31 May 2014

ESPN's MLB Power(??) Rankings

The bottom rungs of the
May 26, 2014 rankings
Each week since week 15 of the 2005 baseball season, ESPN has been compiling "Power Rankings," a list from 1 to 30 of the quality/current trends/strength of each team in Major League Baseball. Each team appears with its ranking, logo, name, current record in wins and losses, and a witty factoid about the direction the team is going in or about a player or two who exemplifies that trend. (Though the meaning of these trends is sometimes obscure as I've discussed before.)

It always rankles me when my favorite teams appear lower than other teams with the worse records than my teams' own. Of course that's part of the point of these rankings: to show subjective impressions of who is actually better than someone else despite having the record not showing their "inner strength" or some mumbo-jumbo like that. Still, when week after week, it appears that your team is getting the shaft, you begin to wonder if perhaps the rankings aren't so much about power on the field as power to draw viewers to the Sunday Night featured game, which doesn't exactly showcase all thirty teams equally.

To test whether there's something more happening, I looked at all of the power rankings from Week 15 of 2005, when ESPN started them or at least put them online in their current format, to the latest incarnation, Week 10 of 2014. (No, I didn't look at them all directly, I have computers for that.). For each team each week, I assigned a "bias point" for every rank that a team was above them with a worse win-loss record than them. For instance, if a the Jays at 25-20 were ranked 2 and the teams in ranks 3 and 5 had better records then them then the Jays would get 4 bias points, one for the team with rank 3 and three for the team in rank 5. Similarly teams would get negative bias points for teams ranked above them with worse records. In the illustration above, the Padres (my favorite, generally losing team) would get -1 points for being a position below the Red Sox (my next favorite, generally winning team). The Cubs and Diamondbacks would exchange no bias points even though percentage wise the Snakes are slightly ahead.

Over about ten years of Power Rankings, the teams most often ranked above their records were

1. Yankees (2052 points)
2. Red Sox (1101)
3. Tigers (968)
4. Angels (945)
5. Blue Jays (918)

The bottom five were

30. Pirates (-1141)
29. Astros (-1113)
28. Orioles (-977)
27. Rockies (-848)
26. Marlins (-805)

Since I started this to look at the Padres (-774), I'll just note that they ranked 24th.  For the most part, these numbers make sense -- even though these bias rankings already take into the power ranks assigned on the basis of current records, the people at ESPN wouldn't be earning their keep if they didn't take into account historical trends. In fact, looking at weeks 8 and beyond, the amount of bias is  less, and there's some shuffling throughout the ranks, though not at the very top:

1. Yankees (285)
2. Indians (241)
3. Phillies (210)
4. A's (178)
5. Angels (122)

30. Brewers (-380)
29. D'backs (-326)
28. Astros (-220)
27. Cardinals (-175)
26. Mariners (-153)

(Padres jump to 18th at -30 points, actually beating the Red Sox who drop all the way to 22nd at -70! Clearly those second half spurts and slumps, respectively, make a difference.)

A stronger showing of bias would be whether there's a long-term difference between the record of a team and its bias measure. The five most winning teams over the ten-year period were 1. Yankees, 2. Angels, 3. Red Sox, 4. Cardinals, and 5. Phillies and the most losing were 30. Royals, 29. Pirates, 28. Astros, 27. Mariners, and 26. Orioles (Cubs fans, be glad I didn't extend this list one more spot! oops.; Padres hit #21).  Subtracting the win-loss rank from the Power Rankings bias rank gives a ranking of systematic difference that cannot be explained by records alone.

Most biased for:
1. Cubs (WL rank: 24.5; PR rank: 12; = difference: 12.5)
2. Blue Jays (17 - 5 = 12)
3. Royals (!! Rob Neyer?) (30 - 22 = 8)
4. Indians (19 - 13 = 6)
5. Tigers (7 - 3 = 4)

At the bottom:
30. Cardinals (4 - 16 = -12)
29. Brewers (11 - 20 = -9)
28. Rockies (22 - 27 = -5)
27. Reds (15 - 19 = 4)
26t. D'backs, Padres, Marlins (20 - 23 or 21 - 24 or 23 - 26 = -3)

In both of these lists there are good, average, and pretty bad teams. There are some adjustments that could be made. For instance, since the records only reflect the regular season we could adjust for World Series wins and pennant wins, subtracting a system bias point for each (i.e., 2 pts for winning the Series and one for losing). To keep the numbers exactly constant we'll add one point per team (I love it when the math works out: 3 pts per season and 10 years in the dataset = 30 points, or exactly one per team). At the top nothing changes, since none of the top four teams have done anything in late October, except that the Tigers disappear from the most overrated (two pennants will do that for you) to be replaced by the Nationals. At the very bottom, nothing changes -- in fact four World Series appearances for the Cardinals just makes the bias even worse. San Francisco and Philadelphia make their way into 27 and 26.

