How many Henry Fords are there today?
I have always wondered how great business managers of the past, like Jack Welch, Michael Eisner, or even further back all the way to Henry Ford, would compare to today’s top business talent. But how can we make such a comparison? A recent paper on top chess players has given me a couple of ideas on how to approach this question.
To begin with, a much easier question to answer is: “are sprinters getting better with time”. The answer now is simple: Yes, Bob Hayes was fast in 100m in 1964, but if he were to compete with a 2009 Usain Bolt, Bolt would finish first and still have half a second to turn his head around and wink.
Unfortunately, in tournament sports like business, objective measures of ability do not exist. Management is an immensely complicated and multi-faceted endeavor and, in the end, all it matters is how the firm performs vis-à-vis its competitors.
What can we learn from chess?
Initially it would seem that the same would be true for chess. How can you possibly compare Kasparov with Kapablanca from half a century earlier, or a brilliant young Bobby Fischer in 1960’s with today’s prodigy Magnus Carlsson? Chess buffs do have some aids to make such comparisons, but they have primarily relied on the Elo rating system devised by Arpad Elo to measure a chess player’s ability (and, according to the movie “The Social Network”, to rate “hotness” among Harvard female undergrads by pre-Facebook Zuckerberg and Saverin). According to the Elo rating system, while Kasparov is (as of late 2011) the highest rated player of all time, only 3 of the top 20 Elo rated players of all time, peaked before 2000. The 17 remaining top rated players are, more or less, active today.
However, even the Elo system relies on the performance of a player against others and, since it does not measure chess ability against an objective metric, is susceptible to inflation or deflation. Just as you cannot compare an A+ student from a 1970 MIT to today’s A+ MIT students, due to grade inflation, you cannot compare a 2785 Elo Bobby Fischer from 1972, to a 2785 Elo Vassily Ivanchuk from 2007. Indeed, within the chess community there is a lot of talk about real or imaginary Elo inflation, with many claiming that there is no way that there are today so many chess players, so closely rated to Kasparov, when Kasparov was so clearly better than everyone else.
And yet, one would expect that this should exactly be the case… Today’s chess champions have at their disposal an array of training aids that would be beyond the wildest dreams of the champions of even the 1980s. Instead of a team of grandmaster-level coaches that would work with the player for hours to analyze a single position, today’s champions employ small teams of coaches armed with chess-playing computer programs (called chess engines), such as Crafty or Rybka, that can, within minutes or even seconds, calculate a hidden trap in a position more than 10 moves in advance. The faster feedback cycle leads to more positions analyzed per hour of training, which leads to more efficient training, which leads to faster improving championship-level players. So we should expect that today’s top players are better than ever. But where is the data to support this, data that are untainted by Elo inflation?
An alternative way to evaluate players is to use computers to rate players’ moves. This process works by having an engine check the moves that a player played and rate each move against the best possible move that the player could have played, given the position. Thus, the average “error” per move for each player can be calculated. For example, one engine calculated the average error of 1920s world champion Capablanca to be 0.1, which means that Capablanca was dropping a pawn every ten moves, by playing less than perfect. For comparison, a modern tournament level player rated at 2,100 Elo drops a pawn every four moves, against a perfect opponent.
Today’s chess champions have at their disposal an array of training aids that would be beyond the wildest dreams of the champions of even the 1980s.
Of course, this requires engines that play better than the 2800+ Elo players that they try to rate and this has not yet been achieved: a 2006 effort by two computer scientists used a chess engine (Crafty) that was then rated at 2650 Elo. In 2011, K.W. Regan, a computer scientist at the University of Buffalo, and Guy Haworth, a systems engineer from the University of Reading used an improved Rybka engine at the 2700 Elo level (pdf here). This is still shy of the 3000+ Elo that would give the engine, in many critics minds, the undeniable authority to judge Kasparov’s game. Note that the Rybka engine is indeed rated 3000+ Elo, but Regan & Haworth had to process all the moves in 6,000 games to do their experiment and decided to sacrifice engine strength for speed of computation (even at this much-lowered rating, it would still take nearly 5 years for a single processor core to analyze the data set). Let’s then treat the Regan & Haworth results as provisional, but let’s also be content in our certainty that Moore’s Law assures us that we will know with certainty in the near future.
What do the engines tell us then? According to the Regan & Haworth paper, not only there is no Elo inflation today, but, if anything, there has been deflation in top players’ ratings, as many, equally rated top players drag each others’ Elo ratings down. In the past decade, more than thirty players have exceeded the 2700 Elo rating that would make them formidable opponents for Kasparov. Carlsson, currently rated at more than 2800, is only twenty-two years old…
What about business then?
But what does that tell us about the quality of today’s top business leaders? I will not claim that business management resembles the game of chess, but I will claim that today’s top business leaders “train” better than ever. The reason is the faster feedback cycle that managers enjoy, due to the widespread use of Information Technology on the business side and the proliferation of social media and Web2.0 tools in general, on the consumer side. Indeed, today’s C-level executives get to learn the outcomes of their actions much faster:
- it took Palm only nine months to develop WebOS, which was a full-fledged smart-phone & tablet operating system
- it took Hewlett Packard months – not years – to learn that their tablet was not good enough to compete with Apple. The project was killed only seven weeks after product release
- it took video renting/ streaming service Netflix a few days to learn that customers hate the company’s proposed split. Netflix was able to reverse the decision (but only soften the damage done) within weeks
- it took just one day for US telecom provider Verizon to reverse its proposed $2 “convenience fee” for customers who chose to pay their bills by phone or online, after consumer outrage
- it takes clothing retailer Inditex just days to learn that a new line of clothes does not resonate with customers
- it takes hours for social gaming giant Zynga to learn how a new game feature works out in practice
Today’s top business leaders “train” better than ever.
It is thus not surprising that boards demand that top leaders learn and improve faster than ever before and that the average CEO tenure is on the fall. With such fast paced feedback cycle, we should expect that top business talent has never been better that it is today, but that it is also much harder for these leaders to stand out among equally talented business competitors.
So there you have it. I would guess that there are many Henry Fords today, working late into the night, trying to bring about the next business revolution, but we will not know for certain until programs are smart enough to be able to objectively evaluate managerial decisions, with the same authority that they can rate chess moves. Only then can we learn who these extremely talented people are. At that point, I’m afraid that the computer will tell us that the question is no longer relevant.