[Mod Edit] Must Read: Drew Fustin Article on Deflategate Science

[upstater1]

That seems like a key piece of information, but it's also the only time that the gauges seems to be internally inconsistent. Otherwise, the same-gauge variance would seem pretty small, less than .15 PSI. (I'm getting that because the difference between the gauges is consistently between .3 and .45 PSI, so the same-gauge error is likely less than the difference between those two.) I'd like to know more about the conditions behind those three tests, particularly the timing of them relative to each other and to when the ball was brought inside.

Honestly, the variance is the only part of the physics that I'm having trouble explaining. This article kind of hand waves it away–the author explains that variance will be higher because more time passed when the Pats' balls were being measured and it was during a time when they would have been further from equilibrium and thus gaining pressure faster, but the numbers still don't add up. Min/Max difference in the Pats' halftime readings is 1.35 PSI, which is above the effect of the transient warming curve even if you consider the entire halftime (which certainly wasn't the case).

So, yes, I get the logo vs. non-logo gauges, and I get the warming curve, and I get the difference in initial temperature assumptions. I feel safe in saying there are entirely reasonable scenarios where the average pressure of the Pats' balls can be explained perfectly well by physics. That's average, though, and the variance is harder for me to explain. The only thing I can think of that makes that work is that the initial variance was higher, but that directly challenges Anderson's statement that the initial pressure was consistently 12.5 PSI. I don't have a real problem saying that and think it's plausible, particularly since the initial pressures weren't written down; but it's a different type of argument than the rest of the physics corrections since everything else can be explained without challenging someone's testimony, only their assumptions and models.

The high variance of measurements on the intercepted ball is useful, but not entirely consistent with the measurements of the other balls and thus I'm tempted to say it's a result of something like there being a few minutes between tests of the intercepted ball rather than inaccuracy of the gauges (calibration aside).

Is there a mechanism that's been demonstrated that would account for high variance? Maybe something to do with how long the ball was in play or how wet one ball got compared to the rest?

Too many variables. Some balls were waterlogged for instance and used extensively in the field of play. others were drier and in bags.

I'd also venture to say that it's asking a bit much for someone to be so exact on the measurements pre-game. I'm sure he was aiming for 12.50 and thought that 12.35 or 12.65 was just fine. AND, if we had all the Colts balls, your question would be answered more completely. Then we'd have another set for measuring variance.

Here's something else everyone seems to miss.

The report set aside one of the 4 Colts balls because the reading with the non-logoed gauge was actual HIGHER than the reading with the logoed gauge, and it was .45 higher, while in every other time that logo gave lower readings, for the Colts AND the Patriots balls.

So the 11 Patriots balls were measured against 3 Colts balls.

 

[Tony Cahncuhd]

The cited article explains the variance very simply. The patriots balls were warming up so the first ones measured had lower pressures than the last ones. Colts balls were measured last when temps were already leveling off hence less variance.

 

[Tony Cahncuhd]

That seems like a key piece of information, but it's also the only time that the gauges seems to be internally inconsistent. Otherwise, the same-gauge variance would seem pretty small, less than .15 PSI. (I'm getting that because the difference between the gauges is consistently between .3 and .45 PSI, so the same-gauge error is likely less than the difference between those two.) I'd like to know more about the conditions behind those three tests, particularly the timing of them relative to each other and to when the ball was brought inside.

Honestly, the variance is the only part of the physics that I'm having trouble explaining. This article kind of hand waves it away–the author explains that variance will be higher because more time passed when the Pats' balls were being measured and it was during a time when they would have been further from equilibrium and thus gaining pressure faster, but the numbers still don't add up. Min/Max difference in the Pats' halftime readings is 1.35 PSI, which is above the effect of the transient warming curve even if you consider the entire halftime (which certainly wasn't the case).

So, yes, I get the logo vs. non-logo gauges, and I get the warming curve, and I get the difference in initial temperature assumptions. I feel safe in saying there are entirely reasonable scenarios where the average pressure of the Pats' balls can be explained perfectly well by physics. That's average, though, and the variance is harder for me to explain. The only thing I can think of that makes that work is that the initial variance was higher, but that directly challenges Anderson's statement that the initial pressure was consistently 12.5 PSI. I don't have a real problem saying that and think it's plausible, particularly since the initial pressures weren't written down; but it's a different type of argument than the rest of the physics corrections since everything else can be explained without challenging someone's testimony, only their assumptions and models.

