How Moneyball Ruined Baseball

I haven’t written about sports for a little while, so let’s do something a little different for this post holiday week. As I’ve discussed in the past, there is a lot of overlap between trends in sports and markets, particularly around analytics.

While baseball made analytics mainstream through Moneyball, I believe baseball is now highlighting the flaws with taking analytics too far and not planning for unintended consequences. Just as efficient market theory led to the development of behavioral finance, I believe we are due for a reaction to analytics that values common sense.

The Moneyball Revolution

How about a quick review for those who don’t follow baseball closely or perhaps are reading from overseas? Moneyball was a book (later a movie) about the Oakland A’s and their general manager (i.e. COO), Billy Beane, who looked at non-traditional statistics that better correlated to winning than the more common metrics that teams had used for 100 years.

For example, the most common statistic in baseball was the batting average (how often do you get a hit relative to how often you make an out). Beane realized on base percentage (which included hits AND walks, the latter of which batting average ignored) was a more relevant measure. Thus, he traded for players who had a high OBP because other teams didn’t value them as much as players who hit for a high average but had a low OBP.

Sounds brilliant, right? It was actually pretty simple. The innovation wasn’t the math. OBP had been around since the 80s. Everyone in baseball – and many fans – were aware of it. The breakthrough was the willingness to buck convention and actually act on an alternative measure of player worth.

Moneyball was actually more about behavioral finance than it was analytics! Beane wasn’t smarter at crunching numbers than the rest of baseball. Rather, he was better at exploiting fallacies the rest of the market incorrectly believed.

Learning the Wrong Lesson

Unfortunately, the rest of baseball learned the wrong lesson. Instead of focusing on pursuing alternate strategies that others overlook because of their blind spots (e.g. you can’t draft a short quarterback no matter how well he played in college), the other teams turned to data mining.

Teams began hiring PhD physics grads to look for new statistics that better correlated to “win probability” than simpler metrics like OBP. So we got OPS, WAR, BABIP, and FIP among others. These all have their merits but they violated the spirit of what Beane had done.

It was no longer about taking advantage of conventional wisdom and rules of thumb. It became a data race to find the best advanced statistic only your team had knowledge of.

But those advantages only lasted so long. People like to brag about how smart they are and word would leak. Or, other teams would hire away the best quant guys from the winning teams to “learn their secrets”. The duration to capitalize on a newly discovered inefficiency got smaller and smaller (this happens with every financial market new product as well).

Disturbing the Balance in the Force

And then things really took a turn for the worse. Once most teams had their crew of analytics experts and were looking at similar ways to value players, what else could you do to gain an edge?

There was only one thing left to do. Tilt the playing field, i.e. change the way you instructed your players to play the game.

Teams decided the way to win wasn’t by being passively observing what players do and acquiring the ones they thought the most undervalued (akin to stock trading). Rather, they would be like activist investors telling players how to create more value on the field.

Hitters should walk more and change their “launch angle” to try to hit more home runs (since swinging for the fences has higher expected value than a high OBP made up of walks and singles).

Pitchers should throw differently to get more strikeouts, since that is controllable while where a ball goes when it’s hit is more random. (This led to the whole “spin rate” focus which is the current controversy in baseball.) The defense should “shift” to neutralize hitting tendencies of certain batters.

This all led to a focus on what are called the Three True Outcomes, meaning the only things a player could control, and that were of value, were home runs, strikeouts, and walks. Everything else (i.e. every time the batter hit the ball but it didn’t leave the park) was “noise”.

This was Pandora’s Box, though nobody realized it at the time. They were messing with the balance of the game. They had forgotten Goodhart’s Law. You all know this one…when you begin to measure something, it changes the relationship in a way which warps the usefulness of the metric.

Well, I propose a corollary to Goodhart’s Law. When you begin to actively interfere with a metric, you disrupt the natural balance in such a way that, not only is that metric no longer valuable, you often lower the productivity of the entire system so that the new equilibrium achieved is worse than where you began.

In simpler terms, when you tell players to only focus on maximizing the Three True Outcomes, you make baseball unwatchable. The game was better before the statheads tried to improve it.

Global vs. Local Maximum

There is, ironically, a mathematical way to look at this corollary I created. It’s the concept of local vs. global maximum.

Simply stated, in any curve, the global maximum is the absolute highest point on the graph. The local maximum is the highest point of the top of any other curve on the graph (or, in more mathematical terms, any point where the first derivative = 0). Take a peak at the graph below.

The data nerds instructed players to play the game differently to try to raise their local maximum, i.e. the value of the individual player. In doing so, they lowered the global maximum by making the game boring for most fans.

If this sounds to you a little bit like what has happened to financial markets over the last decade plus, yes, there are an awful lot of similarities. Quant investing, ETFs, factor models and other analytical trends that improved local maxima (asset returns for those who invest that way) have made fundamental analysis less relevant to investment outcomes (which is part of the reason value stocks have underperformed for so long).

The global maximum in this case would be the extent to which stock prices reflect company’s prospects. Capital markets function best (i.e. the risk premium is lowest) when issuers believe their stocks are efficiently valued.

When prices don’t reflect fundamentals, there are incentives for artificially overvalued companies to issue more stock (hello, AMC!) while undervalued companies suffer from an artificially high cost of capital and can’t pursue growth that is nominally attractive. This leads to misallocation of capital across the economy. In other words, a lower global maximum.

Customer Experience As a Global Maximum

Let’s briefly bring this to insurance so we can have something for everyone. Does the idea of technology disrupting insurance remind you an awful lot of analytics changing sports? It should! Often technology is being used to pursue local maxima (what is good for my individual company’s odds of success) rather than for improving the global maximum.

What is the global maximum for the insurance ecosystem? Lower combined ratios? Lower expenses? Lower prices? New products? New distribution?

I would argue it’s none of the above. It’s more accurately meeting the buyer’s needs. That may include lower prices or different products, but that’s only a part of the puzzle. The simplest way I can put it is it’s giving the customer what they ask for, not what you want to give them…even when they don’t know quite how to ask for what they mean.

The industry is a long way from elevating that global maximum to where it needs to be. And before you ask, no, I didn’t write this whole post as a set up for an Informed plug, but you can imagine I think about this issue an awful lot and there will be a point in the future where I will talk about how we plan to address this challenge.

(For a preview, check out yesterday’s LinkedIn post.)

Regardless, no one company will be able to do it alone. It will take a lot of others rethinking how they approach their local maximum to really change the way consumers view the insurance industry. I can’t say I’m very optimistic about the prospects (if I were, there wouldn’t have been a reason to launch a new company) but I will remain hopeful.