Saturday, July 24, 2010

On Luck, and Definitions

Consider the stars (she said, oratorically, with a great flourish of arms) (yeah, it’s going to be one of those). Consider them, more specifically, as they look to the unaided eye- which is to say, human eyes, on earth, at night, without benefit of telescope, perhaps lying in a field on a summer night at some nonspecific high-altitude point on the globe. From this perspective, stars are points of white-to-yellowing light of varying size and brightness, strewn across the sky. With the exception of the Milky Way stripe sometimes visible, their distribution appears chaotic. Some cluster close together, others are isolated, and their position doesn’t correlate in any way with their size or brightness or color tone. They might as well be flecks of sawdust on the floor, Cinderella’s lentils in the ashes, for all the symmetry or order they show. Stars look random.

There is nothing in the universe, except perhaps nothingness itself, which is less satisfying than random. The human brain cannot fathom random. It hates and fears random. It sees random and it reaches for its gun, and then uses that gun to shoot out assorted bits of random until a nice floral pattern is created with a kind of an arabesque around the side, and then it calls other human brains over and has a highly contentious discussion about the aesthetic merits of the design. This is not hyperbole. We are not just pattern-seeking creatures, we are pattern-inventing creatures.

So obviously people are not going to sit around all night under all these twinkly lights and just contemplate the senselessness of them. People are especially not going to do this before the advent of television. Or roofs, for that matter. We are not going to look at something all damn night, every night, except when it’s cloudy, and not wonder what the hell that shit is and what it’s doing and how it might conceivably benefit or harm us. And once people start asking questions, people immediately start having answers for them, because just like we are never going to serenely contemplate randomness, we are also never going to tolerate unanswered questions.

Most of what people did with the stars, once upon a time, was tell stories about them. Looking at them long enough, we came to see our own reflection in them- or, at least, the reflection of our world, with the same animals and objects, in dream-or-nightmare outline. We imagined the sky more mundane than mythic, a place where hunters chased bears and bulls and only the odd mer-goat. Just like us, really, only bigger.

We looked for patterns and we found them, we asked for stories and we told them, we feared influence and we felt it. For a good 3.6 millennia, astrology was how man understood the skies- in terms of old gods and spectral influences, ancient powers radiating across the universe, seeping into your soul at the moment you took came screaming out of your mother, and making you the charismatic, sensitive soul you are to this very day. Now you can look at it as a joke or a novelty, an occasional diversion in the paper, but back in the day, astrology was science, bitch. Astrology was a goddamn profession. If you could send your kid to any kind of school, you were hoping they might be smart enough and study hard enough to become an astrologer. Governments kept astrologers on the payroll, like freelance consultants, to advise on decisions of the day. For a long time, astrology and divine revelation constituted the totality of all knowledge about outer space and the universe beyond terra firma.

Here’s the thing: astrology did some good work. If you leave aside the explanatory aspect of it, the old time astrologers accumulated an awesome collection of observations about how these star-sun-moon-planet things behaved. They could tell you the rising and setting positions of all of those orbs at any time of year, explain and anticipate all the seasonal changes in star patterns, chart the movements of the planets from night to night and month to month. It took thousands of years of astrologers watching and waiting and studying to produce all the underlying data that astronomers later used to support all their Rational Thinking. It was the astrologers who mapped the skies, and if you had asked them what stars were, their answer would crack you up, Mr. or Ms. Modern Educated Person, but then again, they could navigate by them, and you can’t. So whose understanding is better?

This actually does relate to hockey, by the way.

Right now.

Hockey events look random, particularly goals. To an uninitiated person, goals seem to bubble up abruptly from a sea of purposeless motion, and this is not so very far from the objective, observational truth. There is no definable pattern to when goals occur in a game, which team scores them under what circumstances. A goal is just as likely this minute or the next or the next, which is why hockey is so entertaining to people with short attention spans. A deep knowledge- even the deepest knowledge- of the game only mitigates this unpredictability slightly. A hockey expert can see a scoring chance developing perhaps two seconds ahead of a lay person, and has a better understanding of who to watch and when, but even the sages are helpless to know which shots will go in.

