THE BOOK--Playing The Percentages In Baseball

By the way, the gray-background thing is the genius of studes.  I love that part.

Similarly, there’s alot of wasted white space, right?  Especially since the graph always goes the same way (top left to bottom right)?

Why not reset the scale, and show the “WAR” as “wins above hall of fame replacement level” (WAhR).  The bottom of the gray area now becomes the x-axis at y=0. 

Basically, tilt your head about 30 degrees to your right. Studes, how will that come out?

Interesting idea about moving the axis, Tango.  My first reaction is that I don’t like it, because you then lose the notion of better vs. worse years.  I’ll keep an open mind, though.

There’s absolutely nothing wrong with white space, by the way.  In fact, white space is a very good thing if used correctly, even in written text. Sometimes people are afraid of white space and fill it in for no good reason.

And I also absolutely disagree that you “have” to have the y axis.  I understand that you’d like to see it (cause you’re a data kind of guy), but if the point of the graph is the relative performance of different players, then the absolute numbers provide no extra information. It also depends on the way the graph is presented and how it’s used in the article.

Adding in the career totals is a great idea.

I was thinking of the graph if you have a second graph that has Bonds and Mays and Aaron on it.  How do I know that you keep the same scale?  The only common reference point will be the thickness of the gray HOF band.

But, I’ll bet you can shrink the scale for Bonds/Mays/Aaron by 20%, and no one will notice that the thickness of their gray bar will be different from the 3B’s gray bar (even though both gray bars represent the HOF levels of all nonpitchers, presumably).

So, I like the y-axis so that I can compare different graphs.

I will conceded that the y-axis can be removed as you have if you treat each graph as its own universe.

While I agree that the WAR graphs can become confusing, I believe that some of the simplification removes context.  I like graphs that can stand on their own, and without knowing the background of the graph I don’t know if Dick Allen’s highest WAR is 2 or 20 or how long he played.  If I look at just the plot I have no clue what the gray area means.  At least the y-axis needs to be shown and I would prefer a horizontal line at zero to indicate below replacement.  I like year indications but I can understand removing them.

My preference would be to include only one HoF line and to use a background color like the gray above.  The original intent was to provide a baseline for a simple comparison, much like the HoF stats on BB-ref.  Graying the area between them appears (at least to me) to indicate the total range of HoF WAR, not the area between average and replacement (As a side note, I don’t think the HoF lines have been recalculated since Rally provided all the historical WAR data.).  Presenting this graph on a general blog would require explanation which is avoided by leaving a few items on the graph.  I’m not color blind so I can’t provide context, but it also appears that the program used, the line-weight, and the image size/resolution have a significant effect on readability.  If the plots would have been 50% wider and the WAR scale between 0 and 10, the graphs would have been much easier to read.

I don’t think tilting the graph would help since as shown the lines have decreasing value, where tilting would result in peaks above and below the baseline.

I love the grey zone.  Now I just need to figure out how to create the zone in Open Office.

I agree the total in the name is great.

I am not great on graphics, so will see how hard it is to move the legend in OO.

I am actually planing one soon so let me see what I can do

I agree with JBrew that you need the x-axis at y=0 (without the year indicators).  This is important for active guys who have years that are close to 0 as they will be pushed out farther to the right of the graph as they play more seasons.  For example, Raines’ 1979 and 1980 seasons, if you drew the graph in 1987 would look terrible.  But, the more years he plays, the more those seasons simply take their rightful place to the right of the graph.

I disagree with one HOF line, as it shows in/out, when really it’s a gray area.  I think Studes hit it out of the park on that one.

“I don’t think tilting the graph would help since as shown the lines have decreasing value, where tilting would result in peaks above and below the baseline. “

That’s the feature, not the bug!  I look at those charts, and the first thing that popped into my head is that Dick Allen had alot of peaks and valleys compared to the HOF baseline!

Compare for example to Pinto’s defensive graphs, where he shows the outs made at each zone slice for a player, and he has the “boundary line” that follows along to represent the average player.  But, then he shows the differential at the bottom of the graph.  Which one do you prefer?  I mean, I like both, but after a while, I simply look at the differential graph.

Do you prefer the dotted line (which represents the rebaselining against the standard), or the two colored lines?

As I said, I like both.  But after a while, you realize that SS get outs to balls hit to them, and have a tougher job at the hole and over 2B.  So, they all start to look like the same kind of shape, with some zigzagging across the “average” line.

But, the differential really stands out.

I’m seeing two different arguments about the gray area.  If you assume the plotted gray area is an actual gray area where they may or may not get in, then anyone above the gray area should be automatic, which may not be the case.  Second, if you tilt the graph and then use the replacement line as the baseline you remove the gray area and thus create an in/out scenario.  I agree that with the plot in #7 that the difference is easier to understand, but WAR plots present different information than fielding plots, and have different baselines.

WAR plots would probably be better presented with only positive WAR values, while the total would still include the negative values.

The tilting would keep the gray area.  The differential would be relative to the bottom of the gray area. Wins Above HOF-replacement.

Tom - this what you are looking for in the 30 degree rotation.  I am not sure if I like it.

The extra legend should not be there.  Flickr won’t let me replace an image without paying $25 a year.

Jeff, I can’t see flicker from the office.  You can email me:
tom~tangotiger~net

Good comments!  Some points to throw in:

- Good point, Tango, about having different widths in y axes among multiple graphs.  Then you would definitely need to have the y axis.  But I also understand why, as JBrew and you say, some would want the y axis.

