A point I want to make more clearly is that one major reason that accurately predicting events that people are particularly interested in is so hard is because many of those events are the result of some kind of tournament.
We are fascinated by tournaments. (Just look at all the complaints that tonight's college football championship game only represents a quasi-tournament rather than an explicit tournament like the NCAA basketball championships).
So many of the things we most want experts to predict for us are explicit tournaments (e.g., the Super Bowl playoffs) that have been carefully designed to create maximum uncertainty in the later, more climactic rounds by matching the best contestants against each other.
For example, in about 90 or 100 tries, a #16 seeded team in the men's NCAA basketball team has never upset a #1 seeded team in the opening round, so basketball games are actually quite predictable when there is a fair-sized difference in quality between teams as determined by their seasonal performance. But subsequent round games become less predictable as the quality gap narrows, so public interest builds.
Or, the things we are interested in can be semi-explicit tournaments (e.g., the Presidential primary/general election process).
Or, unplanned events take on some of the nature of tournaments.
For example, people in the 19th Century were utterly fascinated by the Battle of Waterloo (June 18, 1815), which determined the basic political arrangements of Europe up through 1914. It was often remarked that the next century of European dominance was determined by the events of a few minutes in the crisis of the battle in which Napoleon's hitherto-undefeated Imperial Guard nearly broke through the British lines, but were stopped just short. Then, they faltered, broke, and ran.
Waterloo -- which Wellington called "a damn nice thing -- the nearest run thing you ever saw" -- was seen as evidence against large-scale deterministic theories of history, since so much depended upon something so close.
Contributing to Waterloo's fame was its numerous tournament-like aspects. For example, Bonaparte was the old champion making a stunning comeback. Wellington was the challenger who had never faced Napoleon before, but had worked his way up to the top by defeating his best marshals.
Finally, much that interests us are forged by vaguely tournament-like processes. For example, stock prices are the result of, in effect, competitions between those who think the price is too low and those who think it is too high.
My published articles are archived at iSteve.com -- Steve Sailer
Posts
37 comments:
Is Waterloo really an example of a critical turning point that determined European power relationships for a century? It seems unlikely to me. Had Napoleon won, he would just have had to face a half dozen more Waterloo's. At that point, the other European powers were as implacably opposed to him as the Allies were to Hitler in '42. It's sort of like saying that if the Germans had beat the Russians at Stalingrad, the outcome of the war might have been radically different. Not likely.
"Had Napoleon won, he would just have had to face a half dozen more Waterloo's. "
True, but it mattered who beat him. If it had been a Russian-Austrian force that defeated Napoleon after Britain and Prussia had been defeated then the post war balance of power might have been very different.
Tournaments vs everything else is not a viable distinction, because it's hard to find anything that is NOT a tournament.
It is no harder (in fact it is easier) to predict Super Bowls than the social-science data such as crime. What is different is resolution at which predictions are desired. In one case we want to know "microscopic" outcomes (who wins a given football game this week, what is the point spread) and in the other case "macroscopic" ones(metropolitan annual murder rates).
Obviously the larger-scale effects are easier to describe.
Let's compare Florida vs. Oklahoma in football (tournament) and in foreclosure rates (social science).
It's hard to predict Florida vs. Oklahoma in football because they are the survivors of a semi-tournament. It's easy to predict that Oklahoma has a lower foreclosure rate than Florida because they differ on various factors that predict foreclosure rates.
On the other hand, it would be easy to predict Oklahoma vs. South Dakota in football because we've been through a tournament and know Oklahoma is a lot better. On the other hand, I couldn't really guess whether OK of SD has a lower foreclosure rate since they are pretty similar on the factors that drive foreclosures.
It would be easy to predict Oklahoma vs.
I've only read a couple of online essays and interviews (and listened to the Charlie Rose interview), not read his books. But I think a big point your review seemed to be missing is that whole issue of payoff.