The best of hockey knowledge right now is not so far from astrology. It has accumulated a massive body of observational evidence, and it has noted certain patterns and correlations in that evidence, and it has developed a number of very complex theories about how the game works. Nevertheless, for all that, it is uniformly terrible in predicting the sport- what will come of a particular shift, game, season, career, or franchise- and is routinely baffled and befuddled by particular game events. The descriptive knowledge hovers at slightly better than guesswork, the explanatory knowledge sinks to somewhere just above “Because I said so.” The best of all possible hockey knowledge is, frankly, not very good, as knowledge goes.

The problem is this pattern-seeking impulse. We see something mystifying, and we look for the explanation closest to hand, which is like as not to be an anecdote from a previous mystifying incident, and so we link together unknowing with unknowing until it starts to feel like knowing. But it’s not. Orion was never actually a hunter. Chris Drury was never actually clutch. It just looks that way from a certain angle, particularly if you’re bent on finding something.

Some of our knowledge is real, and some of it is just shit we made up, and we have not the slightest idea how to tell the one from the other. And the biggest problem is that we still do not know what luck is. Most of being able to tell good hockey knowledge from bad hinges on being able to tell the difference between luck and, for lack of a better word, skill- that is, the real patterns from the things that just look like patterns, or conversely, the truly random from that which looks random. Skill is the things we know, the things that can be controlled or at least impacted by human agency- the skill of GMs building teams, the skill of coaches in creating strategy, the skill of players in executing it (N.B.: Obviously a certain amount of player-skill is luck at bottom, in that one is lucky to be born with whatever mental or physical capacities can be grown into hockey genius, but for now we’ll consider that, broadly speaking, players have a large degree of control over how they play.)

Luck is the catch-all term for everything else- the uncontrollable, the external, the not human. The hockey gods. The forces of nature. Something out there that determines the things that we do not. Somewhere out there, there is a line separating skill from luck, the controllable from the uncontrolled, but it’s a difficult boundary, a long invisible line across rough country, you can walk for days and never be sure exactly when you’ve crossed it. At one point- say, the point where a player is able to take a wrist shot- is definitely in the land of skill. Another point- the point where a puck bounces erratically off the back boards- is definitely within the realm of luck. But everything else about this shot- where the shot is taken from, it’s accuracy, whether it is deflected or blocked, whether the goalie makes the save, where the rebound ends up and who gets to it first- is somewhere in the middle. Depending on who you ask and how they parse their evidence, it might be either.

Players and coaches define luck narrowly. To them, luck is things that are uncaused and, generally, uncausing. Flukes. Bounces. Momentum. These are the names of the constellations. They are the ineffable mysteries of the game. They are powerful and strange and not to be disrespected, but they are also a comparatively small percentage of all hockey events. Them that play have a very high estimation of their own power over the game- most of what happens, they believe, is somebody’s fault. It is, after all, their hands that hold the sticks.

When a statistician speaks of luck, the feeling of agency is irrelevant. For them, luck is closely allied to random, it is the very meaningless happening that describes why one random option occurred over another. Anything that is not consistently replicable over a large sample size; anything that does not, over the very long term, differ significantly in its incidence from the theoretical coin flip- that is all luck.

One sees the difference, yes? The latter definition is far broader than the former. It’s the difference between presumption of agency, according to the old wisdom, and the presumption of luck, according to the new. A good many things that players would consider controllable- like, say, whether or not their team is going to score a goal this period- are, statistically speaking, apparently random.

