Copy
“Learning is not attained by chance, it must be sought for with ardor and attended to with diligence.” – Abigail Adams
View this email in your browser
Proper Tack #63 - 19.10.03

Analytics & Being Less Wrong


Is this thing still working? Hello?

Well, that was a weird 3 months. It started with a 2-week vacation to Italy and then a cascading series of very exciting but very time consuming events and you, dear readers, had to suffer and I apologize for that.

But now we're back and we're going deep on analytics. The genesis for this edition was a trio of articles from The Athletic so it will be a little sports-heavy at times, but the underlying concepts are applicable across all walks of life.

Now, if you aren't familiar with it, The Athletic is a fantastic publication and you should subscribe to them. On top of articles like the three today, they have, I believe, the best collection of soccer writers covering the American game. And not just MLS (which obviously I have a soft spot for) but anything to do with soccer in America. But that's a conversation for a later date!

🏟️  Just one write up today; a deep dive into 3 articles from The Athletic.

🎏  But, as always, I've still got a collection of great stuff I've found recently to share with you.

If you are enjoying Proper Tack, the best thing you can do is forward this newsletter to anyone you think would enjoy it and ask them to sign up here.

This entry's header palette is alba sul mare.


The Art of Being Less Wrong

Today, I want to talk about three articles published by The Athletic over the last month or so. All three focus on analytics in sports, but from a conceptual standpoint rather than another dive into the latest stats and what they mean.

Each article was written entirely or in part by Justin Bourne, a former professional hockey player and Maple Leafs video coach, or Seth Partnow, a former poker player (which is actually where I first encountered him), turned Nylon Calculus managing editor, turned Milwaukee Bucks director of basketball research. Fun fact, Seth is only the second most successful 2+2 NBA season thread contributor, shoutout kbfc

But back to the point of this all, analytics. And specifically what they can and cannot accomplish for a team. I loved these articles, so much that I am straight up stealing the title of one for this email. Sorry Seth.


The first article to check out is Bourne's piece on the coming revolution in NHL data. For the first time, the league will be compiling and releasing player tracking data. Not just shots, goals, hits, but things like on-ice position, distance to the nearest defender, and total distance covered.

Like analytics in any field this is going to mean a lot of people saying “whoa this is really cool” and also “what does this mean?"

Hockey stats are really hard because you never truly know what a player is trying to accomplish outside of “score more goals than the other team.” In free-flowing sports like hockey there tend to be fewer pure 1-on-1 battles like those a pitcher has with a catcher, or an offensive line has with a defensive line. As Bourne puts it: 
On a shift-to-shift basis in hockey, you may get a coach in your ear saying, “Hey, the past couple plays it’s looked like your hands have been giving you a bit of trouble in trying to make moves at the blue line. Let’s get everything in deep the next couple periods and just get your line playing in the right end, OK?” Maybe he makes that ask because your linemates are turning it over too much, so you’ve spent long stretches playing in the D-zone. Maybe it’s because the coach thinks the opposing D looks off and so he just wants the team to gain the zone without risk, thinking the puck will be easier to retrieve that night.

Maybe after the game, it’s written that Player X entered the zone with possession zero times on a half-dozen neutral zone touches, and the discussion around his play ends up way off base, given he was actually executing the coach’s vision to a T.
This sounds daunting, and it is. We can try to infer what an individual player's goal was, but with significantly fewer direct engagements than baseball, significantly more opportunities to engage with the coaches than soccer, and significantly less scoring than basketball, hockey presents a unique challenge in trying to turn raw data into actionable findings.

And this is where some people may stop. Why be good if we can't be perfect, right? Bourne also speaks to this point head on.
That’s not to say we disregard numbers. It doesn’t even really devalue them. It’s just an intermittent reminder that eyes and stats and even inside information all play instruments in the band called Analysis.

And that takes us very nicely into our second article, a conversation between Bourne and Partnow. The two of them go back and forth on the relative value of data and what successful analysis looks like for team. 

The whole thing is a nice easy read, but it was this framework, laid out by Partnow, that really stood out to me.
There are sort of three levels of statistical analysis:
1. Descriptive — What happened?
2. Evaluative — Why did it happen?
3. Predictive — What will happen next?
We generally have to answer in that order
I think this is applicable to every type of business out there. You have to see the data, then you have to analyze the data, then you have to use it to make decisions going forward.

The hockey data we'll have this year is firmly in level 1. People will start working on level 2 almost immediately, trying to crunch and analyze all the new data and turn it in to new stats. But level 3 is what people think of when you talk about analytics, in any context.

And level 3 is really hard. Which brings us to the third and final article.

The entire premise of Partnow's piece is that analytics done right is a series of small incremental gains that add up to a significant competitive advantage over time. It's not a bunch of geniuses working on solving a single equation that will revolutionize how the world works, and the process of doing this analysis is not going to give you the right answer 100% of the time.
Being “better” is not being right about everything; it is being less wrong.
And that, to me, is what this thing we call “analytics” really is: The art and science of being less wrong. However, the specter of perfection often gets in the way.
...
[T]here is a temptation among traditionalist skeptics but also from within the analytics community itself to expect perfection. There is something about a number that implies precision. Accuracy. Certainty. Immutability. The implied message is a problem solved in totality. While exacting standards are generally a good thing, the notion of perfection can be a mortal enemy of improvement in light of how progress actually happens.
Interestingly, Daryl Morey and Jeff Luhnow, the GMs of the Houston Rockets and Houston Astros, respectively, both agreed in a recent WSJ conversation that they would rather have bad data no one else had than good data they had to share. The idea being that it’s VERY HARD to beat someone else when you’re all working with the same inputs. As Partnow said in his conversation with Bourne, "Baseball is relatively easy to analyze, but all that means is now they have to have literal rocket scientists to learn new things."

So don't be discouraged if every analysis doesn't give you a clear answer, or if every test doesn't improve performance. Just keep pushing forward and trying to be a little bit less wrong.

 

More Good Articles to Sound Interesting


I've had a LOT of time to read stuff since our last newsletter, so hopefully this edition's links will be even better than usual.

🏦  A 350-year British banking dynasty you’ve probably never heard of. The article features this great quote "There are two things that can destroy a family business: the business and the family, and they both have to be kept in order."

🚁  The helicopter team that films the Tour de France is amazing.

👨‍🏫  A photo survey of mathematicians' blackboards.

🏃‍♀️  Want to feel lazy? This 71-year-old grandmother runs every damn day and puts up marathons in under 3:30.


Thanks for making it through another ProperTack email. Please forward on to anyone who would enjoy it.

See you next time. Which won't be 3 months from now, I promise.
If you enjoyed this content, please consider sharing it
Share on Twitter
Share on Facebook
Forward to Friends
If someone referred you to this newsletter and you like it, please consider subscribing
Follow me on Twitter
See pictures of my dog and baby on Instagram
I also, very rarely, write on Medium
Copyright © 2019 Proper Tack, All rights reserved.

Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list

Email Marketing Powered by Mailchimp