Lots of stuff is approximated by a bell curve safely, even though real-world distributions tend to have non-bell-curve properties, like being a little skewed one way or another, or having somewhat fat tails. It's safe to approximately model it with a bell curve because outliers don't dominate the outcomes.
For example, suppose you're designing building standards, and you want doorways tall enough that nearly everyone can walk in safely--say, that less than one person in a thousand will bump his head on that doorway. For this purpose, it doesn't matter if there are a few outliers whose height makes them impossible (say, +6 sigma) in bell-curve terms. They don't change the result much--they're just some folks that bump their noses on the top of the door, instead of the tops of their heads.
When outliers don't matter much, then you can use some approximate probability model that handles 99% of cases well, and it's okay. (This isn't just about the bell curve, it's about any probability model where you're making pretty good predictions most of the time, but missing the outliers.)
When outliers dominate the outcomes, then using that kind of probability model means you make really bad predictions. If you want to know how many people will pay their mortgages, you're in the realm of probability models that aren't all that sensitive to outliers. If you want to know where the US economy will go, or where science in the big wide world will go, you care a lot about outliers. For scientific advancement, one Gauss or Newton or Darwin trumps untold millions of normal, non-outlier people.
Sure, in terms of wealth, that's true -- one Warren Buffet accounts for a whole lot of Steve Sailers. But there are a lot of realms where it's not true.
Most financial crashes come about because a fairly small number of high-fliers over-invest, like Texas oilmen in 1981. But that's not true with the mortgage meltdown, which traces back to a huge number of not at all rich people taking out big mortgages.
"True, but it mattered who beat him. If it had been a Russian-Austrian force that defeated Napoleon after Britain and Prussia had been defeated then the post war balance of power might have been very different."
Well, the Russians did beat him in 1812 and occupied Paris in 1813. Disaster-wise, Waterloo was nothing compared to the Russian campaign. The army Wellington faced was a pale shadow of the pre-1812 French force. The "100 days" episode was really an afterthought.
Of course, the 1812 victory did not catapult Russia to the top of European politics. Britain was the top dog of 19th century international politics not because of any military victories that it achieved, but because it was the home of the Industrial Revolution. It was like China today, only more so - the world's manufacturing floor.
Napoleon's fortunes began to decline when the czar refused to abide by the European-wide anti-British embargo that the French insisted on. A refusal to trade with Britain would have meant an economic disaster for anybody in those days. Napoleon thought that if he let Russia opt out of the embargo without punishment, then everybody else in Europe would want to opt out, so he started a doomed war to punish Russia for trading with Britain.
In an indirect way he was slain by British economic might. Really, by the technological gap that Britain then maintained over the continent. See, IQ is important, but it's not everything. In the 19th century many European countries, not just Britain, maintained colonies in East Asia in spite of the Asians' slight advantage over Europeans in average IQ. That was just another technological gap at work.
So no, European power relationships would not have been radically different if Russians and Austrians defeated the French again in 1815. Either way the French would have been defeated and either way the British would have kept their place as the world's premier power for a century afterwards.
Steve, those are examples AGAINST the claim that football games and "social science" are dissimilar in their inherent predictability.
In both football and foreclosure rates, there are easy matchups to predict and hard matchups, as you point out.
In both football and foreclosures, the hard matchups can be created by pre-selecting the contestants to have "similar factors", as you also pointed out.
To be a bit pedantic, the only difference your examples indicate is that there are particular pairs of states (FL vs OK) whose football and foreclosure matchups have opposite degrees of competitiveness, i.e. that a state's football and foreclosure-rate rankings can differ. That is true but irrelevant to the question of which domain is more predictable.
The analogue (in the foreclosure world) of predicting an individual football game is to predict foreclosure of an individual loan. The "macro"-level forecasting of state or national foreclosure rates is of the same nature as statistical modelling of all NCAA football games the past century. At either the macro and micro level there is no essential difference in the nature or difficulty of the football and foreclosure (or "social science") problems.
