Tarasov

There is an idea- or perhaps nothing so fully formed as an idea, but a sense or an impression- that advanced statistics are a new thing in hockey. The debate between quantitative and qualitative methodologies is often framed, in a narrative that goes back at least to Moneyball, as one of the new vs. the old, a bright revolution of young radicals setting fire to the staid customs of their fathers’ age.

This is both true and untrue. True, in that before recent years there existed neither the data nor the technology necessary to do the kind of work currently being done. No matter what one’s values or interests, it simply wasn’t possible in the 1970s to count all shots directed at net for every team in the NHL, then to aggregate and distribute that data widely among many different thinkers, all with extensive computing resources at their disposal. The kind of large-scale, league-wide, multi-year analyses that constitute the meat and muscle of contemporary fancystats are a product of the internet as much as of a new ideology.

But despite their limitations, there were plenty of premodern hockey men who experimented with using quantitative methods to get beneath the skin of the game. Conn Smythe, who was so traditional that his views virtually define tradition in Canadian hockey, who believed hard in good bloodlines and beating people in alleys, recorded all Leafs games on film and rewatched the footage, noting who was on the ice for which types of events in pursuit of objective data about quality of competition. Roger Neilson, dissatisfied with shots on net as a measure of team offense, kept his own count of on-ice scoring chances. And, of course, the metric for shots directed at net is named Corsi for a reason. While technological limitations made it nearly impossible for early innovators to do work on the advanced statistics of the NHL as a whole, it is clear that within franchises, some GMs and coaches have been pursuing fancystats for several decades.

In fact, it’s remarkable how much early NHL stats experimentation lines up, in spirit if not in method, with the work presently being done. They share a common focus on shot counts, scoring percentages, and quality of ice time. They share a common fixation on the front of the net and the direct showdown between shooter and goalie, and a common fascination with shaking out which players are doing easier or harder work. As much as the advanced stats community contrasts itself with the values and practices of “traditional” Canadian hockey thought, it also shares a lineage with those traditions. The ways of looking at the game, breaking down its events, and chasing its potential are not so different.

For a more intense contrast, consider Anatoli Tarasov. Called “the father of Russian hockey”, Tarasov founded the famous Red Army team and ran it for nearly 30 years, as well as running the Soviet national team from 1958-1972, a period in which it came to dominate international competition, including nine consecutive World Championship and Olympic victories from 1963-1971. To appreciate the scale of this achievement, you have to remember that there was no ice hockey in Russia until after World War II, and that Tarasov built this program with no support from any existing hockey power. What Anatoli did, basically, is invent a completely new theory of hockey out of his very own head with nothing but a few textbooks, and his theory was so f*&king brilliant that it took Russia from “country with no hockey players and no indoor ice rinks” in 1946 to “second best hockey country on the planet” by 1972. There are not enough italics in the world to sufficiently emphasize how brilliant this guy was.  Ironically, for a Soviet coach who openly scorned the individualism of the Canadian style, Tarasov has a fair claim to being the greatest individual genius in the history of the sport.

Tarasov’s systems and tactics proceeded from the theory that it was impossible to beat Canadian hockey at its own style of play. Canada, he acknowledged, was the greatest hockey country in the world; however, he also felt that Canada’s way of playing was so peculiar, indeed so culturally specific, that no other country could ever successfully adopt it. There’s no way to play the Canadian way better than Canada does, he argued, so if another country aspires to actually challenge Canadian supremacy, it must have its own style of hockey. And so the Russian style was born, as an explicit contrast to the Canadian game.

Like Smythe, like Nielson, like Corsi, Tarasov counted microstats for his teams. However, because he was trying to develop a different way of playing hockey, he valued different things and counted different things. I don’t have access, sadly, to much of his actual data- presumably it survives in Russia, in an archive I will never be able to visit, in a language I will never be able to read. But some of it is in his book Road to Olympus, which was translated into English and published in Canada in 1969, so although we cannot fully assess his use of stats, we at least know something about what he was counting and why. I don’t know if his way of breaking down a hockey game is better or worse than the Canadian ways, but it certainly is distinctive, and as such provides a stunning example of how cultural context influences even “objective” thinking.

Firstly, Tarasov was positively obsessed with passing. While he did count shots and track shooting percentages, he throws that data out almost incidentally, with little explanation, as though it was too dull to be worth discussing. But his pass counts, my God, the man can go on about pass counts for paragraphs, with so much evident pride that you can virtually see his chest puffing out of the page. Whenever he wants to make a very strong point about Soviet dominance, he goes to the pass counts, noting happily that the USSR outpasses the Canadians by a ratio of nearly 2:1, and other European countries almost as much.

