the 140s-160s when retested and converted to deviation scores.  Marilyn Vos Savant's true, modern, deviation IQ is about 187 (this is still terrifyingly high).

 Deviation IQs of 190 probably reflect the limits of human "g" intelligence.  Such scores are expected to occur only once in 1,000,000,000 observations, so at any time, there are only a handful of such people on earth.  Keep in mind that with deviation IQs, the most rigorous and commercially most important IQ tests have test ceilings of 155 (WAIS-III) or 160 (Stanford-Binet 4 or 5; WAIS-IV).  The Stanford-Binet 5 can be used to derive ratio-type IQs above 160 using special procedures, but again, such scores won't mean the same thing as the deviation IQ scores.  Certain seldom-used and less reliable tests have deviation ceilings reaching into the low 200s, such as the Woodcock-Johnson III.

It is quite hilarious to see imbeciles boasting about their (usually fictitious) 170+ IQs on blogs and newsgroups.  I once asked one of these people to tell us what test he took to get such a high IQ.  He didn't like that! He told me to "go play with marbles in the street" (so I would get run over by a car, you see).   These people usually also assume that a person with an IQ of 170 is just like anybody else, but with an IQ of 170.  In other words, the boasters have no idea what an IQ of 170 MEANS or what it would be like to have such intelligence (let alone what it's like to have an IQ of 145, or 125).

One last comment before I wear my fingers into little nubs:  the relation betwen genius and high IQ.  I believe genius is like being a world-class chess player (in fact, I'd say that top chess players ARE geniuses), or a world-class composer (like Mozart) or scientist (like Einstein).  I think there is a connection between IQ and genius, but, just as I've argued in the case of chess, a superior IQ (with a frequency of between 1 in 20 to 1 in 100 (IQ 125 to IQ 139) is sufficient  genius, but genius also requires extensive acquired expertise, at times an abundance of specialized mental abilities, such as situational memory or pitch discrimination, and in many cases, creativity and motivational factors as well.  I do not believe that having an IQ of, say 160, is sufficient, in the absence of these other factors, to be a genius in any field.

Having said that, I do believe that profoundly high levels of g intelligence (IQs of 180 or more)  are enough to ensure that someone will be very precocious and based on that, may develop genius at early in life.  This occurs because such high initial capacity (mostly fluid IQ at that stage) enables a person to acquire and categorize expert knowledges, and so use that knowledge to be the basis for ever-larger amounts of crystallized intelligence.  That means that their brains start out with very large capacity, and then are filled with knowledge that is used to increase the capacity further via positive feedbacks.  By adolescence, such people are almost invariably acknowledged to be geniuses, even though the IQ pretty much stood alone, among mental talents, at least initially.  Remember, though, such  people are very, very rare:  1 in a million.

in fact I believe that a person with 100 or under, could easily be a NM, everyone has to think your above average, and everyone thinks they are above average, 

One question:  which one of the below statement is more achievable?

Is a person with high(er) IQ.  more able to grasp the advaced ideas of chess

                                                            or

A chess master has a good shot being close to a 180 IQ 

In other words, one who only knows chess ( B. Fisher ) How could he have a 180 IQ?

The maximum score on the WAIS-III is definitely not 155...

ContraBlue,

you have a rare combination of logical thought process (cf. partial-logic-al thought process of majority = 10% logic + 90% egoistic emotionalism) and extraordinary patience in answering each and every reply!

BTW, I think chess is all about the visualising capacity (64-squares), capacity of Randomly Alterable Memory (just joking) and a speedy mind - nothing to do with intelligence per se. But as you say, there might be some correlation with high IQ.

I think the guy is right. I mean face it, how many dumb can play chess (well)?

Hi Again Rubixcuber -

"The maximum score on the WAIS-III is definitely not 155..."

I did a bit of research on this, in the off chance my shrink was misinformed. 

The answer is:  the highest IQ score allowable (Full Scale, verbal or performance) on the WAIS-III is indeed 155. 

This is not just a limitation of the scoring algorithm or something.  It's a structural limit because the WAIS-III does not have sufficiently difficult items to test higher levels than 155.  Even if that were resolved, the WAIS-III was calibrated or "normed" using a sample of only 2500 people or so.  They probably didn't see IQs in this group beyond the high 140s...

This would not prevent a psychologist from contriving his/her own scoring system for the WAIS-III (possibly using ratio scores).  But such scores would be highly dubious, to say the least.  That would be like having your physician whip up a batch of penicillin using rotten pumpkins he keeps in his back yard...

Regards, CB

dsarkar, you said:

"I think chess is all about the visualising capacity (64-squares), capacity of Randomly Alterable Memory (just joking) and a speedy mind -- nothing to do with intelligence, per se"

I agree with the first part of your statement (my own made-up terminology for this is "situational memory"). 

However, as "g"  general intelligence can be thought of as "symbolic reasoning fluidity and capacity", I'm not sure it can be entirely distinct from memory, either working memory (where you "hold" the chessboard being manipulated), or long-term memory (where playbook moves and scenarios are kept).   The WAIS-III test I was just given, which is a standard IQ test, included 2 explicit memory tasks (recalling number lists of increasing length, backwards and forwards).  The test also included items that implicitly tested working memory:  I had to do every arithmetic problem in my head, without scratch paper.  For the authors of that IQ test, working memory appears inseparable from "g" intelligence processes.

I will say that the memory tasks associated with chess are far more formidable than the memory-related tasks on that IQ test.  That's one reason why I would say a superior IQ is necessary, but hardly sufficient, for master-level chess play.  Extraordinary "situational", or as you may say, 64 squares R.A.M. memory is needed as well.

Regards,

CB

*sigh*

While a popular thing to say, it is untrue.  The APA paper many consider definitive can be read here: http://www.gifted.uconn.edu/siegle/research/Correlation/Intelligence.pdf

IQ tests, when administered properly, give scientifically sound data.  The results, however, are frequently misused for political reasons.  THIS is the crux of the debate concerning IQ being a valid or invalid measure of intelligence.

Psychometrics (the branch of study IQ falls under) is a fully realized field of study considered valid by the scientific community.