"Steve, those are examples AGAINST the claim that football games and "social science" are dissimilar in their inherent predictability."
But that's not what I'm saying in this post. I quite agree that most _possible_ football games (e.g., USC v. Cal Tech) would be as boringly and depressingly predictable as most social science comparisons are. It's the tournament-like organization of college football, which is largely designed to create pleasurable uncertainty, that creates the illusion that football games are hard to predict. (And hardcore fans up the ante of uncertainty even farther by betting not on who will win but on who will beat the point spread).
What I am saying is that those particular football games that seem interesting and "important" to us, such as Florida v. Oklahoma, are ones that are hard to predict. Florida v. Citadel (70-7, if memory serves) was not interesting or important-seeming before, during, or after the game.
If it were an annual football game, the San Marino v. Compton public school test score competition would have been canceled long ago for lack of interest, like the old College All Stars vs. NFL Champions exhibition game at Soldier Field was called off in 1976.
But San Marino v. Compton isn't a tournament contrived to entertain fans betting on the outcome. It's real life, and it just goes on and on.
Steve,
that recapitulates the undisputed (and very well stated) parts of your blog posting. It does not address the disputed part, which was the claim (last paragraph of blog post) that social science observables aren't subject to the same type of tournament-driven phenomena.
In reality, most of the large scale Soc Sci quantities one might be interested in, are driven by various forms of tug-of-war that play exactly the same role as the "tournament" process used to pick Oklahoma and Florida for the championship bowl game. That is, they
(a) restrict range
(b) reduce predictability
(c) focus interest on the unpredictable part (e.g. in market prices of CDS based on foreclosure rates), and so on. The similarities in the two situations are fairly precise.
Thus, if foreclosure rates got high the lenders would adjust the loan availability or loan terms accordingly. The profitability of banks would depend on their success rate in dealing with the hard-to-predict loans (not the obviously accepted or rejected ones). Markets tend to concentrate their interest on the less predictable aspects of the foreclosure rates (by holding tournaments to price securities dependent on those rates), the easily predicted component already being "priced into" the calculations of the market participants.
Conversely, one can predict football game winners (resp. point spreads) reasonably well using the odds offered by various bookmakers, with success rate similar to the prediction rate of loan default (or losses due to the default) in the foreclosure models used by banks.
The level of interest in questions such as Compton and San Marino foreclosure rates is higher than that in predicting NCAA football games.... when measured in dollars rather than audience size.
Crime rates aren't that easy to predict, at least not accurately.
In fact, a lot of stuff is not easy to predict accurately since stuff that happens in the future depend on the past in several complex ways.
For instance, what will the murder rate in Los Angeles (in homicides per 1000) be exactly 10 years from now? Could you make a prediction that is within + or - 1%? Would you bet your house on that prediction?
Can you predict what the unemployment rate in California will be exactly 5 years from now? Again, could you make your prediction within + or - 1% and would you bet your house on that prediction?
I could go on and on forever with predictions that are difficult to make accurately, but you get the picture.
You seem to think that because you can make comparative predictions in some areas of social science (such as test scores of San Marino being higher that those of Compton), that you can make accurate predictions about everything in social science.
Sorry, Steve, you don't have a crystal ball.
Q. "For instance, what will the murder rate in Los Angeles (in homicides per 1000) be exactly 10 years from now? Could you make a prediction that is within + or - 1%? Would you bet your house on that prediction?"
A. I don't know. No. No.
But when did I ever say I could?
In contrast, could I make fairly accurate _relative_ homicide rate predictions ten years out for Los Angeles v. Memphis or Los Angeles v. Portland? After all, that's the main thing sports experts are called upon to do: pick which city is going to come out on top.
And, yes, I think I could.
There are tournament-like aspects that make city crime rates somewhat variable over time -- for example, if things get really bad, big cities hire William Bratton to be police chief. So, there is a lot of variation over the years, but it's certainly easier to predict than to predict which of a pair of NFL cities will have a better record in ten years.
