Tuesday, November 13, 2012


Over the weekend I saw Skyfall, one of the best ever entries in the 50 year old James Bond series. (50!)  And I think I've seen all of them, so I know what I'm talking about.

This movie has as much relevance to our current political situation as most  anything you'll hear on cable news this week. Seriously, check out the new Q, played by Ben Whishaw, who represents the triumph of serious computer knowledge over the silly toys offered Bond by previous Qs. Notice the old-fashioned action-hero spy standing in the background waiting for the analytics to come in.

In politics, the old-fashioned operatives are also taking a back seat to the new breed of science geeks. The LA Times ran a tribute to the computer geniuses who provided sophisticated metrics for the Obama campaign. Here's Daniel Wagner, age 29 (about the same age as Ben Whishaw), the Q of the Obama campaign, leading a staff of more than 50 (50!) who crunched the numbers that helped the Obama team win. Apparently they have been working in secret throughout the campaign in a place they called the cave, and can only show themselves now that the election is over.

photo: John J. Kim, Chicago Tribune
These are the faces of the future. If this year's election represented anything, it represented the ultimate revenge of the nerds. Look how shocked the true right wing believers were when the election went exactly as predicted by the number-crunching whiz kids like Nate Silver, who correctly called 50 out of 50 state results. (50!) No more can politicos rely on their gut instincts and feelings. Politics might finally be a science now. And you can't argue with math.


  1. You can't argue with math that uses good data input. For example, the normalized power (wattage) for a cyclist who has a power to weight ratio of 4.2W/kg, weighing 185lbs and with a coefficient of drag of (CDA=0.4m^2) is 4 hours and 4 minutes on a flat course. But the 4:04 assumes the numbers input are correct.

    My question, given that Silver's projections were spot on, what did he input?

    I agree with you, Skyfall was tremendous, the best, and I have seen them all. Craig is a bad ass.

  2. That's an interesting question, because we know that analysts like Nate Silver input some bad data. My understanding is that he uses some kind of weighted averaging of various polls, and somehow has to take into account that some polls are more reliable than others. But it's even more complicated than that I'm sure.

  3. We now use maths to measure training loads, predict physiological response, fatigue and to help predict performance. Essentially Performance = Fitness - Fatigue.

    So we have to be able to measure these variables:

    How to accurately measure training loads? Let’s say we do 2 rides – one that is completely even paced (on a trainer or velodrome) such that if we took a random sample at the beginning, the middle and the end, they would all read 200 watts. The other ride is an increasing power ride where we focus on being strong at the end. We ride the first 1/3 at 100W (watts), the second 1/3 at 200W and the last third at 300W.

    Average power for both rides is identical:
    (200+200+200)/3 = 200
    (100+200+300)/3 = 200

    However, from a physiological perspective, they are quite different, with the first being much more ‘pleasant’ For this reason, Dr. Andy Coggan came up with a formula to weight this variability according to its physiological difficulty. In this case, samples are raised to the 4th power, an average is taken and then the fourth root is taken of that. From our example above:

    (200^4+200^4+200^4)/3 = (1600000000+1600000000+1600000000)/3 = 4800000000/3 = 1600000000

    Then taking the 4th root of 1600000000 = 200NP (the same as the average power)

    However, in the second scenario:

    (100^4+200^4+300^4)/3 = (100000000 + 1600000000 + 8100000000) = 9800000000/3 = 3266666667

    Then taking the 4th root of 3266666667 = 239NP (39 watts greater than the average power).

    The 4th power curve looks a lot like a lactate curve. This general trend of physiological effort, measured by things like blood lactate increases exponentially with increasing workload. This is the very concept behind Normalized Power – if you put out double the wattage, say go from 200-400W, anyone who has trained with power can attest that it is a whole lot more than twice as hard. Furthermore, when an average athlete jumps from 200-400W, the time to exhaustion is significantly more than halved.

    I suspect Silver weights the polls and other data using algorithms in a similar way.