Tuesday, February 28, 2012

Mississippi Valley State's Unusual Win/Loss Distribution

My friend Gregg from Los Angeles recently drew my attention to Mississippi Valley State's unusual distribution of wins and losses in men's basketball this season. As seen in this game-by-game log, the Delta Devils completed non-conference play with a 1-11 record (the only victory coming by a single point in double overtime). However, once league play in the Southwestern Athletic Conference (SWAC) began on January 3, MVSU really took off, winning all 17 games from then on! After the 15th straight win, the national media started taking notice.

Quite a turnaround, right, from nearly winless to unbeatable? Well, even a quick glance at MVSU's schedule reveals that the non-conference segment (played nearly all on the road) was much harder than the conference games. I created the following table to document this fact, using Ken Pomeroy's computer rankings (as of Monday night). Non-conference games are shown in white text, with conference games in yellow. For comparison purposes, Mississippi Valley State's ranking is 246.

Opponent MVSU Opponent Rank
@Notre Dame L 32
@DePaul L 169
@N. Carolina L 7
@S. Carolina L 160
Tennessee St. W 182
Cal Poly L 184
@Arkansas L 111
@Northwestern L 56
@Ole Miss L 103
@Florida L 17
@Wisconsin L 5
@Iowa State L 31
Arkansas-Pine Bluff W 333
@Alabama State W 321
@Alabama A&M W 341
Alcorn State W 342
Southern U. W 327
@Prairie View A&M W 329
@Texas Southern W 283
Jackson State W 335
Grambling State W 345
Alabama State W 321
Alabama A&M W 341
@Alcorn State W 342
@Southern U. W 327
Prairie View W 329
Texas Southern W 283
@Grambling State W 345
@Jackson State W 335

Looking at the table, MVSU's turnaround may seem less astounding and more explicable than when first hearing the team's records during its losing (1-11) and winning (17-0) stretches. In the non-conference portion of the schedule, all opponents were good enough to be ranked 184 or better as of yesterday, and five were ranked in the top 35. In the SWAC, with the exception of Texas Southern (283), all other opponents were ranked from 321-345 (Grambling State, at 345, is dead last in the nation).

One could argue that MVSU's season simply consists of losing to superior teams and beating inferior ones. I don't think this characterization does justice to what the Delta Devils have accomplished, however. First, winning 17 straight games under any circumstances is not easy. Second, the team clearly did not feel its season was lost after starting out 1-11 and kept plugging away as conference play started (one might say the team had a "Devil may care" attitude).

At worst (if it loses in the SWAC tournament), MVSU will be playing in the post-season National Invitation Tournament (NIT), which guarantees a berth to all regular-season conference champions. At best, if it wins the SWAC tourney, the Delta Devils will be playing in the NCAA "Big Dance."

As I suspected, MVSU's road schedule had to do with "The Delta Devils crisscross[ing] the country, playing some of the nation’s top teams on the road in exchange for money to supplement the athletic budget," according to the above-linked Washington Post article. That MVSU lost nearly all of these non-conference games is not that surprising. That the Devils have won all their conference games (even in a relatively easy conference) is impressive.

Monday, February 20, 2012

Accessing My BBC Radio Interview on Jeremy Lin

I had a nice discussion Sunday night with the host of BBC Radio's "Up All Night" show, about the recent success of New York Knicks' guard Jeremy Lin. If you want to listen to the archived audiotape of the interview, click here first to get to the appropriate BBC webpage. Then, unless you want to listen to 3 and a half hours of programming before my interview, advance the time marker as shown below:


Friday, February 17, 2012

There's a nice article on Jeremy Lin and hot hand research in the Daily Beast. I am interviewed in the piece, as are several other researchers who study statistics and streakiness.

Wednesday, February 15, 2012

Putting Jeremy Lin's Hot Hand in Perspective

The athlete creating the greatest amount of "buzz" in U.S. sports is almost certainly Jeremy Lin, a 6-foot-3 second-year New York Knicks point guard from Harvard. In addition to his ability to hit game-winning shots, Lin has gotten attention for his scoring. In fact, he holds the modern-era (i.e., post NBA-ABA merger) record for most points in a player's first five starts (136).

According to David Berri and Martin Schmidt's book Stumbling on Wins, points scored appear to carry great weight in determining basketball players' salaries and receipt of awards (e.g., being named to the All-Rookie Team). However, shooting percentage (or efficiency) appears more important for actually winning games. Thus, I decided to focus on Lin's game-by-game shooting percentages.