"In reality, most of the large scale Soc Sci quantities one might be interested in, are driven by various forms of tug-of-war that play exactly the same role as the "tournament" process used to pick Oklahoma and Florida for the championship bowl game. ... Thus, if foreclosure rates got high the lenders would adjust the loan availability or loan terms accordingly."
Right. Participants in _markets_ respond to tournament results. In fact, they are supposed to anticipate, but in this case they signally did not. Now, I'm trying to explain to them why they screwed up so they don't make the same mistake twice, and instead go make some new mistake.
But, a lot of social science patterns don't respond over short periods to tournament pressures. In the test score tournament, San Marino has beaten Compton 50 years (or whatever) straight. And it's not likely to change in the next few decades due to the differences between the two municipalities in housing stock, terrain, tree cover, and street design. Short of the Big One that reduces all of SoCal to rubble, San Marino is going to have smarter kids on average than Compton.
Steve, you're 100% right here -- these criticisms are incoherent.
First, Taleb is basically another person pushing the idea that the exception *is* the rule. That's the animating idea of our age -- the Hollywood-aided idea that black doctors and white racists make up the warp and woof of our society rather than the exception.
Second, like Jared Diamond, Taleb spends just enough time on indisputable facts that it is possible to fashion apologetics for his work.
Diamond: "He's not saying race doesn't exist, look! On page 237 he admits that the Xhosa and Swedes are actually distinct population groups"
Taleb: "He's just telling us to be on the lookout for fat-tailed distributions"
Of course, such apologetics do not capture the focus and intensity of the work.
Third, your insight regarding tournaments is absolutely right. A Bernoulli random variable has maximum variance when p = .5, and minimum when it is near 0 or 1.
Fourth, I think the right response to king obama is this: in the long term crime rates are going to increase as the NAM proportion of the country increases, assuming no genetic engineering or widespread use of tranquilizing drugs. Similarly in the long term the US is predictably going to decline in world standing as it has a smaller and smaller proportion capable of doing the tasks required in a technologically advanced economy.
Plot GDP vs. IQ and you'll see that the US is well above the regression line. Who knows how the actual correction will happen? Will it be govt. handouts to minorities on steroids: reparations for blacks *and* Hispanics in Obama's second term? Will it be disaffected youfs burning cars like in France? Will it be the collapse of Social Security? Perhaps all of the above. What we know is that American society has AIDS, a form of immunodeficiency that prevents distinguishing between ingroup and outgroup. The decline is guaranteed -- all we don't know is what particular condition it will inevitably succumb to.
As for Ivy League Bastard:
Thus, if foreclosure rates got high the lenders would adjust the loan availability or loan terms accordingly.
This is just another pie-in-the-sky equilibrium argument. In reality, high foreclosure rates did not stimulate rational responses, because the rational response was so-called "redlining".
The only reason credit was extended to these people in the first place was because of government lawsuits against redlining. It was the disparate impact doctrine yet again -- any statistic re: blacks and Hispanics which is out of line with population expectations must be the fault of whites.
"On the other hand, the kind of phenomena that the social sciences (and much of public policy) are concerned with -- crime rates, test scores, and the like -- tend not to be very tournament-like at all, and thus tend to be fairly predictable."
A banal point: The public tournament in social sciences tends not to be fought for the result or even accurate prediction of the result, but for interpretation of the result so that particular policies can be maintained or promoted by the interpreters. These tournaments can be skewed heavily toward one side by the acceptability or pleasantness of one interpretation compared with another. In fact, you could make the argument that the easier it is to predict a social science result, the easier it will be to promote a contradictory or merely distracting interpretation of the result.
Steve:
"In contrast, could I make fairly accurate _relative_ homicide rate predictions ten years out for Los Angeles v. Memphis or Los Angeles v. Portland?"
Memphis will be higher than LA will be higher than Portland, because Memphis is mostly black, LA Hispanic, and Portland white? Is that right?
