On Wire Management, Flyball pitchers, and an awkward-looking shortstop

I recall a mathematics professor using an HL Mencken quote to describe a particularly intractable physics puzzle:

"Explanations exist. They have existed for all time; there is always a well-known solution to every human problem - neat, plausible, and wrong."

(I'm certain that the abovementioned professor paraphrased Mencken, but thanks to the power of the internet we can get the quote right).

Three discussions I had over the past week brought this idea to mind, and have me reflecting on the limits of intuition and of statistics, both professionally and recreationally:

1. A discussion of the defensive ability of New York Mets shortstop Wilmer Flores
2. Predictions on future success of Mets pitcher Carlos Torres
3. A discussion on the AV Installation Nightmares Facebook group on the merits of plastic cable ties as opposed to hook-and-loop straps to bundle category cable.

What do the above have in common? All are, to some of us, important questions. None are within our ability to intuitively answer. What it comes down to, in the end, is almost a question of epistemology: how does one know that an assertion is true?

A Data Driven World - When Statistics are the Answer
We know things today. Gathering data is a skill at which we have become quite adept. One high-profile example of this is in election predictions: there has always been a great deal of polling and data gathering before elections. In recent years it's become possible - even easy - to gather data from wide areas over years, make comparisons, and see which correlations appear stable over time. This lead, for example, to several statistics-based analysts (ie, Nate Silver of "538" and Wang of the Princeton Election Consortium) to predict the last Presidential and midterm election with nearly frightening accuracy. More traditional pundits relying on fewer numbers but more experience and intuition ran into one of Mencken's neat, plausible, and wrong explanations.

Warmups at CitiField, Queens NY

The key is in knowing what one knows and how to leverage that knowledge. This brings us to the discussion of Carlos Torres, relief pitcher for the New York Mets. In a discussion on the Amazing Avenue blog, one commenter asserted that based on statistical analysis, Torres' early success was based largely on luck. A more "traditional" view would be to look at simple statitics such as "earned run average" - the number of runs allowed per nine innings - or "batting average against" - how often opposing batters hit safely. Looking more deeply, Torres had a greater number of walks per nine innings pitched than expected, and possibly a lower "batting average for balls in play"; looking at these statistics, it appears more apparent that his success is a bit of a statistical anomaly.

What fascinated me is that that didn't end the discussion. Another poster - a young man going by the handle "noahmets" who is about to begin an internship with baseball analysis website Harball Times. Noah's contention is that one can look at a set of statistics called  "Pitch f(x)" which give details of the exact location, result (ie, swing and miss, put in play, pitch taken) and even spin-rate of each thrown ball. This data is available via sensors which have been installed in every major league ballpark since 2006. Having this extra data lets us see more and, hopefully, derive stats which are more stable over time. He was able to add a dizzying alphabet soup of pitching statistics which control for ballpark and other variables. What was once conjecture can now become predictive.

A similar discussion about shortstop Wilmer Flores eventually went nowhere; to the more "visually oriented" fans he looks awkward and appears to have limited range. To the statistically-oriented fans his numbers look just fine. The problem with the statistical argument is that he's played very few games at shortstop thus far, and those numbers have fluctuated wildly over this time. Statistics can only give an answer with a workable statistical model and sufficient data; in the case of defensive range, I'm not convinced that we have either.

How does any of this relate to AV? In the abovementioned install nightmares group, somebody posted a picture of an AV rack mid-wiring. In what was meant to be playful teasing, I "tsk tsk'd" him for using plastic cable ties on category cable. What followed was a discussion on why one should use hook-and-eye (ie, velcro) straps instead of cable ties, IF one should do so, if it even makes a difference. My arguments against plastic ties:

1. Ties fastened too tightly can deform the precisely-manufactured cable, increasing crosstalk between pairs
2. It's very easy to accidentally overtighten, especially if fastening many of them
3. BICCSI standards recomment against using nylon or plastic cable ties
4. A white paper from Valens Semiconductor (maker of the chipset at the heart of HDBaseT transport systems) recommends hook-and-eye straps.

AV professional William Bloomquist had what looked like a data-driven argument against. His point:

1. Many people use nylon ties without a problem.
2. There are no failure statistics detailing how many failures are caused by compression caused by overtightened cable ties.

Do you see the key difference between this analysis and the above discussion on relief pitchers? We can very precisely analyze pitching because we have very precise data; Pitch f(x) is in all major league ballparks and measures every single pitch in every major league game. THis is not so for cable failures; if video doesn't pass on one cable it gets re-terminated, re-re-terminated, and perhaps replaced without anybody performing a post-mortem or a root-cause analysis on the failure. There's no pitch f(x) database listing all of the cable failures  across the industry on which we can perform analysis, no decades of polling.

Hook-and-eye cable straps, detail.

What we're left with are best practices designed to minimize failure. When I write  a bid specification, I'll include hook-and-eye fasteners for category cable because I know that that will eliminate one potential source of technician error and, ultimately, failure. Absent statistical proof, we're with the scouts evaluating shortstops based on experience and hands-on analysis. In short, with no real-world statistics and without my own laboratory, I'm left with what other experts have measured and presented. Whitepaper from Valens. Whitepapers from cable manufacturers. Best practices from those organizations in the IT industry whose job it is to know what they're talking about. 

I applaud people like Bloomquist for being prudent and insisting on statistics to back up our choices. Sometimes one needs to take a step back to understand which data is available and the limits of real-world statistical analysis. This has given me food for thought, but at the end of the day I'll still be requesting hook-and-eye straps to bundle category cable