The strange thing is that the two may not be mutually exclusive. Consider another example. Presume you sit on the top of a very tall building, watching people come out the front door. Presume that you know, from previous close observation, that half the people who come out during this particular hour of the day turn right, and the other half turn left. From your perspective, whether any individual person will go right or left is entirely random. You have no way of knowing. You have no way of predicting. If you make a bet with your friend over the turning of a particular individual, you might say the winning or the losing was a matter of luck. But if I am that person, coming out of the building, for me there is no luck in my own movements. I am going somewhere particular, I know exactly which way I mean to turn and do so. For me, there is no 50% chance that I will go the other way. Perhaps there is some small chance, if something unexpected happens- a call on the cell phone, a road block down the street- but even that won’t be random.

It seems, sometimes, as though the choice presented is leaden and binary: either everything is luck or nothing is. A lot of statistical work finds many aspects of what we conventionally consider skill to have tiny, marginal effects or no discernable effect whatsoever- the best player is only 4-5 wins above the worst, the best goalie even less, shooting percentage is an inhuman force of nature, shot quality is non-existent. But the conventional wisdom will make up explanations for every damn thing, as if the whole game was foreordained had we only but eyes to see. Believe the experts and there is not one single game won or lost on luck in an entire season.

The problem with the experiential definition of luck is that it overestimates human power and credits any pattern that seems to appear with causal influence. Taken to its logical extreme, it becomes little more than superstition. It is not very good at skepticism, and it routinely refuses to consider random as a possibility. The problem with the statistical definition, at this point anyway, is that it’s prematurely apocalyptic in its application. Once something is declared luck, is determined to appear random on a scatterplot, it is the end of the discussion, it is the final word. Random is random, and can be analyzed no further.

The stars are not random because they look random. People’s motions are not unpredictable just because you cannot, at this time, figure out how to predict them. The constellations and stories maybe the wrong interpretive framework- just as clutch or shot quality may be- but it doesn’t mean that the phenomena they explain do not exist, or have no cause or reason whatsoever.

Astronomy emerged from astrology not by means of large-scale analyses, but rather by small-scale ones. It was when people stopped fixating on the whole sky and it’s wide, senseless reaches and began to zoom in- literally and figuratively- on specific bodies and events that truer, clearer, stronger explanations began to emerge. The intense study of one or two or three planets gave us the principles of planetary motion, even though the behavior of those planets is additionally impacted by a hundred individual variables.

Hockey events are difficult to understand because of their very ephemerality. A star will keep doing it’s own highly unique thing, day after day, month after month, year after year, for a very long time. There is time for observation and documentation and analysis. A goal happens once and is finished and no goal exactly like it will ever come again, so it is difficult to sift the causal from the incidental. But the distinction is still there.

Whether it is possible for hockey knowledge to get any better, I don’t know. I don’t know if it’s important that it get any better, frankly. Like the study of the heavens, after we gave up on the idea that they controlled our fate, the only reasons for a serious analysis of hockey are curiosity and exploration and knowing for the sake of knowing. Unlike the stars, however, hockey is a man-made puzzle, and as such is somewhat less fascinating. But if it is interesting, it is for exactly this reason- because in its raw unpredictability, it takes us right up to the threshold of human agency. It puts our bodies and our wills flat up against the hard intractability of random. It is a remarkable case study of these concepts of choice and luck, chaos and control. Two things I believe: Firstly, the best, most useful, accurate hockey knowledge is just as likely to come from the close investigation of small, individual events and figures as from huge data sets, undertaken in a spirit of respect for the received wisdom but not deference to it. Secondly, the single most difficult and significant problem facing any further study of the game is the cartography of luck.

(Part 1 of I-have-no-idea-how-many. This ain’t gonna be over for a long while.)


MathMan said...

Fantastic post. I have thought about the difference between luck-as-physical-chance and luck-as-statistical-randomness, and how they overlap, before but could never express it this nicely. Likewise, talent-as-technical-ability and talent-as-affecting-the-result.