Wellington also made serious use of the Prussian regiments and afterwards of course promptly forgot about them.
Steve,
Certainly you are correct. Most important issues in the Social Sciences are not interesting because the data is so crystal clear. An honest, rational observer would have to admit that if you test every kid in America at the age of five, and then have the government spend lavishly on half of the low IQ kids and spend very little on the other half of the low IQ kids that the life outcome at the age of 21 will not be better for the ones that the government spends lavishly on. All the government money is wasted on these kids. Data indicates that the "die is cast" by the age of five
I myself don't feed at the public trough so I am just fine with the message that government spending on those with a low IQ is wasted and that it is better to just lower taxes on all of us and not spend this money
That being said, Steve can you read up on the Abecedarian Project and give your opinion? I am a pretty cynical bastard and I want head start and a whole bunch of other social programs shut down because they don't work. But to me the data on the Abecedarian Project seems pretty compelling - it looks like there is some evidence that if you take the poorest kids - the kids with the most screwed up parents - and you basically take these kids away from the parents for much of the day from birth to age five and give them intensive preschool that this really has a huge positive financial impact -
Basically the kids that went through the Abecedarian Project avoided teenage pregnancy and actually got decent skilled jobs instead of winding up in prison.
Again, I am against almost all social programs and I want most useless government workers fired so that taxes can be lowered. But I think the Abecedarian Project is worth taking a look at
Steve, you still don't get it.
A Black Swan is not extremistan. It is simply a lack of enough data to form an accurate mental model. Because there are huge parts of the data unknown, and worse, people don't know they don't know.
It's false confidence that the underlying data in the model is descriptive.
An example of a Black Swan might be the theory of disease. Prior to medical pioneers like Lister, it was thought miasmas and such caused disease. Not bacteria or viruses. The theory of infectious disease is a Black Swan.
Not because the events were rare (infectious disease is not rare) but because the mental model of the data was flawed and nobody knew it.
The Mortgage crisis was no Black Swan because people (enough anyway) knew quite well what was going on.
A real Black Swan would be contact by Aliens. The Aliens having of course always been there, but Humans not knowing it.
I'm afraid you are pushing this model places it cannot really go.
Now, I'm trying to explain to them why they screwed up so they don't make the same mistake twice, and instead go make some new mistake.
At the most fundamental level, "they", which is to say everyone in the chain from the home appraiser to the pension fund buying the securitized mortgage bond, screwed up because their incentive schedule was biased. They received bonuses based on short-term results that could not be clawed back if long-term results did not keep up. It was heads I win, tails you lose.
As long as corporate management can privatize profits and socialize losses to their large amorphous shareholders, this will be the case.
Henry Canaday wisely points out:
"The public tournament in social sciences tends not to be fought for the result or even accurate prediction of the result, but for interpretation of the result so that particular policies can be maintained or promoted by the interpreters. These tournaments can be skewed heavily toward one side by the acceptability or pleasantness of one interpretation compared with another."
Well said.
I'm half-way through the book and still don't have a rigorous filter to decide what is a black swan and what isn't. At this point in my study of his book, the bird that most comes to mind is not the swan, but the peacock. I find that the guy's ego renders the book less readable and his ideas harder to understand.
The paleo poet and essayist Wendell Berry once wrote that capitalism was like a tournament to determine the world champion corporation.
If so, then 2008 was the equivalent of the first weekend of March Madness.
I am a pretty cynical bastard and I want head start and a whole bunch of other social programs shut down because they don't work. But to me the data on the Abecedarian Project seems pretty compelling
Right. You are a concerned Christian conservative as well, I assume?
As for Abecedarian:
http://www.campusreportonline.net/main/articles.php?id=1488
You have heard much about studies that are supposed to justify the expansion of the early childhood programs – programs like Perry and Chicago and Abecedarian. Yet, you never hear that all of these studies have significant methodological flaws, small sample sizes, are very expensive, have not been replicated, would be very hard to produce on a large scale, and either don’t have the impacts advertised or positive results are more due to parental influence or involvement than on participation in the program.