The high variance of measurements on the intercepted ball is useful, but not entirely consistent with the measurements of the other balls and thus I'm tempted to say it's a result of something like there being a few minutes between tests of the intercepted ball rather than inaccuracy of the gauges (calibration aside).

Is there a mechanism that's been demonstrated that would account for high variance? Maybe something to do with how long the ball was in play or how wet one ball got compared to the rest?

Evaporative cooling temporarily lowers pressure of the wet balls beyond what they were outdoors hence increasing the variability a little further.

 

[upstater1]

Evaporative cooling temporarily lowers pressure of the wet balls beyond what they were outdoors hence increasing the variability a little further.

Which makes the intercepted ball curious and interesting since it was measured outdoors.

 

[BSR]

One thing though - Exponent did give a more reasonable explanation of using the non-logo gauge than to make the numbers work against the Pats. They said that the non-logo gauge was closer to a calibrated gauge than the logo (Wilson) gauge. And since Anderson said Pats balls measured 12.5PSI pregame right where they presumably wanted them Exponent assumes the more accurate non-logo gauge was used. BUT how does Exponent know the Pats gauge wasn't reading high too. Did Wells try to get the Pats gauge so it could be checked? Do the Pats have the gauge and does it read a few tenths of a PSI high? If so that would completely blow apart the exponent conclusions.

I noticed that too but I don't think many have seen it. It is actually a pretty good point by Exponent. Both the Colts and Patriots would have had to be using gauges that were off similar to the logo gauge or both would have had to supply balls that were under the requested specs by the same variance. Seems unlikely. Then again, since Walt's recollection is off as to which gauge he used, it could certainly mean that it was off elsewhere as well. And with such a vary low tolerance between passing and not, every little detail counts. Its way too many assumptions to really be a conclusive judgement. This is my biggest problem with the report.

 

[Palm Beach Pats Fan]

Is there a mechanism that's been demonstrated that would account for high variance? Maybe something to do with how long the ball was in play or how wet one ball got compared to the rest?

I'd bet that 3 or 4 were very waterlogged, 3 or 4 were dry and maybe even unused, and a couple were in the middle. Colts footballs were not used in the last 5 minutes except for a kneeldown, so they were baagged uo and drier.

 

[VJCPatriot]

Just read the article carefully. It explains variance well in relation to time. The balls measured later in time will vary more compared to the balls measured earlier in time. NFL halftime has 12 minutes allotted. So the very first Pats footballs vs the Colts footballs measured near the end of half time will have more variance due to time.

Also considering that water can affect PSI, and you are not going to have uniform wetness of the balls exposed to the elements, and uniformity in the amount of snaps the balls were in the water, of course you will have variance due to mother nature.

 

[jeffro]

Both the Colts and Patriots would have had to be using gauges that were off similar to the logo gauge or both would have had to supply balls that were under the requested specs by the same variance. Seems unlikely.

Really only the Pats gauge would have to be reading high. The report just assumed that the colts wanted their footballs initially at 13.1 psi. I wonder about the logo gauges though. Maybe it is something Wilson gives to the teams and possibly they all tend to read high.

 

[PATS16N0]

What can you do beyond just face palm at this point?
What a pathetic ordeal. This is Twilight Zone stupid.

 

[ChesBay]

I think all parties on either side of this investigation would agree that...at best...there's a 0.3 psig differential that can't be explained. And in the worst case, where the 0.3 differential occurred for one ball, the differential was less than 0.3 for the remaining 10 balls. So....

1) has anyone attempted to see how precisely an individual could deflate 13 balls (from a starting psia of 12.5) in increments as small as 0.1 psig?
2) furthermore, could such preciseness be uniformly replicated given 100 seconds to accomplish this feat??
3) and lastly, what would be the motivation to deflate the balls by such a small increment - i.e. to make such a practice worthy of the inherent risk?

 

[Bruins29]

This thing gets stupider by the hour as more stuff comes out. This is the most embarrassing moment of professional sports ever. This makes the Malice in the Palace look good by comparison.

 

[PATS16N0]

At least we won ring #4.

Just imagine how ****ty this would be if Brady and the Boys didn't end Seattle's repeat dreams.