Until the hockey grognards and the statisticians use the same definitions -- or understand that they don't! -- they'll never see eye to eye, especially since "luck" is a dirty word in sports circles!

greywall said...

May I ask you out on a date? If you're ever in Vancouver and are looking for a good time...

Jonathan Willis said...

I really enjoyed reading that.

E said...

mathman- a professional poker player told me once that, in his field, the preferred term for luck in the statistical sense is now 'variance'. i wonder if it's possible for an alternate naming to catch on in this context, although it seems to me as though some statisticians are very invested in the term 'luck', precisely because they want to wean people away from what they see as an illusion (or delusion) of agency.

greywall- i'm charmed, but unfortunately, you're way, way too late. i wouldn't mind going out for a platonic drink, though, if i'm ever in the region.

mr. willis- thank you. it means a lot, particularly considering i was anxious as hell about this one.

Lowetide said...

You are such a splendid writer and even more impressive you understand hockey and bring clarity to the conversation.

I'd buy you a beer but you're not in the neighbourhood. :-)

greywall said...

sigh... i suppose its that "Julian" character isn't it?

well, a platonic beer is better than no beer. still, hands of Mr. Lowetide. I asked her first. what game are you trying to play anyway?

btw... new favorite theoryofice post.

Anonymous said...

Wow. Hockey as a muse: no damn kidding! You've been pulling veils layer by layer from over our hockey watching eyes, over our very act of watching now, in this ongoing investigation of yours.
I'm enjoying it very much.
With the underlying fear that the muse might at some point be deciphered, named, crystallized. Or something.
But don't stop.

Anonymous said...

I meant to sign L. not E.

E said...

greywall, mr. lowetide can say or do anything he wants on any hockey blog on the internet and we will thank him for the privilage. and i am more than capable of having different drinks with different people in different parts of canada at different points in the vague future. but not beer, it would only embarrass me.

and, incidentally, i don't know hockey all that well, i just have a lot of hypotheses about it. which is probably better in a way, since (as you point out), certainty is generally unjustified and often dull.

Anonymous said...

There is an ocean's worth of hockey knowledge constantly offered out there and (i-uncultured craftsman that I am) I shudder to think of whatever else is uttered and shared on the web at large in the way of learned opinions and substantial writing. I don't think it's about knowledge, although you've obviously been a quick study. I suppose it has also to do with your training (have you yet revealed what it is that you do?), but your own talent in analyzing the wealth of impressions/perceptions you get from the game (+ the hard work you surely invest in it) makes for some enlightening reading for this (recent, tries not to be sycophantic) commenter.

Scott Reynolds said...

A fantastic post E. Thank you so much for writing.

Snap Wilson said...

Very much looking forward to the rest of these. I also have the urge to share your definition of "professionalism" with some of my colleagues.

Alice said...

Splendid, thank-you for finding the groove again lately, we hang onto your craft out here in the un-parsed wasteland.

Btw, I am not your original 'alice', I think the first time I posted here I decided to mix it up with RO, but looking back in your archives - Balance might be my favourite - I have found a previous iteration. Just so you don't attach my ramblings to your imagined incarnation of that original alice soul. :-)

To further your confusion I come from the male side of the spectrum, in order to set up the next bit correctly:

I love women's hockey, and I think you should write a book about it.

Embed yourself with the US national team, hang with them like no male journalist ever could, find out how they - too American! Too female! - came to love and devote themselves to the game in such a different way than you came to it. Bench. Dressing room. Practice. These could all be accessible to you.

This occurred to me at the cottage, reading the strangely artless 'Gretzky's Tears' and thinking if Brunt can throw that out there, then there's some room at the top of that food chain. You are Roger Angell-good. And that's small company.

My boys retired from the game after a bit of Bantam. The girls, Intermediate (17) and Peewee (12) still play. Once you get used to people going into a corner and not throwing it away out of fear of being crushed, it's the game and it's joy. Think about it, and whatever you do throw your dart high - you're a terrific writer.