Perry – It only included 123 children; the beneficial results have never been replicated, which means that it would be hard to reproduce on a large scale; there were significant methodological flaws admitted by people such as the co-founder of Head Start; and it required a mother to be at home in the experimental, but not the control group, which could well account for the gains made. In addition, even though program children did better than controls, still nearly one-third of participating children dropped out of high school, nearly one-third of the children were arrested, and three of five participating children received welfare assistance as adults.
Chicago – “It is possible that parental involvement explains more of the variance in outcome among inner-city children than do structured programs. . . . If policy makers mistakenly accept the conclusion that preschool intervention results in less criminal activity later, they may mistakenly invest in these programs when the money might be better invested in parenting skill programs and other interventions to increase parental involvement.” This indicates a need for parents to be involved. It also points to the reams of research that children perform the best academically and socially in parentally involved families and even better in two parent families.
Abecedarian – It cost $20,000 per child per year back when it was performed; it took children away from their mothers for 8 hours per day, five days per week, starting at age 4 months; they had to combine the IQ results from all 4 groups in order to see any benefit, because they actually had two groups that lost IQ points; and the mother’s IQ was a more powerful predictor of the child’s IQ than participation in the program.
-----
The NLSY data is on 10000+ people. The SAT data is on easily 1M people. Yet naive liberals cling to uncontrolled studies like Abecedarian -- with approx. 100 subjects! -- like life preservers in the ocean.
There is just no question that the totality of evidence for any rational observer is on the hereditarian side of the ledger.
Headache - to be fair I dont think Wellington 'used' the Prussian regiments.
I think the arrival of Blucher and his Prussians confirmed the allied victory. Im not sure they were ever under the command of Wellington
Sooners and Gators is on Thursday night, January 8th, not the night of this post.
Gitcher sample swans straight!
;)
Off topic: Steve, you apparently got bored with the Infosys/Comverse Israeli spy story revealed shortly after 9/11...
Israel Is Spying In And On The U.S.? Part 1
But this reporter James Bamford has stuck with the story:
James Bamford describes the NSA's subcontractors
Israeli companies Narus and Verint are now operating as the information gateways and gatherers for our National Security Agency (NSA). The Israelis (Mossad) are in control of the eavesdropping and tapping software embedded in the AT&T and Verizon networks which represent 90% of telephone traffic in the United States.
http://en.wikipedia.org/wiki/Narus
http://en.wikipedia.org/wiki/Verint
Bamford has a lot of other videos up at youtube
Steve,
As every Polish, German and French person (and probably every continental European) knows, victory in Waterloo was 'bought' by the British (courtesy of Rothschild's Bank), when the Prussian General Blucher was paid a very substantial cash sum in gold to bring in his cavalry at a very decisive moment to relieve the exhausted British.
That Rothschild gold saved British skin is one of the most salient features of modern history.
I am sure that almost everyone on this board agrees that the IQ of an adult is 90% set in stone on the day that that person is born.
Almost all efforts to raise the IQ of a person after birth are doomed to failure.
However, Steve wrote that there is evidence that if you take a newborn and breast feed him, he will have slightly higher IQ as an adult than he would if he were bottle fed.
This is by no means proven - best way to prove it would be to take a thousand identical newborn twins, breast feed one and bottle feed the other and then check back 40 years later and see what happened.
I am not arguing one way or the other for breast feeding, what i am saying is that there is some possibility that doing something for kids under the age of five will raise their IQ.
I may be mistaken but I think Steve has cautiously suggested (1) putting iodine in salt in the countries where that is not now done (2) encouraging breast feeding. So I'd like Steve to do his own evaluation of Abecedarian
Boy, very long replies on this thread - too long to comment on any one of them individually.
