Too many anecdotes, not enough data

The plural of anecdote is not data

Not the least bit about baseball, but very applicable to sabermetrics and sportswriters.

Believing something doesn't make it true, nor does hearing a couple fans complaining about something on the local talk radio.

over 2 years ago Dan Turkenkopf 8 comments 0 recs |

Story-email Email Printer Print

Comments

As a scientist, I whole-heartdly agree with this point of view and fight to support it.

100%.

That being said, I’ve learned over the past five years that data doesn’t convince people, anecdotes do. So my current model, which I still struggle to follow, is to convince myself something is true with data, have the data ready to back me up, but use anecdotes and hard-hitting emotional stories to convince people of things. It’s just the way it works.

My guess, and I’m no psychologist, is there’s something going on in the brain that needs to be overwhelmed in order to change opinion. Looking at data is not overwhelming, unless you do it for long enough, but it’s tough to get someone to put that much time into it. But if you hit them over the head with a story, boom, they’re convinced.

Convincing stories don’t have to be statistically significant, they just have to be plausible.

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Jul 20, 2009 9:08 PM EDT reply actions

Just read the comments...

And someone mentioned the word “narrative” vs. “anecdote”. I think “narrative” fits quite well. Given the data, you can tell the narrative of what the data says in order to convince people of things. Quite well put, I think.

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Jul 20, 2009 9:11 PM EDT up reply actions

Hmm... I think the problem is that most of the stats we use can't be applied to in-game situation as often/easily as simpler stats.

Like wOBA – how can it be applied to predicting a single AB? That is what most fans are focused on while watching – what’s happening at that moment. And they don’t want to have to do a calculation that would likely take longer than the AB.

@bs_uf15bosox9be:OverTheMonster-ALLERGEN WARNING:May contain PB.

by bdalebs on Jul 20, 2009 11:33 PM EDT up reply actions

In game that's probably true, although the anecdote can be even more misleading

“So and so is 4-9 in his lifetime against Pitcher X with 2 doubles”

Overall player evaluation on the other hand should follow Sky’s advice – use the data and supporting anecdotes to craft a narrative, instead of letting anecdotes in a vacuum drive your point.

by Dan Turkenkopf on Jul 21, 2009 7:43 AM EDT up reply actions

Well, what's the point of a single stat for a single AB?

Win Probability seems quite useful and narrative. And wOBA is quite useful if you want to compare the decision to use the current hitter/pitcher versus making a bullpen change or pinch hitting.

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Jul 21, 2009 7:49 AM EDT up reply actions

True.

I mean that most fans think they can look up on the scoreboard and see a batters AVG or a pitchers ERA and get a general idea of what should happen. While that’s not necessarily true, they prefer to think so, which might explain why they don’t care about stats like wOBA, because it doesn’t seem to have a quick prediction ready. Or maybe I’m just losing my mind…

I agree about Win Probability though – as long as you use at the right moments, and can explain how it was calculated. On Sportscenter they were quoting Win Probability for the Twins/A’s game last night, but they were saying how they don’t know how the number was calculated, and then proceeded to credit random staff assistants for coming up with the numbers.

@bs_uf15bosox9be:OverTheMonster-ALLERGEN WARNING:May contain PB.

by bdalebs on Jul 21, 2009 1:32 PM EDT up reply actions