I wouldn't mind a NFL post every week or two. What bothers me is the accuracy of the numbers. Even if you knew based on a lot of data that your team converted 4 and 1 65% of the time, is that for this exact offense line? This line that just played hard for 59 minutes? This weather? What about the defense? It's really more like 65% plus or minus 20%.
Right, and Olsen even admitted that he was taking context into account. "You’ve been good all day with Kenneth Gainwell running the ball behind this offensive line and the big tackles."
The analytics people love to tell you that their calculators take strength of opponent into account, but never tell you how much this widens their error bars.
Isn't this simply an argument in favor of no-regret dynamics? Their fundamental principle is that you give past performance some weight, but not overwhelming weight.
In politics I think this is achieved through party systems. The electorate of a party has varying loyalty, so performance affects some more than others, and the overall effect over many elections is a no-regret dynamic.
One nice thing about no-regret dynamics is that they seem to work even if outcome is very noisy signal for the quality of decisions.
When politics become too personalized, and long-term institutions (such as parties) no longer matter, I think the system malfunctions.
Very nice. Your piece resonates with the John Ganz piece which made the rounds about opinion polling. And the great Shaw quote, "The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man." Decision theory doesn't model the unreasonable man...
Really nice (though you did almost lose me with the football details)! Bashing quantitative decision theory because the real world is more complex than card games or analytic models is a popular pastime in ecology. I'm really curious what you might make of the arguments from 'resilience alliance' folks like https://doi.org/10.1016/j.tree.2011.04.007 or https://doi.org/10.1016/j.tree.2009.03.020 ... I suspect some of it will resonate and some won't.
I wouldn't mind a NFL post every week or two. What bothers me is the accuracy of the numbers. Even if you knew based on a lot of data that your team converted 4 and 1 65% of the time, is that for this exact offense line? This line that just played hard for 59 minutes? This weather? What about the defense? It's really more like 65% plus or minus 20%.
Right, and Olsen even admitted that he was taking context into account. "You’ve been good all day with Kenneth Gainwell running the ball behind this offensive line and the big tackles."
The analytics people love to tell you that their calculators take strength of opponent into account, but never tell you how much this widens their error bars.
Isn't this simply an argument in favor of no-regret dynamics? Their fundamental principle is that you give past performance some weight, but not overwhelming weight.
In politics I think this is achieved through party systems. The electorate of a party has varying loyalty, so performance affects some more than others, and the overall effect over many elections is a no-regret dynamic.
One nice thing about no-regret dynamics is that they seem to work even if outcome is very noisy signal for the quality of decisions.
When politics become too personalized, and long-term institutions (such as parties) no longer matter, I think the system malfunctions.
But I think my government's systems *are* malfunctioning. I certainly have regrets.
Very nice. Your piece resonates with the John Ganz piece which made the rounds about opinion polling. And the great Shaw quote, "The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man." Decision theory doesn't model the unreasonable man...
Totally. I should have linked to Ganz, who coined the phrase "vulgar positivism."
Really nice (though you did almost lose me with the football details)! Bashing quantitative decision theory because the real world is more complex than card games or analytic models is a popular pastime in ecology. I'm really curious what you might make of the arguments from 'resilience alliance' folks like https://doi.org/10.1016/j.tree.2011.04.007 or https://doi.org/10.1016/j.tree.2009.03.020 ... I suspect some of it will resonate and some won't.
I'll give these a read and report back!