[BTW, I've got lots of thoughts on Waterloo, but I'm nursing a really neat movie script in my mind, so I suppose I ought to just keep my mouth shut about that for the time being.]
Anyway, I think a lot of you are being unfair to Taleb - if you actually read Taleb, his fundamental beef is with intellectual/egghead/ivory tower ARROGANCE and especially how that arrogance has a really bad tendency to transmogriphy elite opinion into something which is essentially indistinguishable from the legalistics of a pagan religion.
In particular, the defining [Freudian, if you will] event in Taleb's life was his witnessing of the Beirut of his childhood - the Paris of the Mediterranean - disintegrate into chaos and war and living [or dying] hell.
And the reason that that childhood experience was so horrifyingly unexpected [by the standards of the elitist Lebanese salon] is PRECISELY because people made the mistake of assuming that Muslim Arabs are just like Christian Arabs: That - to paraphrase Shrubby Bushitler - Muslims, LIKE ALL PEOPLES, yearn for freedom [and peace].
But, of course, Muslims most certainly do not yearn for freedom [or peace], as anyone who had spent even five minutes glancing at the Koran and the Hadith and the history of worldwide Muslim conquest could have told you: the lifeforce of the Muslim is the bloodlust, and he yearns for the absolutist tyranny of Sharia - just as anyone who had spent even five minutes glancing at a century's worth of educational data [neatly summarized by Murray and Herrnstein, way back in 1994 - ironically enough, almost exactly when the Boston Fed was initiating their sociopolitical Jihad against classical lending standards] could have predicted that various peoples [especially those aboriginal to Western Africa and Central America], with IQ distributions a good standard deviation to the left of Caucasian IQ distributions, would prove to have catastrophically greater default rates on more or less any non-trivial loan [much less a loan the size of a house mortgage].
But the legalistics of the pagan religion known as Political Correctness [or, more accurately, the subsect of Political Correctness within the greater theocratic tyranny of Nihilism] prevented even the Masters of the Universe from anticipating the consequences of ignoring [or at least intentionally obscuring] this fundamental tautology.
PS: On the other hand, as some of the cynics on this board have posited, it's difficult to be familiar with the historical record and not at least wonder whether the Masters of the Universe - like Mr & Mrs Golden West and the Broad Street Boyz - knew exactly what the hell they were doing...
The comments coming in from Henry Canaday, Steve and others continue to make the opposite of their intended point. That is, they show the SIMILARITY of social-science and sports observables.
The "public tournament for interpretation of results [in social science]" not only exists in sports, it permeates it. Consider how much press and internet is devoted to comparative debates about sports teams and their players, or the many official rankings published, or the multitude of honors (All-Star, All-American, Hall of Fame, MVP, Heisman, etc) which are awarded not through game outcomes but through voting tournaments among sportswriters.
Those tournaments to interpret outcomes are, however, distinct from the tournament-style processes that drive the outcomes themselves.
Such processes drive things to a similar extent in both social science and athletics, except that sports prediction tends to be easier.
Talking about "markets" also does not point to any meaningful distinction between Soc Sci and athletics. Certainly there are explicit and implied markets that are sensitive to most social observables (such as people deciding where to live, attend school, or open a business based on local crime or test score statistics, which is then reflected in various prices). So that too is a point of similarity, not difference, between sports and social science observables.
The only difference between athletic and social-science observables indicated by Steve's examples, is that DIFFERENT (SCALES OF) QUESTIONS are of interest in the two situations.
The reason "test scores" sounds like an easier prediction problem than "football scores" is that in social science the question almost always refers to large populations (averaged scores from thousands of high school students in Compton vs San Marino) while in football we want to know individual game results. The logically correct analogy would have been predicting or comparing test scores of one student from Compton to the scores of a one student from San Marino.