Technically by this method, the Red Sox tie Philadelphia, but as #2 on the PR positive bias, they're hardly being discriminated against by any measure, but they do have a legitimate beef against the way they've been treated in the last two-thirds of the season.

There are a lot of ways to slice the data; some of which make the Padres look exploited, and others the give them more credit than they deserve. However, any way you look at the numbers -- adjusting or not for postseason success, including or excluding the start of the season -- there are some teams that the ESPN staff are definitely fans of: the Cubs, Royals, Indians, Nationals, Rays, and Jays. And there are some that get no love: Rockies, White Sox, Reds, Rangers, Giants, Diamondbacks, and Brewers. But most of all its the St. Louis Cardinals who time and time again get the shaft on the Power Rankings.  Maybe it's time for the crew to stop batting their eyelashes at the Friendly Confines and look for inner strength down I-55.

(attachments: Excel Spreadsheet and, in case anyone wants to look at NFL, NBA, or NHL Power Rankings, the Python program that generated the data)

28 May 2014

Conference/Journal Convergence

The latest issue of Journal of Musicology just appeared (in virtual form) for me today, and I've been reading through the fascinating articles by Karol Berger and Michael Gallope this morning (I omit Peter Schmelz's contribution not because I don't think it's fascinating, but simply because I haven't gotten to it at the time of writing). On broad topics cutting across decades and centuries, these are the types of articles that I wish appeared more often in my field. The arrival of this issue has made me set aside (again in a metaphorical sense) the journal that I had been reading through with interest, Gamut, a recently reestablished and exciting music theory journals that publishes entirely online.

Unifying the issues of these two quite different journals, and many of the other publications I've read recently, is that the articles selected for publication came not through open calls for submissions but as proceedings of conferences or as designed Festschriften for a major figure (well, important man, to be technical) in the field.

Yard from Dguendel/WP (CC:BY) 
Sign from arkadin55(CC:BY-NC-SA)
The JM issue comprises papers from the 2012 conference in honor of Richard Taruskin, "After the End of Music History," which included talks by so many of the brightest stars in the musicological firmament and which was covered in the New York Times. James Steichen's introduction to this and the following issue, which will contain more papers from the same conference, helps those of us who were not there feel the energy of the weekend: the dual performances of a lost Eugene Onegin, the premiere of a new work, the evenings on at a club at Princeton.

Beyond placing these articles together within the cultural context that produced them, what musicologists usually aspire to do with their writings on musical works, there is much that is gained from publishing articles from a single conference together. Responses from the audience or the dedicatee can be transcribed, edited, and included (the old IMS reports are fascinating in this regard). Interconnections among papers can be made by the writers themselves. And there is the opportunity, which Karol Berger takes, to keep articles shorter, nearer the length of the original presentation. This facet alone makes publication of conference proceedings and Festschriften in journals worth advocating. Too many of our journals contain articles that are far too long for their topics or narrow audiences. Whether encouraging such length is the editors' intent, or indeed even if they continue to publicly solicit shorter articles, those students and scholars who want to be published look at the tomes that do get accepted and are smart enough to know what to do.

Conferences as journal issues mitigate against this trend. They also help scholars, especially the untenured or underemployed, stay active in the vital intellectual life of the field by attending meetings and not needing to decide whether they will contribute their work to a special conference proceeding book, which will likely count less for tenure no matter how selective the press or how strong the process of peer review, or go out on their own in search of a peer reviewed journal.

Yet I am uneasy about the implications as more and more space in journals, especially at the top journals, becomes devoted to conference papers, or to colloquies, symposia, and round tables (common now in the Journal of the American Musicological Society). Conferences turn into journal issues either by the editors of the journal seeking out conferences or selected papers from conferences to publish or by the organizers of the conference proposing their papers (or, again, a subset) as ripe fruit for an issue. In either case, the scholars who are or could be included come from the group of people invited to present at the conference. Many conferences begin with an open call for papers, but these are rarely the conferences that turn into journal issues. Journal paper selection moves from selecting articles from the broad scholarly public to the group of people who were invited to prestigious conferences. And these invitations rarely, if ever, go out on the basis of the work to be spoken of at the conference--which is almost never written at the time of the invitation--but on the basis of past work.

