Hacker News Reader

Comment by mjn | Hacker News Reader

It's interesting that he would go to a sports network over the culture clash between statistics and narrative. Sports has traditionally been extremely negative on "stats geeks", and favored narrative explanations, along with factors like "grit" and "determination" and that kind of thing. Silver's own former work in sabermetrics was a frequent target of pot-shots from ESPN commentators for years. There used to be a blog on that kind of anti-intellectualism in sports commentary (written by people who post-blog went on to write for the TV series Parks and Recreation): http://www.firejoemorgan.com/

Sign of a more general change?

View on HN

I actually think the opposite: Sports is full of stats geeks. The whole essence of sports is statistical: Who gets how many points? Who runs faster? Who throws faster? Who's taller? Heavier? Who's highest ranked? Stories glue all of this together, but sports = stats.

I think of that as more akin to how political pundits use polls: they cite numbers, but they object to the use of statistics as an analytical tool to tie them together. The Moneyball approach in baseball was hugely controversial, as were (are?) sabermetricians like Silver trying to push predictive statistics like VORP and FIP over the more traditional descriptive statistics. Perhaps more to the point, also to challenge traditional narratives (not that different from the political horse-race narratives) of "momentum" and "choking" and "playoff runs" with statistical analysis. The above-cited Joe Morgan (ESPN commentator 1990-2010) was a famous example of ranting on air constantly about this kind of thing. One of his more notorious comments was that on-base percentage (OBP) is a stupid statistic because he doesn't like people "clogging up the basepaths". Perhaps that genre of ESPN commentator is finally on the way out.

>Perhaps that genre of ESPN commentator is finally on the way out.

There is room for both, IMO. The 40+ year old crowd are largely comprised of people who are "done learning things" (a quote my father actually told me when I told him about my baseball research), so you need to have the Joe Morgans of the world.

But since all of the winning teams in baseball rely on an analytical approach, it's never been more obvious that nerds dominate the level of the sport where decisions are made. And commentators are shifting as a result of that, since it's just natural.

But you do need both, often because the people in the 2nd group tend to have worse social/presentation skills than the individuals in the 1st group. No slight intended (I belong mostly in the 2nd group, for god's sake), but it is not an unfair generalization.

I think, at least in baseball, the culture has come around to the importance of statistics and numbers in the game. The fact that Silver has a decent track record in his analysis and that sabermetrics has been showing some positive results probably make him a welcome voice in ESPN's baseball programming. Numbers are a fun thing to show a viewer and to make comparisons with and I think ESPN is coming around to that (they developed their own QB Rating for football a few years ago).

That said, it also makes him an interesting counter to some sports pundits when there are invariably differences between the numbers and perception of players and teams. This will make for some good television. Heck, ESPN just started a show called "By the Numbers" a few months/years ago, so I think there really is a move to numbers-based analysis in sports. The key is just making digestible to the general public, which Silver has been very good at on 538.

Readers prefer reading human narratives over rational analysis in all areas, whether sports or politics. It's not even really about a distaste for mathematical analysis so much as a distaste for cold rationality in general. There's lots of fascinating strategy involved with blocking schemes in American football ("if I step here, he'll step here, so instead I move this way...") that are analytical but not numerical. With the occasional exceptional comment by John Madden, these are rarely discussed.

So you are right about the typical role of stats in sports. Commenters like chasing and lnanek2 are mistakenly keying on the fact that sports, having well-defined rules and human-constructed elements, are very easy to measure and so lots of numbers exist. As you correctly point out in reply to chasing, these numbers are simply thrown out there without actually being used in a coherent analysis.

Are you sure the blocking stuff has no numerical solution or reasonable analysis?

I would not be surprised if the "how to block" mini-game could be productively analyzed in a game theoretic fashion in terms of a few discrete tactics and empirical success probabilities given the tactics choices. From that data, you can compute equilibrium mixed strategies for both the offensive and defensive positions. This was kind of analysis was by economists for the determining equilibrium strategies for making/defending penalty kicks. http://www2.owen.vanderbilt.edu/mike.shor/courses/game-theor...

Of course, the blocking in American football situation is much harder to analyze/collect data for, because it's not just a 1 on 1 successfully blocked or not, but rather the outcome of the play that matters.

Given a discrete list of tactics and enough data, it's of course possible to generate a best strategy for any particular situation numerically. So yes, if your question is "should I pull block or chip block here?" and you are willing to throw out all the information specific to the situation not found in the data (e.g. the fact that the left tackle's ankle is a bit sore today), then you can do it.

But the point is that detailed blocking foot work is not adequately captured by such a framework (unless you're going to do detailed numerical simulations of every part of everyone's body). There is no concise way to explain exactly why chop blocks work well in certain situations, and why a hypothetical block which seems mechanically similar is completely useless in the same situation. Human bodies are complicated, and the answer is usually just "empirically, that's how it works out".

>"Sign of a more general change?"

Possibly, ESPN had some struggles this year [0].

I hope they're looking to change up the formula. I find the traditional sports pundit and the fans who ape them unbearable.

0: http://deadspin.com/espns-ratings-are-way-down-this-year-wha...

