, , ,

As promised, the second half of my review of the Signal and the Noise, by Nate Silver.

Premise: This is the second half of the Signal and the Noise by Nate Silver. Here he discusses how statistics and probability can be effected and our own biases improved. Sometimes by acknowledging them. Sometimes by knowing the limits of the questions we are asking

Historical Significance: Previously covered in part one.


  • Rage Against the Machine, Chapter 09 – Chess and using it to understand the constraints of man processing ability. We know all the rules and all the outcomes. Compare to weather for data determinism. Computers can see far ahead of us but they aren’t perfect. There are almost always tweaks to be made and errors that occur. The problem for humans is seeing that these are errors and not assuming the computer must inherently be smarter than us. That’s the signal in the noise of precision. Basically, we can’t assume the computers know best and live in fear of that. Human instinct and ingenuity still plays some role.
  • The Poker Bubble, Chapter 10 – a murky meeting between the signal and noise. purpose of chapter. poker is skill and luck. shows how like baseball, odds play out long-term. great example of over confidence when data suggest otherwise The key is recognizing your own weaknesses when attributing the data points. This is over attributing the signal portion, instead of the noise. Thinking it’s bigger than  it is

Basically, poker is about results verses data and the volatility that ensues. Poker shows that as long as you’ve read the data correctly, even losing shouldn’t bother us in the long-term. What’s the point? I guess that signal and noise are always present. There’s no way to really remove either, just understand what they are and what they represent in context.

  • If You Can’t Beat ‘Em, Chapter 11 – Upon reexamination of the stock market I find it harder to believe in its existence. The inherent instability will even out over time. At least according to the data available. For me, I agree with Silver’s assessment of the stock market, in that Winston Churchill was correct about democracy. It’s the worst system invested, except for all the rest. I find the market less reliable now than I did when I started reading. After all, if human nature pushes to do exactly the opposite of what’s best, how is this the best system we could come up with?

Still,  his assessment that signal and noise are intrinsically linked is appealing. Using the stock market, he is able to demonstrate that each will balance out the other, and that our perception of signal and noise influences each in turn.

  • A Climate of Healthy Skepticism, Chapter 12 – The most important note, I think, that can be said here is the problem with projected analysis. What I mean by that is, we have a hard enough time predicting the weather in five days, let along five or five hundred years. As with earthquakes, out ideas about those models is based on past events and cannot be said to have widely held predictive accuracy.

This isn’t to say the data is wrong, just the further out one goes, the harder it is to distinguish what you’re looking for from noise. Long term patterns are a nice starting point, but don’t hold enough specific data when interests are rooted in quarter-to-quarter, year to year results. Said another way: uncertainty is certain. Dealing with that uncertainty is the problem.

I’m not really sure what to take from this chapter other than an argument should be made for the specificity of word selection, but I think I’ll save that discussion until Friday.

  • What You Don’t Know can Hurt You, Chapter 13 –  Most people act like they’ve read the data when they really haven’t. TV pundits are a close approximation, just saying things to make headlines. Conspiracy theory is just lazy data analysis.

Here’s the real problem: hindsight is perfect. Looking back, it’s easy for everyone to claim that Pearl Harbor could have been avoided or that a 5 run inning would have been stopped if the pitcher was taken out. Except, it’s almost impossible, in real-time, to distinguish the randomness of the universe from the facts and path we’re looking for.

Also of utmost importance is the idea that unfamiliar is not improbable. Just because 9/11 was not expected or Bobby Fischer’s sacrifice against Donald Byrne had never been considered, does not make them improbable. Their probability does not decrease just because they were not considered, that stays the same.

There is a difference between not knowing something and uncertainty. Uncertainty is knowing you are unaware of a particular fact. And, by definition, not knowing you are unaware means you can’t know. This is an important distinction.

  • Conclusion – Our bias is to think we know more than we do. Using data is the most effective way to see the bigger picture of sports or politics, but even then we need to acknowledge of biases.

The accumulation and probability must start with the belief that something, no matter what is it, holds true. Information is only knowledge in context and that context can be distorted. This awareness helps us understand data. Despite all the data in the world, in fact, because of it, there is more noise for the same amount of signal. This might not be a good thing

Recommendation: Verdict? Absolutely I recommend this.

Sometimes the water is over my head, sometimes it is exactly at my head,  but Silver always uses analogies that I can relate to. Both my parents are lawyers. The finer details of what they do in a day I couldn’t begin to comprehend. But my father always took the time to make sure complex concepts of law were understandable when I did ask. That’s the highest compliment I think I can pay this book. That the theories of data analysis, why important to our lives, are presented in a straightforward, well-reasoned structure that I can understand.

What I can say for sure, is that like Money Ballcritics of Silver are missing the point. Money Ball wasn’t any more about small market teams competing with big  market teams than the Signal and the Noise was about definitely knowing weather patterns or climate change. Climate prediction or baseball or the housing bubble: these are all just repeated examples of distinguishing signal from noise. Arguing whether or not his findings are substantiated or even correct is a faulty case. He even says no one can remove bias or be correct 100% of the time. Rather that our current methodology for approaching that assessment is flawed and should, as a society, be reevaluated. Arguing his answers were wrong is immaterial if the methodology, his approach, is correct.

And that’s his point. The true signal must be distinguished from the heavy noise that hides it, even if sometimes Silver himself is the noise.

Next week I’ll be reviewing; A Prayer for the Dying, by Stewart O’ Nan.