HATE US CUZ YOU AIN'T US

 

[Tony2046]

Awesome work by Drew !
But why were all the Pats balls below 12.5 ? !

Here is what you have to understand.

All footballs or basketballs or volleyballs will lose pressure, not air, when moving from a warm environment to a cold environment.

Air when heated expands and when cooled contracts. That is what the ideal gas law predicts. How much will it expand and how much it will contract.

So if you take a football inflated to 12.5 psi in a 70 degree environment and put it in a 10 below zero environment the footballs will lose about 4.0 psi. They will not lose any air molecules. They will only lose pressure because those air molecules are moving slower.

If you took a balloon and put it in the freezer it would shrink in size because the air or gas inside the balloon would move slower and create less pressure. BUT if you took the balloon out of the freezer it would return to its initial size and inflation pressure.

That is the key to understanding why Brady never asked to have air removed from those footballs. The whole wells report is based on a possible 0.3 psi loss in those footballs

Hope that helpsl

 

[wahwho]

Just read the article carefully. It explains variance well in relation to time. The balls measured later in time will vary more compared to the balls measured earlier in time. NFL halftime has 12 minutes allotted. So the very first Pats footballs vs the Colts footballs measured near the end of half time will have more variance due to time.

Well, no, not really. The article provides a principle that demonstrates that some variance is to be expected, but the magnitude of the recorded variance can't be explained purely by the principle they're suggesting. The recorded variance is 1.35 PSI. If I'm really generous on the time and include both the wet and dry curve, then using Exponent's graphs I can squeeze maybe .9 PSI of variance out of the temperature change.

I have seen some people say that their wet ball experiments weren't done well though, and I agree with that. But I haven't seen anyone produce an alternative transient curve yet. (Some have suggested that the values of the curve should shift up or down based on different gauges or starting assumptions, but I haven't seen anyone suggest a curve with a significantly different shape or magnitude.)

That said, there are significant problems with how the curves were displayed in the report that skew how they are being perceived. For example, the two graphs on page 55 of the Exponent appendix show non-logo and logo graphs next to each other as if they are comparable. They aren't. In order to treat the Colts balls as a control, they changed the assumptions (pre-game temp) between creating the left and right graphs.

For the record, I don't think they did a great job of that, either: they fit the curve to the average Colts reading at a time that's likely earlier than when the actual reading took place, throwing off the midpoint of the curve. There's also a question of if the order of events was Measure Pats -> Measure Colts -> Inflate Pats or Measure Pats -> Inflate Pats -> Measure Colts; if it's the latter, then the graphs and controls are way off. (There's an assumption that it was the former throughout the document but I don't think I ever saw a direct statement that it really happened that way. Maybe I missed it, though.) Similarly, though the graph shows the entire halftime period, only minutes ~2-7 should be compared to the Patriots' average bar–including minutes 7-14 just deceptively makes the average bar appear further away from the applicable part of the curve than it really is.

But yes, I agree that both the wet/dry ball factor and the likely lack of precision in the initial measurements contribute to and may explain the high variance. I just haven't seen anything that would let me put specific numbers to either factor. Lacking that, it's hard to claim that they d0–and we're obviously never going to get an answer on the initial measurement question, so...?

 

[Marko]

What was time of possession, I would imagine the Colts balls were in the bag longer too. Got to wonder if he's a Patsfans member, not to brag but his killer without a murder reference in the texts was extremely close to one of my examples posted.
I agree with all he writes, but even if the science by exponent is right .5 psi is nothing and would be ludicrous for Mcnally to sneak around to basically go pfffft. with 12 balls.

 

[nowayback]

Here is what you have to understand.

All footballs or basketballs or volleyballs will lose pressure, not air, when moving from a warm environment to a cold environment.

Air when heated expands and when cooled contracts. That is what the ideal gas law predicts. How much will it expand and how much it will contract.

So if you take a football inflated to 12.5 psi in a 70 degree environment and put it in a 10 below zero environment the footballs will lose about 4.0 psi. They will not lose any air molecules. They will only lose pressure because those air molecules are moving slower.

If you took a balloon and put it in the freezer it would shrink in size because the air or gas inside the balloon would move slower and create less pressure. BUT if you took the balloon out of the freezer it would return to its initial size and inflation pressure.

