Features: Similarity Scores

We recently leased a newer, more powerful machine to host our football, hockey and basketball sites. Basketball hasn't moved yet, but today we flipped the switch on the football and hockey sites. This new machine boasts 16 cores and 16GB of RAM, so it will be a lot faster than our previous machine.

This should mean you can now view Tom Brady and Jurrell Casey's pages much more quickly.

With the Hall of Fame's inductions tonight, just a reminder that you can check out a list of Hall of Famers here... Also, you can select HoF induction status as a criterion in Play Index searches.

We don't know what this weekend's games will hold, but maybe one of them will end with an historically-unique final score. If so, as a Pro-Football-Reference user you'll know first because you can run a search on any combination of final scores to see how often they've occurred all-time.

Using the main table on this page, and clicking the "count" column header to sort by he number of instances in the database, you can find the rarest and most common scores in pro football history. You might be surprised to see the number of combinations that have taken place just once, running the gamut from 66-0 blowouts to brutal 5-3 affairs. And at the other end of the spectrum, 20-17 is by far the most common all-time final score.

Play around with the tool for a while, and you'll always know where to find an answer when somebody asks, "how often does this score happen?"

If you want a variety of detailed splits for a player's stats, check out the Splits [+] option on his player page. By mousing over "Splits [+]", you can access career splits, as well as the numbers for any individual season of his career.

Might as well face it -- most people are going to use our new Play Finder tool to search for clutch stats on quarterbacks. How do I know this? Because, when Basketball-Reference rolled out its Play Index+, the most popular searches (by far) involved some combination of Kobe Bryant and clutch/late-game shooting splits.

And if there's anything fans love more than the fabled NBA "closer", it's the crunch-time QB.

So let's do this. Much like I did in the clutch receivers post, I'm going to dig up clutch stats for QBs from a variety of different angles. There is no one single, unifying definition of "clutch", but we can at least try to get close to a consensus by using the splits most people would think about when looking for clutch performances.

Read the rest of this entry

Want to find a chain of teammates between any two players in MLB history, Six Degrees of Kevin Bacon-style? Then check out our Oracle of Baseball, which can help you find things like this:

Chain from Mickey Mantle to Mike Trout

One old football aphorism was always that teams needed to "establish the run" -- running early, as the theory went, would set up the pass later in the game. Often it was cited that "Team X is 10-1 when they run 30 or more times in a game," or some such number that failed to see the difference between correlation and causation.

Later, researchers like Football Outsiders' Aaron Schatz would point out that teams have such great records when they rush so frequently because the running game is used to run out the clock late in games by the team in the lead. Teams run when they win, not win when they run.

Using our new Play Finder tool, you can illustrate this by looking at rushing yards by half. Here were the teams who had the most rushing success in the first halves of games a year ago:

And here were the most successful rushing teams in the 2nd half:

If you look at how those numbers correlate to winning percentage, total 1st-half rushing yards have a Pearson coefficient of -0.090 (meaning there's practically no relationship -- and whatever relationship there is is negative!), while yards per rush in the 1st half have a 0.179 correlation with winning. 2nd-half total rushing yards have a correlation of 0.289 with winning percentage, and yards per carry in the 2nd half has a correlation of -0.303.

These are small correlations, but the point is that total 2nd-half rushing success is much more correlated with winning than 1st-half rushing (which has practically no relationship with winning whatsoever). In other words, so much for "establishing the run."

More interestingly, per-play rushing success in the 2nd half has the strongest correlation of any of the variables I looked at above... and it's negative! This makes sense, because teams running out the clock are often calling safe, straight-ahead plunge plays, while defenses with the lead are willing to concede effective runs to the trailing offense.

Because of those factors, the most telling numbers of all are simply that 1st-half rushing attempts had a -0.312 correlation with winning, while 2nd-half rushing attempts had a correlation of 0.612.

Which basically proves once again that teams need to establish the pass early, not the run, and use rushing to run out the clock once they have the lead.