Just a lie under oath by Exponent that has been bothering me for a while.
Dean Snyder (NFLPA statistical expert) makes several points regarding the use of the Colts' balls as a "control" group to test the difference between their level of inflation and the Patriots' balls. Essentially the point is that the Colts' balls make for a terrible control group because, among other things, the balls were not tested at the same time, the balls may not have been equally wet, and the sample size of Colts balls is only 4. All of this renders the Exponent analysis extremely suspect.
Here's just one bit of that when he talks about the timing aspect. As I said, there's a lot more. It's the crux of his argument.
From the Snyder Direct:
Q. What did you conclude about Exponent's difference in differences work?
A. Well, it's wrong. It goes back to their basic theory, the basic idea that the Colts' balls a control. If you want -- and I understand the idea of using the Colts' balls as a control, but they have to be a good control. If they are a good control, then you can isolate on whether the question of whether the additional pressure drop exhibited by the Patriots' footballs is or is not likely to have occurred by chance.
Q. And what was your conclusion as to whether 24 the Colts' balls served as good controls in their analysis?
A. They didn't, because I mentioned this earlier. There was a sequence of events at halftime. And the sequence of events at halftime was that the Patriots' balls were measured first. The Colts' balls were measured second, or even later, depending on the sequence of halftime events.
And what does Exponent have to say about this when asked by Kessler? He says, "oh no, they weren't a control! What are you talking about?"
Q: Were the Colts' balls used as controls in any way with respect to the statistical significance analysis performed by Exponent?
A. Absolutely not. I don't know where that idea came from. We did not consider the Colts' balls as controls in other statistical analysis. We looked at them both equally. And looked at the variation in the data for both the Patriots and the Colts. There was no assumption about control or anything like that on the Colts' balls in our statistical analysis.
I don't know why Kessler doesn't hit him on this point. It's a blatant lie that can also speak to bias.
Now here's Ted Wells calling the Colts' balls a control in the Wells Report:
Exponent also conducted a series of experiments to evaluate the impact of environmental conditions on the air pressure of footballs. Among other things, these experiments attempted to replicate the likely conditions and circumstances on game day and the results recorded by the game officials at halftime. In these experiments, the Colts footballs and the Colts halftime measurements were used as a “control” group because there was no plausible basis on which to believe there had been tampering with the Colts balls.
And where would Wells have gotten this from? Let's turn to the Exponent section!
For the purpose of the experiments, Paul, Weiss informed Exponent that there was no plausible basis to believe that there had been tampering with the Colts footballs; therefore, the Colts footballs were used as a "control" group when evaluating and determining test parameters for the pertinent experiments. In other words, because we could reasonably assume that the Colts measurements collected at halftime on Game Day were the result only of natural causes, a combination of environmental and timing factors was identified (within the realistic ranges provided by Paul, Weiss) for the purpose of our experiments that resulted in measurements for the Colts balls that matched the Game Day measurements. Aligning our experiment in such a way confirmed that the test conditions were a good approximation of the environmental factors present on Game Day, and allowed us to concurrently assess what the Patriots measurements would be under the same conditions. We could then assess the physical plausibility of the Patriots measurements recorded on Game Day.
1. They were used as a control group.
2. Exponent lied under oath.
3. Further evidence that they're just hired expert witnesses paid to reach a conclusion and support that conclusion in court.
There's no reason to believe anything else in his testimony or his findings.