Over at The Book Blog, Tango has a post about standardizing Pitch F/X graphs to be catcher/batter/umpire view. While I don’t think this should necessarily be the case for all Pitch F/X graphics, it brings up another issue that I’ve had with some graphics that I see online: they’re often not marked with labels or legends.

There are some fantastic looking graphics that sabermetricians do at places like Beyond the Boxscore, Fangraphs, etc. Graphics are an important part of sharing your research, especially with a more general audience. Unfortunately, I’ve read some articles where the graphs aren’t explained fully, despite looking like they could be extremely useful. The first lesson you learn in a multivariate statistics class is to ensure your graphs are readable and your audience understands what they are looking at. Otherwise, what’s the point?

If you’re using Excel, the right graph does this automatically. If you’re using a more advanced program like R, Stata, GAUSS, MATLAB, etc., just Google ‘adding legend in R’, and you’ll find great resources for improving your graphics. The website Stack Overflow is one of the best places to find this type of information (or, simply the R Cran Network if you’re using R).

So anyone writing up articles, please PLEASE explain what the graph is showing me. Include a legend…ALWAYS include a legend. Include a title. Make sure to use color to your advantage. Color can be abused like crazy in graphics, so be sure to use it intelligently (this seems to be less of a problem, as I’ve seen many people with much more graphical capability than I at many sites). Even if at your site you’ve used this graphic before, make sure to include a legend and a title. If you don’t want to explain what it’s telling us again, that’s fine…just link the article where you do. But ALWAYS mark the graph in every post you put up.

But as long as everyone agrees to tell their audience what they’re showing us, and mark things accordingly, I don’t see a problem. Different formats are appropriate for different problems, so standardizing the view may be a bit stringent.