Blue Jays/Red Sox

Studes, good points.  Thanks for the feedback.  I realize now that my plot was unclear. 

Speaking broadly, I think these graphs are a GREAT way to show the ebb and flow of a game and the relative contribution of individual events toward a team win.  We’d both agree that this type of graphic would make a great supplement to the line scores we see in the paper every morning.  In a way, they are like Tufte’s Sparklines, a graphical ‘word’ that summarizes a lot of data.

But. . . what is “done right?”

The standard way of presenting this info is a Y-axis that shows the probability of winning the game for just one of the teams.  If you’re plotting the White Sox, and they win, the line “stops” at the top.  However if you are following the Cubs, and they lose, it stops at the bottom. 

Now, there’s nothing wrong or bad about this, but I thought I’d try something different.

My notion was to conceptualize the Y-axis as a measure of WINNING.  I was thinking of the area under the line as “possession” or “advantage” or “progress” toward the goal of victory.  If your team is winning, and the pitcher is mowing down hitters, then the line rises as they slowly progress toward that goal.  If they flounder, then the line falls.  And it may even fall below 50%, when they “give up possession.” 

At this 50% line, to indicate this change from Winning the Race to Losing the Race, I used a color change.  It’s a change of possession.  It’s a turnover on the race to victory.  Red gives up, Blue takes over. 

But it’s obviously confusing.  Perhaps the solution is a clear labelling of the Y-axis?  A different title and label for the colors?  A different use of the colors?  No colors at all?

Or perhaps the standard way of showing the WPA suffices?

Anyone have comments or suggestions?

-Tybor

ps - I’m leaving aside the colorblind issue, which I think could be remedied by grayscale or choice of appropriate colors.  I don’t know much about that, however.

pss - Btw, thanks again.  It’s awesome to have the spreadsheet that easily generates the data, to make this discussion even possible.

I like what Tybor was trying to do—use colors (or grayscales) to show which team is ahead—but the final product doesn’t quite work. What about a graph set up the way Studes has been doing them, but coloring the space between .500 and the WPA? I hope that’s clear.

How did you score the Shea single in the ninth when Reed got thrown out at the plate? I was WPA scoring this game, as well, as that play (in my mind) almost had to be broken into two because you can’t score it as one as that penalises Shea for singling and rewards Foulke for giving up a hit.

Also, does WPA not seem to be a bit hard on the starters? I got Arroyo at 0.054 and Halladay at 0.030 for this game, and both posted pretty good lines.

I wish I had half the workability with graphs that you guys have.  It’s a lot of fun to see what people with actual skills can do with their graphs.  My lack of Excel-lence (ha) is turning out to be a real pain in the ass.

I do find that a lot of plays are necessarily broken down into numerous happenings, but I’m not sure I’ve noticed the severity with regard to pitchers.  I don’t know much about the Arroyo and Halladay lines, but when I’ve done it with Clemens/Hudson, Backe/Thomson, and Oswalt/Sheets recently, it’s been pretty fair.

I’ve also wondered if WPA is a bit hard on starters.  I recently covered the 1-0 ten-inning game between the A’s and Angels, and found that Harden and Washburn got all the WPA they deserved.  That’s just one datapoint, however.

Yes, many plays do need to be entered as two plays, essentially.  That will give you the best allocation of WPA to players.  Breaking baserunning into a separate play from a hit might be valuable, as an example.  I’ll post an article on this subject soon.

By the way, “Awrr”, that’s a very nice job with WPA on your blog.  I’ll add a major post referencing it.  And your graphs look fine!

Great job with those Astro graphs!  I like the use of the big textbox and the numbers that reference it.  It keeps the audience reading from left to right, to follow the flow of the game.  Well done.  And that dip with the Franco R/L match-up is very cool.

Your graph is simple and very readable, yet at the same time chock full of information.  Perhaps my color attempts were nothing but “chart junk.”  Vince, I’ll try your suggestions with a game or two this weekend and post a link to the results.

I’m also very interested in hearing more about how to break down plays.  Here’s a tangential topic - if Guillen has Konerko bunt with men on 1st and 2nd and nobody out, should we really dock Paulie for lowered expectation?

To follow up on Tybor’s tangent, the Reed Johnson play from the aforementioned Red Sox-Blue Jays game is another example. That was clearly a case of Butterfield waving Johnson home and Johnson obeying his coach’s orders.

Now, Johnson did nothing wrong in the situation and even made a pretty nifty effort on the slide to avoid Varitek’s tag, but he still got thrown out.

Breaking the play into two parts, which is fairer to Shea/Foulke and Varitek/Payton for their defence, penalises Johnson for doing, essentially, nothing wrong.

Is it a stupid idea to give credit to coaches (for sac bunt situations, as another example)?

Thanks for the kind words and the front page shoutout.  I’m glad I finally found something to make use of the space with.  I made a similar post for the Sheets/Oswalt showdown, but Blogspot killed it.  I posted the graph but was too disheartened to redo the write-up.

As far as the hypothetical of penalizing coaches goes, my inital thoughts are that it doesn’t make too much sense.  WPA is already a system stacked in favor of coaching decisions.  Even though it’s not Konerko’s fault he’s following orders, a reliever is (similarly) rewarded by being used in significant situations.  Even if X and Y are equal pitchers, their success with respect to WPA is completely dependent on where they’re used in the game.  You could probably say the same for late-game pinch hitters.  So much of the system revolves around coaching decisions that it seems overly difficult and perhaps unreasonable to try to separate positive or negative WPA between players and managers/coaches.

Yes, I agree 100% about coaches.  You’re right that WPA speaks to coaching decisions, and a seasonal review of using best relievers in high-P situations is a great use of the data.

But this spreadsheet is best used for tracking players.  To judge coaches, collect the data and see how pitchers and batters were used vis-a-vis critical situations.