That is the key to understanding why Brady never asked to have air removed from those footballs. The whole wells report is based on a possible 0.3 psi loss in those footballs

Hope that helpsl

Thank you but I was joking >>>

 

[Chevy]

One valuable thing that he explains very well., something that I have scratched my head about often, is the variability within the group being so so much higher for the Patriots footballs.

Sample size is surely part of it, but it always seemed not to be the whole picture.

But...

He totally nails this point in the notes section.

The great differences in the steepness of the warming curves that the two sets of footballs are on, when they were tested, explains it completely, clearly, and in bullet-proof fashion.

My God, the Princeton guy needs to read this. He'd squirm uneasily about he study to which his name became attached.

Wouldn't surface-to-mass-ratio come in to play as well? I would assume that the Pat's balls (12) were in a bag - bunched up with those under the first layer more insulated from the ambient room temperature and therefore even slower in the climb to equilibrium. Meanwhile, the Colt's balls (12 - bagged the same way) sat longer, and only those from the outer layer were checked (most likely since they only checked 4).

 

[Tony2046]

http://drewfustin.com/deflategate/

First article I've read that hits all 3 of the big science/logic failures in the Wells Report and gets really into the science in an approachable, easy to understand way.

I'm thinking this Dr Drew Fustin and our Palm Beach Pats Fan could get to be good friends.

One thing that blows my mind is that if you have evidence of cheating why do you destroy it? They say the footballs were under inflated at half time and they re-inflate them?? Why not keep them as proof? That is what leads me to believe that this was more than a ignorant mis calculation. In essence they are trying to destroy Brady's legacy over 0.3 psi or whatever. It definitely was not the original 2 psi as initially reported or leaked. Anyways I do not hear a lot about the evidence being mishandled. The NFL botches these investigations time and time again and still claims it is all in the name of integrity.

 

[Tony2046]

Thank you but I was joking >>>

Whoops. Yeah I see that now.

 

[Tony Cahncuhd]

Well, no, not really. The article provides a principle that demonstrates that some variance is to be expected, but the magnitude of the recorded variance can't be explained purely by the principle they're suggesting. The recorded variance is 1.35 PSI. If I'm really generous on the time and include both the wet and dry curve, then using Exponent's graphs I can squeeze maybe .9 PSI of variance out of the temperature change.

I have seen some people say that their wet ball experiments weren't done well though, and I agree with that. But I haven't seen anyone produce an alternative transient curve yet. (Some have suggested that the values of the curve should shift up or down based on different gauges or starting assumptions, but I haven't seen anyone suggest a curve with a significantly different shape or magnitude.)

That said, there are significant problems with how the curves were displayed in the report that skew how they are being perceived. For example, the two graphs on page 55 of the Exponent appendix show non-logo and logo graphs next to each other as if they are comparable. They aren't. In order to treat the Colts balls as a control, they changed the assumptions (pre-game temp) between creating the left and right graphs.

For the record, I don't think they did a great job of that, either: they fit the curve to the average Colts reading at a time that's likely earlier than when the actual reading took place, throwing off the midpoint of the curve. There's also a question of if the order of events was Measure Pats -> Measure Colts -> Inflate Pats or Measure Pats -> Inflate Pats -> Measure Colts; if it's the latter, then the graphs and controls are way off. (There's an assumption that it was the former throughout the document but I don't think I ever saw a direct statement that it really happened that way. Maybe I missed it, though.) Similarly, though the graph shows the entire halftime period, only minutes ~2-7 should be compared to the Patriots' average bar–including minutes 7-14 just deceptively makes the average bar appear further away from the applicable part of the curve than it really is.

But yes, I agree that both the wet/dry ball factor and the likely lack of precision in the initial measurements contribute to and may explain the high variance. I just haven't seen anything that would let me put specific numbers to either factor. Lacking that, it's hard to claim that they d0–and we're obviously never going to get an answer on the initial measurement question, so...?

And just to be clear, when you say 'variance' you actually mean 'range' because statistical variance is calculated by summing the squared differences between observations and the mean and dividing by sample size. Using the right terminology might keep things less confusing.

Without knowing the precise beginning pressures and temperatures it is impossible to interpret the variability of these footballs. The esteemed Mr. Wells should be able to understand this. Not acknowledging it trashes his claims of independence and objectivity.