Stats in the Wild

Djokovic recently (a few weeks ago at this point) won the US Open and his rise to number 1 has been incredible.  The data for that graph in the article was collected from the ATP website using the R package XML (which I learned about from The Log Cabin).  In that graph, I was only looking at the top 8 players and their points since 2009.  (In 2009 the ATP changed the scoring system, so it’s difficult to compare players directly from before and after 2009).  Anyway, I had all this data and I figured I should mess around with it some more.  This led to my tennis masterpiece graph.

This graph contains the career trajectories for all of the 19 players who have been ranked number 1 in the tennis world since 1990.  Rather than display the total points of players, I am displayed the percentage of points a player had relative the the number one player in the world at any given time.  (Bearing in mind that the points system changed in 2009.)

The bottom part of the graph contains the time series plot for each player containing the percentage of points they had relative to the number one player in the world at the time.  Dashed and solid lines indicate whether a player is inactive or active, respectively and the width of the line is relative to the number of Grand Slam tournaments that a player has won.  The top part of the graph indicates the time period that a player was ranked number 1 as well as the number of gran slam tournaments a player has won.

All told this graph contains (1) the number of graph slam titles for each player (2) the time period each player was number one (3) a time series of their ATP points relative to the number one player in the world at any given time and (4) whether the player is currently active or inactive.  It’s no Napolean’s March on Russia, but what is?

Cheers!

Republican presidential debate, Obama addressing the nation, AND the start of the NFL season. It’s almost too much to handle.

Before we get to any NFL predictions, I’ll make a presidential prediction.  Mitt Romney is going to win the Republican nomination.  I don’t care what Perry’s poll numbers are right now.  I don’t think Republican’s will vote for a guy who’s first google auto-complete term is “gay”.  (Maybe I under-estimate Republican’s tolerance, but then again, maybe I don’t.)

Anyway, I’ve been toying with the idea of simulating the NFL season for a little while now (I did a bit of this last year, later in the season.) This year I’ve done it before any games have been played, so we’ll see how this model works out. it’s a pretty simply model and uses only data from the 2010-2011 regular season and playoffs.  Using that data, I used a logistic regression model to model the probability that one team beats another team.  Then I simulated the upcoming season 5000 times.

Let’s begin with some pictures.  The first has nothing to do with the simulations, but it’s interesting.  It also give me a chance to quote myself.  So, here is a plot of some Chernoff faces based on the final 2010 NFL regular season team statistics.  (I posted about this before here)

My comments from before:

The face represents the offense and the defense is represented by hair. The size of the nose indicates sacks, the ears indicate turnovers (ear width is interceptions; ear height is forced fumbles).  The eyes indicate penalties and, finally, the size of the mouth indicates wins with a smiling face if the team made the playoffs (a really nice touch, if you ask me.)  The face at the bottom right indicates the league leader.

Some observations on the NFL faces:  The two superbowl teams last year (Pittsburgh and Green Bay) are both located at the bottom of the graph and there faces look very, very similar.  San Diego looks similar to to both Green Bay and Pittsburgh (similar face, nose, eyes, and hair), but the big differences are the ears and, of course, the San Diego face is frowning.  Another thing that pops out at me is how similar Houston and New England look to each other.  They have very similar face shape, eyes, and hair.  The big differences are the nose and ears (sacks and turnovers).

Here is a graph with 32 side by side boxplots representing each of the NFL teams.  Each boxplot displays the distribution of the predicted number of wins for each team.  The teams are in order of the SITW power ranking (which means it’s mostly made up).  I have also included a red W for how many wins the team had last year, a green dollar sign for the over-under betting line, and a blue P indicating whether or not the team made the playoffs last year.

Now it’s time for my Super Bowl favorites table.  The first column lists the team, the second column lists the my predicted odds to win the 2012 Super Bowl, and column three displays my predicted probability of each team making the playoffs.  One interesting thing to note in the first couple lines of this table is that Pittsburgh is more likely to win the Super Bowl than Baltimore, but Baltimore is more likely to make it to the playoffs than Pittsburgh.  This is a result of the NFL scheduling system.  Pittsburgh and Baltimore share the exact same schedule except for two games.  Those two differing games for Pittsburgh are New England and Kansas City whereas those two games for Baltimore are the New York Jets and the San Diego Chargers.  So what is happening is that because Pittsburgh has the chance to play New England in the regular season, in the simulations, when they do make it to the playoffs, they are most often making the playoffs as a 1 seed.  Baltimore is making the playoffs more often, but they are rarely (relative to Pittsburgh) simulated to be a 1 seed.  Remember, this table is ordered by odds that a team wins the Super Bowl; it’s not ordered best to worst team.

Below are some over-under bets that I like.  It seems like a lot of times forget that they are betting on the NFL. Every single team can beat every other team (See Miami beating New England as 13.5 point underdogs a few years ago.) Betters over value good teams and under value bad teams.  My two favorite bets here are Green Bay and Cincinnati.  Green Bay had a great run through the playoffs last year, but they still only won 10 regulars season games in 2010.  Add that to the fact that Aaron Rodgers is a concussion waiting to happen and winning 12 games seems like a difficult task.  Cincinnati has been blessed by the scheduling gods.  Not only did they finish fourth in their division last year earning them games against Denver and Buffalo they have also drawn the NFC west division giving them games against Arizona, Seattle, San Francisco, and St. Louis.  Then add to that two games against Cleveland and it’s not to hard to see 6+ wins in their future.  And think about this, they start their season Cleveland, Denver, San Francisco, and Buffalo.  Is it that far fetched that they start 4-0?  (Yes, it is that far fetched.  Just saying is all….)

Other bets that intrigue me.

And finally, let’s make some predictions that will ultimately prove to be way off.  But it is fun to try here is what the playoffs will look like.

The AFC.

I know, I know.  It’s boring and it’s exactly the same as last years AFC playoff teams down to the seeds.  But wait until you see my NFC picks!

NFC

Ok.  Those weren’t that exciting either.  At least I made a stand with Tampa Bay, right?

And now for my Super Bowl prediction.  Based solely on the numbers I am taking new England over Atlanta.  That’s wicked boring though.  So my gut is taking Tampa Bay over Baltimore 21-20.  And I’m still picking Mitt Romney.

Cheers.