To recap: a game outcome in sports corresponds to a measurement of an "individual" (one person, one mortgage loan, one family) in the social science statistics. The individual data are of course partly easy to predict and partly very random; a football team from a huge Texas public high school is likely to beat one from a Massachusetts prep school but the latter might have a player who makes the NFL draft. The statistics aggregated from many such data points (townwide test scores, home runs per at-bat, foreclosure rates) are much more predictable and people can and do build usable mathematical models of those in both social science and athletics (Moneyball).
The sports prediction models tend to be more accurate than the ones for social science questions, because the games that they measure are simpler, and relevant data are more available.
A real Black Swan would be contact by Aliens. The Aliens having of course always been there, but Humans not knowing it.
But some of us do know it. Check out the flick "They Live".
Similar to Gettysburg, where Pickett's Charge, now seen in retrospect as tragically futile, almost succeeded, and might have been much less close had the Confederate artillery been angled just a little lower and hit the Union center directly rather than churning the ground well behind them.
Furthermore, Stalingrad was indeed decisive. Had the Germans secured it the Nazis would have had a free hand to conquer the Caucasus and the all-important Baku oil fields. That would have starved the Soviet tank armies of desperately needed fuel, and they would have ground to a halt just as the Nazis did in the Battle of the Bulge (after the loss of Baku, the Romanian oilfields at Ploesti, and the synthetic oil plant in Bavaria).
It's sort of like saying that if the Germans had beat the Russians at Stalingrad, the outcome of the war might have been radically different.
Waterloo wasn't Napoleon's Stalingrad, it was his Battle of the Bulge.
Well, the Russians did beat him in 1812 and occupied Paris in 1813.
In 1814, you mean?
Anyway, despite the Russian catastrophe, Napoleon's decisive defeat didn't come until the campaign in Saxony in 1813. The battle of Leipzig ("Battle of the Nations") was ten times more crucial than Waterloo. I wonder why no film has ever been made about that colossal fight, which was the biggest clash of armies in world history until WWI - three days of bloody battle, Europe's Gettysburg.
Dear Ivy League Bastard:
What exactly are we arguing about anymore?
_Of course_ sports and social science topics such as school test scores share much in common.
The point is, however, that if a competition in sports gets too boring and predictable, it gets called off. For example, the Penn State vs. Harvard football game was a big deal in 1921, but they don't play it anymore because it would be a wipeout.
In contrast, Beverly Hills versus Compton in student test performance just grinds on forever, even though everybody knows who will do better and everybody is thoroughly sick of the topic by now.
dear Steve,
your post on Colored Swans and tournaments had eleven paragraphs, of which the first ten are correct, concise, best-of-the-blogosphere quality material. The last, eleventh paragraph appears to be a (mathematically bogus) attempt to stretch that material into snide commentary about the underclass and its test scores and crime rates.
Since your entire posting was about an interesting mathematical technicality, I assume you are interested (as were others in the Comments discussion) in actually getting the math right.
Thus, what we're arguing about is the claim that social science outcomes such as crime rates and the rest are (a) more predictable, and (b) less tournament-like, than sports outcomes. The first statement is virtually always false (social science "games" involve more players, more variables, vague and variable rules, and so are harder to predict). The second statement is false more often than not.
The correct statement is that large-scale outcomes (aggregated statistics about numerous math tests, murders, mortgage loans or home runs) are fundamentally easier to predict than outcomes that are small-scale (who wins the football game, will there be murders in my neighborhood this year, determine the point spread). Since most social science outcomes of political interest are large-scale,
while most interesting sports outcomes are small-scale, that is a sufficient explanation of any recurring difference in predictability. The business about tournaments does not explain it, as elaborated here and above.
As for Compton vs Beverly Hills, nobody is running a "tournament" between them any more than Harvard and Penn State are playing each other (and of course there are objective metrics that can be used as "football test scores", to compare teams in different leagues). The difference is, again, only in what interests us: nobody cares whether there is a permanent football underclass of Northeast private universities, while having a permanent racial underclass in all the important life outcomes (money, education, safety) has implications for the whole society.
Post a Comment