Lin recently had a four-game streak of shooting .500 or higher from the floor in each contest: February 4 vs New Jersey (.526, 10-of-19); Feb. 6 vs. Utah (.588, 10-17); Feb. 8 at Washington (.643, 9-14); and Feb. 10 vs. the L.A. Lakers (.565, 13-23). How does this stretch compare to those of other leading guards?

It's been over a year since Kobe Bryant had such a four-game streak. Between January 17-25, 2011, Bryant had consecutive games of shooting .583, .556, .533, and .636.

It's been almost two years since Derrick Rose had one. Between Feb. 22-27, 2010, Rose had consecutive games in which he shot .500, .526, .600, and .643.

Of course, Lin is not the only NBA guard who has shot well recently. Miami's Dwyane Wade is currently on an eight-game streak of shooting .500 and above (.500, .571, .529, .625, .611, .500, .571, and .500).

Monday, February 06, 2012

Yesterday's Miami (FL) at Duke men's basketball game, won by the Hurricanes 78-74 in overtime, featured many examples of hot and cold streakiness. Miami darted off to a 9-2 lead, only to have the Blue Devils outscore the Canes 20-10 to take a 22-19 lead. Miami then responded by going on a 23-6 run to close out the first half and lead 42-28 at the break. The Canes upped their lead to 48-32 early in the second half, before a Duke 16-2 run put the Blue Devils back in the game. Things were tight the rest of the way and, of course, it took overtime to decide the outcome.

Ultimately, however, the lasting impression will likely be how Duke went 0-for-6 on free throws in OT (see overtime play-by-play sheet). As shown below, three players -- two of them above 80% free-throw shooters coming into the game -- each missed a pair from the stripe.

Duke Players' Free-Throw Percentages Within Different Time Frames

Player For the Season 
(Before  Miami)

Regulation 
vs. Miami
Overtime 
vs. Miami
Seth Curry .887 (63-71) 1.000 (4-4) .000 (0-2)
Austin Rivers .684 (65-95) .714 (5-7) .000 (0-2)
Quinn Cook .812 (26-32) 1.000 (2-2) .000 (0-2)

To estimate the probability of the six consecutive overtime misses, we simply multiply the individual miss probabilities on each shot (two for each player). Each player's miss probability is 1 minus his long-term free-throw success rate (in this case, I'm using the players' season-to-date FT percentages prior to the Miami game). The calculation, using the miss probabilities for Curry (1 - .887), Rivers (1 - .684), and Cook (1 - .812), is:

.113 X .113 X .316 X .316 X .188 X .188 = .000045 or roughly 1 in 22,000.

The usual cautions apply to analyses such as the present one. It is post hoc, selected after the fact only because of the unusual nature of Duke's free-throw misses. To paraphrase the words of one statistician, presumably no one asked before the overtime started, "How likely is it that Duke will miss all of its free throws in the extra period?"  Finally, although 1 in 22,000 seems pretty improbable, if one considers all the basketball games that have been played -- in the U.S. and internationally; high school, college, and pro; men's and women's -- perhaps a slump like Duke's was bound to happen at some point.

Wednesday, February 01, 2012

One of the definitions used by statisticians for a hot hand is that a player's successes cluster together sequentially, as do his or her failures. A college basketball player whose three-point shooting has been fitting this pattern lately is Brigham Young University frosh guard Matt Carlino, a transfer from UCLA. (One of our Texas Tech grad students, Tim Oblad, who did his undergraduate work at BYU, brought Carlino's stats to my attention.)

As seen in his game-by-game log, Carlino missed BYU's first 10 games this season before playing in the team's last 14 contests. He started off nicely from behind the arc, going a combined 17-of-33 (.515) in his first six games. After a 1-for-7 (.143) game January 5 at Loyola Marymount, Carlino resumed his hot three-point shooting over his next three games, hitting a combined 7-of-13 (.538).

Since then, in his last four games, Carlino appears to have lost his shooting touch, making only 2 of 21 attempted threes. It's also possible that the most recent opponents have figured out how to defend him better than did his earlier opponents. I've plotted Carlino's game-by-game shooting percentages from long distance (with opponents displayed along the horizontal axis), as shown in the following figure. (You may click on the graphic to enlarge it.)


Admittedly, we're dealing with short sequences of games. However, Carlino seems to get "stuck" either in sequences of hot shooting or of cold shooting. As a point of comparison, Creighton's Doug McDermott (whose three-point shooting I analyzed here) seems able immediately to snap back after a poor shooting night. Hence, at this stage of the season, Carlino appears to be more of a streaky shooter -- with both hot and cold spells -- than McDermott.