This development does not harm me personally. I am now at a stage of my career where I receive more invitations to speak at important meetings in traditional and computational musicology than I could attend, let alone write new contributions for. And my newly tenured state enables me to publish more in fora that seem appropriate (even blog posts) than worry about what a promotion committee will think of the surrounding context of my thoughts. The change to conferences as journals does, however, have the potential, especially if it becomes more common, to exclude from publishing those who are less well-connected, who are not invited to be in colloquies or symposia, but who have thought-provoking and well researched ideas. The organizers of "Taruskinfest" and the JM editors are to be commended (and nothing here is directly specifically against them) for including younger scholars in their celebration and publications, but not every conference journal issue, especially Festschrift contributions, are so thoughtful.

The change also has the effect of squeezing out space for ideas and areas of work that are at present too obscure or new to be the subject of a conference or for interpretations that go sharply against those of the senior scholars who are in the position to convene these events. Musicology, and I suspect many other humanistic disciplines, is not yet at a post-publication, post-peer-reviewed-journal state; any changes that may make it harder for the brightest and newest of ideas to be shared in our journals should proceed cautiously and under the scrutiny of the field.

26 May 2014

Country / Capital Challenge

Country / Capital Challenge is a Javascript game I made with Elina Hamilton based on our mutual love of learning geography.

Two players compete to create chains of capitals and countries as fast as they can. The first player chooses a country or capital (any). The second player must choose a country that either ends the same way as the previous country began or that begins the same way the previous countryended. For example, if the first player chose SpaiN, the second player could go with PariS or Nepal. Longer chains are possible, such as SpaIN to INdia.

You can also play in one-player mode.
By default, every 60 seconds of thinking, the opponent gets 1 point. Every time that a player repeats a country or capital, the opponent gets 1 point. If you can't come up with an answer, you can concede (giving the opponent a point) or challenge the other player. He or she then needs to answer quickly; if they can, they get two points. If they need to concede , you'll get a point. Countries that share an exact name with their capitals (such as Luxembourg, but not Mexico/Mexico City) appear only once, under countries. Play to a score that you'd like or a maximum time or total number of countries or capitals. There's no set "Game Over" point.

23 May 2014

Web Pages with musical examples...

Through the music21 Javascript port, music21j, any webpage can now include (simple) musical examples easily.  Copy and paste the following code somewhere in the webpage:

<script data-main="http://web.mit.edu/music21/music21j/src/music21"

If you are writing on Blogger or another platform, you will need to switch from "Compose" to "HTML" mode to enter this tag and the div tags below below.  I wrote the rest of this post in Compose mode and then added the tags later. If you're an advanced user, you can put the script tag in your Blogger template. It will call the scripts asynchronously, so your page loading will not slow down.

The "data-main" parameter says that music21j is needed, and music21j uses the amazing require.js package, specified in the "src" parameter.  Music21j uses new technologies that have been standard in all web browsers released since March 2011. Unfortunately, 20% of people on the Internet still use older browsers. If musical examples are crucial to your page, remove the warnBanner='no' line and a warning message and opportunity to upgrade the browser will be shown. For most of us, a page without examples is better than no page at all, hence why I suggest putting the warnBanner='no' line.

To put in a musical example, put the notation in the TinyNotation format described at moduleTinyNotation inside a
tag with classes "music21" and "tinyNotation", like so:

<div class="music21 tinyNotation"> 2/4 c2 d e4 f#8. g16 r4 c'4>/div<

The "2/4" is, obviously, the time signature. Notes are given by A-G for bass clef notes, a-g for low treble clef notes, a'-g' for higher treble clef notes, AA-GG for lower bass clef notes. Additional quotes or doubled letters represent even higher or lower notes.  #, b, or n are for sharp, flat, or natural (optional) and follow any quote signs.   "r" stands for a rest.

Durations are specified with 4 = quarter, 2 = half, 1 = whole, 8 = eighth, etc. A dot after the number dots the note.  There aren't ways of doing triplets, etc. currently.  If there is no number then the previous duration still applies.

Here is that example rendered.  Click on it to play it.

2/4 c2 d e4 f#8. g16 r4 c'4
There is no way of specifying the tempo, unfortunately. If you don't like the playback, put "noPlayback" in the class tags.  This example cannot be played.

<div class="music21 tinyNotation noPlayback">3/4 C2 D-4 Dn E- E F2.</div>
3/4 C2 D-4 Dn E- E F2.

I'm working on getting other notation formats, including MusicXML, ABC, MIDI, etc. to work soon; however they will not be converted on the browser, so only TinyNotation will work on a completely off-line application.  This summer will see some amazing new developments in this space.

The rendering is done via VexFlow and the playback is from MIDI.js. Both are amazing projects worthy of supporting. The development of music21j is supported by a Brit D'Arbeloff education grant at MIT.