- I don’t think using the WAR graphs for active players is a good idea in the first place, so I don’t see a need to add the 0 baseline, or to only include positive years.

- I much prefer the gray area to the single reference line.  Provides much better visual context.

- I think you know that I helped David with the original fielding graphs.  Here’s an example in development:

http://www.baseballgraphs.com/main/index.php/site/article/good_graph_design/

He added the differential line.  I didn’t like it at first but now, like you, I tend to look at it more than the rest of the graph.

But I think the WAR graphs are different in several ways.  The main point of the WAR graph is that it presents years from best to worst, which is an interesting and useful perspective.  Flattening the graph removes that perspective and gains nothing (except less white space).

Right, I’m not sold on the idea of flattening the graph (i.e., Wins Above HOF-replacement level).  It might be good, it might not.

As for active players, it won’t matter if you like it or not… they will happen!

I wrote up some R code to make the image, so that those of you who use R can make Studes approved HoF WAR graphs. 

All you need to change are the names in the first four lines and the string of WAR values in the next four.  They do not need to be in any order.  You also need to change the location where the file is saved.  That is in the line that starts png, the location is the string in quotes.  It is set up for four players, to add another or take one away would take a little bit of fooling around with the code.  But it should be pretty clear.

Here is what it looks like.  Not quite as nice as Studes’, but pretty close.

It looks like my gray area is a little different than Studes’.  I had a average HOFer with a best year of 8.2 WAR, the worst year at replacement level and a 20 year career.  My replacement level HOFer had a best year of 7.2 WAR, the worst year at replacement level and a 17 year career.  I got the values from the BtB post on the subject.

Here is the code.

Dave, if you add up the total WAR for the HOF line and the HOF-repl line, what do you get?

Off the top of my head, I would say the average HOF would be 80 WAR, and the HOF-repl line would probably be at 70% of that.

Dave, I didn’t know about the BtB post on HOF numbers, so I just eyeballed mine.  I’m sure yours are correct (at least, more correct than mine).

I get 82 for the average HOFer and 65 for the replacement level.

Here is the BtB link:

http://www.beyondtheboxscore.com/2009/5/19/878150/war-graphs-average-and-replacement

Among position-players, Rally’s HOF guys averaged 67 WAR.  The 20th percentile is 47 WAR.

Drawing a straight line from 6.7 to 0.0 for 20 years, that means a drop of 0.35 WAR per year.  The total is obviously 67.

If you start at 7.9 and drop for 0.49 WAR per year for 17 years, the total is 67.

I have to believe this last one makes the most sense.  Gotta love the 0.50 WAR drop, which matches my salary calculator!

Starting at 8, and dropping by exactly 0.5, for 17 years, gives us a total of 68 WAR.  Since I love round numbers, this certainly looks like a pretty reasonable chart (presuming the 8 WAR in best season is around the average for a HOFer).

***

As for repl level, starting at 6.5 and dropping by 0.5 each year for 14 years (to 0) is a total of 45.5.

Not sure how all that matches reality.

Jeff’s chart appears at the top of this thread, just below studes’.

I like it!  It shows that Musial was consistently 3 to 4 wins above the HOF level.  Seeing the chart this way makes me really like it.

Like I said, I would always be mentally subtracting all the lines.  Now not only are things clearer, but all that wasted white space as been filled with the data lines.  In the standard chart, the lines would be so close together, that it would be hard to follow them.  Now, they all stand out.

My vote is for the WAhR charts.

I can see why you like it Tango, particularly if you’re the type of person who is subtracting the lines in his head.  Also, you’re right that the lines are more easily discernable.  To me, that’s the biggest advantage.

I’d still vote for the first one because I like the context of the sloping lines (make the x axis more understandable), but you’ve made me more open to the second.

Tom—In creating the Wins Above Hall of Fame level, do you think the 0 line should be

1. Average HOFer

2. Replacement HOFer

I would say it should be Replacement HOFer and make a gray area up to the average HOFer

Yes, the 0 should be the replacement HOFer, and the top of the gray would come in at say +1 wins above that.

My contribution:

http://www.hardballtimes.com/main/blog_article/more-war-graphing-fun/

I like the “non-aliasing”.

And someone else put the WAR totals, and as I mentioned on THT, for alignment purposes the WAR total should be to the left, not right, of the name.

That would be odd, Tango. People read left to right, and fundamentally understanding a graph is more important than aligning numbers (particularly ones that have little to do with the point of the graph).

90 Boggs, Wade
73 Molitor, Paul
69 Gwynn, Tony
66 Raines, Tim
62 Whitaker, Lou
59 Guerrero, Vladimir

Boggs, Wade (90)
Molitor, Paul (73)
Gwynn, Tony (69)
Raines, Tim (66)
Whitaker, Lou (62)
Guerrero, Vladimir (59)

Whenever I can, I present information using the first chart, not the second.  The letters separate themselves from the numbers that one can easily just look at one or the other.

Nope.  Sorry. Doesn’t work.  In fact, it undermines the point of the graph, making it seem that the WAR totals are the most important thing.

I agree with Tango that it’s useful to have the columns lined up properly, but Dave is absolutely right that this isn’t the way to do it.

I have an idea, but it’ll take me a while to mock it up and put it together with the graph.

Has anybody calculated the WAR median for postion HOFers?  I would bet that it is closer to 60.

Also, how does the WAR mean for pitchers compare to position players?

58 is the median WAR for all HOF players per Rally’s numbers.

60 for Position Players and 56.5 for pitchers.