> Sports has traditionally been extremely negative on "stats geeks"

On the flip side, bookies have always been taken seriously.

The average sports fan might find the commentators more amusing or moving but if you ask them who has a better handle on what's really going to happen, a professional bookie or a commentator, I bet you'll get the right answer.

You should read Silver's book; "The Signal and the Noise". There he reveals that he was a poker and baseball geek long before and much more than a political analyst. In the book he actually shows respects for both views.

Silver will work for ESPN who own ventures like Grantland.

ESPN is owned by Disney who own ABC, on which Silver will sometimes appear to do some political commentary.

It's much more of a package deal than the vertical venture a lot of people make it sound like.

Obviously you don't watch sports. "Teams down by 10 points in the middle of the third quarter while on the road in November during a waxing moon have won 23.3% of the time in the last 16 years."

All the fantasy sports nuts I know are very into stats...

Traditionally, but not recently. Sports has been very focused on statistics for at least the last decade, probably longer. Most large sports organizations have dedicated statistics departments and make many of their decisions based on the same.

That's a good point. I was thinking more about sports commentary, rather than what happens in the teams themselves, but it's true that teams have been all over statistics for years. I do think commentators, perhaps conscious of their role as public faces of an entertainment product, have been more resistant to statistics-driven commentary, and have instead played up horse-race type narratives.

To be fair, when you're watching the game, horse-race commentary is just straight up more fun to hear. I love hearing a radio announcer just get excited about what's happening. However, even live, they do throw in some statistics. Admittedly, these are the sort of post-hoc statistics that aren't good for anything (e.g. "they win 90% of their away games that fall on Friday").

Offline, fans of some sports are almost exclusively about statistics and numbers. Baseball is famous for this.

There is a big difference between "focused on statistics" and a valid statistical approach. Quoting ERA, Wins, RBIs, and even advanced metrics like WAR/FIP/VORP is totally different than validating methods through regression, machine learning, and the like.

Everyone is happy to point at "stats." Almost nobody wants to hear about confidence intervals.

You can only expect so much dedication to the numbers from people who watch a thing for entertainment.

You don't watch much sports do you? Sports is all about stats.

More trivia than stats. That is a distinction Silver was big on in his first incarnation as sports statistician, and he got huge pushback from it. Nobody in traditional baseball commentary wanted to talk about VORP, say, and they certainly didn't want to root their commentary in solid statistical analysis. Traditional baseball use of statistics (I mainly follow baseball) has been very much of the trivia variety: this is the first time a right-handed batter has has two hits in the same inning against a left-handed pitcher while the wind is between 12mph and 15mph on a Tuesday. Not so much about backing up your opinions with solid predictive statistics.

The geeks are winning that culture clash. ESPN.com now includes a lot of advanced metrics in their stats and last year in the Mike Trout vs Miguel Cabrera MVP race they discussed why Mike Trout was the favorite of the new school at length, constantly discussing his obscene WAR[1].

Nate Silver was ahead of his time on those advanced metrics but he was undoubtedly one of the pioneers. If people had been more receptive to his brand of statistics five years ago he likely never even goes into politics. Sports was his first love (along with poker, which he goes out of his way to not mention these days, unfortunately. He even obscured his username on the largest poker forum[2], so it would be harder to find with search engines).

Completely random thought that probably means nothing, but is worth noting: he's openly gay[3]. With the tide on gay athletes turning sharply and quickly, I have to think that factored into his decision. Perhaps he is hopeful he can play a role in that, being the champion of the gay athlete or at least the one guy at ESPN they know will be on their side.

[1]http://en.wikipedia.org/wiki/Wins_Above_Replacement

[2]http://forumserver.twoplustwo.com/members/4548/

[3]http://gawker.com/5969477/sexually-gay-but-ethnically-straig...

slightly off topic, but it's absurd to think that Trout should have won the MVP, when Cabrera was the first Triple Crown winner in 45 years. I don't care what his WAR was.

That was the old school argument, verbatim. The new school disagrees. I don't really have an opinion but I'm pretty certain Nate agrees with the new school and would disagree strongly with your view.

The new school argument is something like: the Triple Crown is an overvalued and superficial accomplishment and giving a player the MVP based solely on that is making a similar error to the scenes in Moneyball where the scouts cared about irrelevant things like whether a player has an attractive girlfriend or how he gets on base (walks vs hits).

WAR is a nice stat as far as a general manager is concerned (or for fantasy), but it shouldn't decide who is the MVP. Last year Cabrera had the best hitting season anyone has had in decades. He finished 4th in WAR because of various weightings given to defense, baserunning, position played, and the bizarre and nebulous "replacement player." WAR is also sadly calculated differently by everyone, making it harder to judge as a useful metric. A single + SB is calculated as being worth more than a double. Why?

I have no problem with using WAR as something to include in the discussion, but it should not be the complete deciding factor.

Interesting side note, apparently Oakland's (famous for using sabermetrics) model for WAR had Cabrera on top - http://bleacherreport.com/articles/1570961-mike-trout-vs-mig...

Reply on news.ycombinator.com