Canadian coaches seem to have found this emphasis on passing to be irrational, and indeed almost comical, but there was a method to it. Canadian players, Tarasov said, are the best in the world at one-on-one battles. At the beginning, none his players could go up against a high-level Canadian individually and come out with the puck. The solution, then, was to make sure that one-on-one situations were as rare as possible. By passing constantly, even when it was apparently unnecessary, Tarasov found that he could sustain team possession for much longer than by relying on individual puck carriers.  Moreover, the constant shifting of geometry confused and exhausted opponents from any country, as they found themselves chasing the puck from one angle to another. He trained his players to pass creatively, abruptly, and frequently, and judged the recipient of a pass to be just as responsible for its quality as the player who made it. When one player has the puck, he said, the other four should be trying to “mastermind” different possible passes for him, positioning themselves to provide more good options than the opposition can properly cover. It makes sense, intuitively: change the angles often enough and the lanes will appear. When he found that his high pass counts correlated with his victories, he made the metric an essential part of his policy.

Another interesting stat he used was something he termed “attack percentage”, which seems to have been the percentage of offensive zone entries that result in a scoring chance (although I’m not sure how closely his definition of a scoring chance cleaves to the contemporary one). Tarasov had no patience for dump-ins, a play he considered boring, dull-witted, and ineffective. His ideal was for his forwards to carry the play fully from the blue line to the goal line, and he noted their success or failure at this carefully. Speaking of the 1967 World Championships, he points out that Soviet forwards had a 59% attack percentage, comparative to 40% collectively for their opponents. The corresponding defensive stat he calls “counterattack percentage”, and measures successful breakouts by counting how many times a defenseman gets the puck in his own end and how many of those times lead to an offensive zone entry at the other end. In this, the Soviet team was even more dominant, with a 72% success rate versus the opposition’s 46%.

Tarasov seems to have been instinctively suspicious of both shooting and scoring as measures of sustainable talent in hockey- perhaps he was too put off by the evident randomness. These kind of zone-based metrics were his way of trying to measure process rather than product, valuing the more trainable skills of getting into scoring position rather than the less reliable skill of scoring. Using these percentages was also a way of refocusing his players’ attention and self-evaluation. Rather than seeing their success in their goals and points, he encouraged them to see it in their good passes and effective zone entries.

Finally, Tarasov also seems to have had a preoccupation with rebounds, both recovering them and, even more importantly, re-shooting them. A good goalie, he figured, will make the first save, but the second one- particularly if it comes quickly and from a different angle- is much harder to stop. Thus, he tracked both the percentage of rebounds his team was able to capture and the percentage of shots they were able to get from rebounds, and speaks excitedly of the first time the USSR outrebounded Canada in international competition. As much as he explicitly dissociated his hockey philosophy from the just shoot attitude he associated with “The Land of the Maple Leaf” (one of his favorite terms for Canada), and was careful not to overvalue shots in themselves, he insisted on making the pursuit of rebound shots a fixture of his teams’ strategies.

All this leads to an obvious question: were those strategies correct? Was Tarasov counting the right things? There isn’t much evidence that he was counting as part of an empirical program of hockey development- that is, he didn’t choose these things after establishing first that they were necessary for winning in hockey generally, but decided a priori that they were necessary for the way his team was going to win. His system was first and foremost an artistic creation, the work of a mad genius born from an eccentric mush of Soviet ideals, quixotic obsessions, old books, and long experience.  He counted his microstats mostly to track effectively how well his teams were cleaving to his ideals, not as a way of evaluating those ideals.

That said, his ideals were the foundation for a dominant league team, a dominant national team, and transformed Russia from hockey backwater to hockey powerhouse in less than two decades. Within their cultural and historical context, his practices and the data he used to maintain them were devastatingly effective. Although his stats never really caught on in North America (although some of the recent zone entry work is reminiscent of them), many his principles have been so widely adopted in the NHL that they’re now indistinguishable from Good Canadian Hockey. And I wouldn’t be surprised if some good Canadian hockey managers have experimented with counting, in a modernized way, some of the same things Tarasov did.

[This will, insha'allah, be the first in a series of posts considering Tarasov's coaching theories and practices, as reflected in his book Road to Olympus, which I highly recommend to all y'all. To be continued...]