## Featured post

### Textbook: Writing for Statistics and Data Science

If you are looking for my textbook Writing for Statistics and Data Science here it is for free in the Open Educational Resource Commons. Wri...

## Wednesday, 11 November 2020

### T1B: Goodhart's Law and Baserunning

When I was about 11 years old, I was good at running, good enough to represent my elementary school as the anchor in a relay at a district track meet. This prompted all the grown-ups in my life to coach me on running. They told me all sorts of tricks about keeping my hands flat and remembering to breathe and to keep in a straight line and to keep from dragging my feet and to start early to get the baton.

The race came and I did all those things – hands, breathing, straight line, no dragging, start early. I forgot, however, to run. I blew a huge lead by running in perfect form, but in slow motion.

What's the lesson here?

I mean, besides that I have low kinesthetic intelligence (I'm slow to learn how to use my body).

What's a more general lesson?

Don't let minutia get in the way of winning.

Often when there's resistance from coaches and players to something a quant suggests, this is why. It's the sporting equivalent of Goodhart's Law, or "When a measure becomes a target, it ceases to be a good measure."  [1]

So are stats worthless for coaching purposes in sports? Not at all, it just means you need to consider the possibility of second-order effects before you rely too heavily on them. Shot attempts is a great indicator of hockey success until your players start shooting from center ice. On-base percentage was at the core of the Moneyball strategy. Aiming for OBP gets more walks and uses more pitches, thereby using up a team's best pitchers faster. That strategy works great right up until opposing pitchers learned to throw strikes more often.

What makes a good stat to measure and aim for? If you're a coach. One with minimal opportunity to reach through some objectively bad strategy, and one with minimal opportunity for counter play.

A good stat for player evaluation is also actionable in that a player can do something to improve it; it should be reliably measurable and attributable to the individual player. Finally, and most importantly, it has to be correlated with success.

T1B, or Time to 1st Base, is an ideal stat.

T1B is measurable. One camera tracks when the ball is hit, and another tracks when first base is reached. The difference is T1B. There's already cameras trained on both locations, so tracking is 'free' in terms of equipment.

T1B is correlated with success. There are many times when a short T1B make the difference between an out and a single, and there is no such thing as having a too-short T1B.

T1B is player-attributable: Nothing the pitcher, catcher, fielders, or anyone else but the runner themselves has a meaningful effect on T1B.

T1B is actionable. Players can improve their times through sprinting training and by incorporating it into their batting practice.

T1B cannot be gamed. There are two ways to shorten T1B – get to base early and hit the ball later. Hitting the ball later earns to a couple of milliseconds at best.

T1B has no easy counterstrategy. The only way to counter a good runner T1B (or RT1B) is with a good fielder time to 1st base (FT1B). That's going to put pressure on the shift, as well as force errors from rushed fielding.

There are some composite stats like wins above replacement (WAR) and American football's passer rating that check most or all of these boxes. But few of them are also as good for fans and media as T1B, because unlike WAR.

T1B is defined absolutely. WAR and passer rating are subject to changing definitions based on changes from year to year as well as differences in opinion.

T1B is easy to understand and discuss. Is a WAR of 4.0 good? Yes, give me 10 minutes to explain why. Is a T1B of 2.8 seconds good? Yes, runner go fast.

T1B is available on the spot. Not only can it be displayed on the screen right as it happens, broadcasters could even have a clock ticking upwards like they do for shot clocks in basketball and hang time clocks in American football.

There's an extra benefit to measuring and aiming for a better average T1B: it prevents players from doing on purpose what 11-year-old me did by accident – jogging.

An average runner who goes all-out every time is going to have a better T1B than a fast runner who jogs when they hit a pop-up on even 10-20% of their at-bats. "Great tacticians learn that consistency often trumps potential."[2]

How often does a player hit a pop fly or something similar and jog to first base instead of sprinting all-out? Very often. How often is there an error exactly large enough that running would have made a difference? Not nearly as often, but more than zero times.

Arhg! Why are you jogging?! You get 4-5 plate appearances per game and it's like 30 metres away. Is the risk of injury really that high? Are you seriously hoping for a double on an error?

Just run! Run every time! Run on strike three! Make every at-bat a fire drill and make fielding, pitching, and catching against you a fearful, stressful ordeal! Make the game a little more fun to watch!

Let's see them shift against that!

[1] Dame Ann Marilyn Strathern, summarizing Goodhart's Law.

[2] Resolute Technique, Path of Exile