APBRmetrics

HoopStudies

There may be a lot to learn from how other sports are studied through statistics. It is quite obvious that basketball sabermetrics got a good head start through some of the insights from baseball sabermetrics. Bill James, Pete Palmer, and the Baseball Prospectus guys have fed ideas to many basketball people.

But there are other sports out there doing research. I thought I'd point some of those out.

Football -- The Football Outsiders from Aaron Schatz, http://www.footballoutsiders.com. Aaron has done a lot to extend what Carroll, Thorn, and Palmer did in The Hidden Game of Football in the '80's.

Football is a very different sport than basketball in many ways I can think of.

First, they have units of people that do certain roles that they can't deviate from. Offensive lineman can block and that's pretty much it. By rule, they can't deviate.

Second, measurements aren't great. Measurements on blocking are poor to say the least. Coaches do a ton to evaluate linemen, but they get grades based on a subjective expert opinion. Defensive backs have stats, but those stats are about as reflective of their ability as Bruce Bowen's blocked shots per game rate is of his defensive ability.

Third, "possessions" aren't as valuable a measure, though they exist because football teams do swap the ball back and forth. But field position matters, giving a quality of possession. Football sabermetricians handle this pretty well with expected point totals given field position.

Fourth, decoys are big in football. A wide receiver who draws a double team every time and doesn't get a lot of catches has plenty of value not reflected in the stats.

Fifth, sheesh, you could go on and on...

Frisbee (as in Ultimate Frisbee) is just getting going and frisbee actually has a fair amount of similarity to basketball. At http://sports.groups.yahoo.com/group/FrisbeeStats/ you can see some of the stuff they're working on. I play this a fair amount, so I am quite interested.

Frisbee has some interesting similarities and differences with hoops.

Every player can do everything in ultimate. If you don't know the basic premise, it's that you pass the disc to teammates with the goal of having them receive it in an endzone. A score is 1 point and that is the only way to score. You cannot run with the disc. If you drop a disc, it is a turnover and the opponents get it. One important rule if we try to make analogies with hoops is that no substitutes are allowed in ultimate except after a point is scored. After a point is scored, teams can substitute all they want before the next kickoff (called a "pull"). This substitution pattern causes difficulty with the danval-type analysis, I think.

Anyway, all players can run patterns, catch the disc, and throw the disc, though there are typical roles. There are "handlers" who are usually the most accurate throwers. They can get away with being slower. There are "mids" who usually don't throw as well, but know how to get open and can then throw downfield if they have to. Then there are "deeps" who are usually quicker and can jump and get open longer; they don't necessarily throw as well. In the most competitive ultimate, everyone can throw well.

I don't think the frisbee people have developed a standard set of stats. Some of the posts in that newsgroup discuss possible ones, including turnovers (forced and unforced), blocks (forcing a turnover), field position of a turnover, scores.

I see a lot of room for growth in frisbee stats and potential for help with us.

Hockey has a rather quiet discussion group at

http://sports.groups.yahoo.com/group/HAG_list/?yguid=84605607

though I expect it may pick up now that those fools have decided to actually play the game again. Hockey and soccer are admittedly sports that drive me crazy for their lack of scoring. It really seems to me that some sort of team stat that actually reflects ability better than goals scored/allowed would help, but I haven't seen one. When they talk about new stats over at "HAG", they seem to talk about individual stats -- penalties drawn, minutes served, shots on goal, deflections, plus/minus. Maybe there is something there that can help understand teams but I haven't seen it. Maybe someone here can take what we do to them and make their world an easier place. (I hear that they are changing the rules in hockey with the new CBA to increase scoring. So maybe it will get easier.)

It seems like water polo should have stats somewhat analogous to hoops. But I dunno. And no one really watches it, do they? Arena football is interesting without all the restrictive rules of the NFL, but I haven't watched it enough to know for sure. I'm definitely thinking that team sports are the most interesting. (An aside: "Golf isn't a sport, but it's a helluva way to spend an afternoon," as a philosophical (aka drunk) friend told me, "Chasing women may be a sport, but it really isn't a team sport.")

Obviously this is a hoops group. I think we are doing some of the best work around these days, asking good questions and actually making progress. But I do think that there are people in this group who have more expertise than I about other sports, especially about hockey, frisbee, water polo, and the like (I lived, breathed, and ate football for the first 13 years of my life). It's asking a lot, I know (as this kind of request usually gets a big blank "Huh?"), but I'd be curious if other people have thought about those other sports and ways to handle their stat issues. This is calling for the random brainstorm ideas, not cohesive manifestos...
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.

Mike G

Ed Küpfer · Joined: 30 Dec 2004 Posts: 786 Location: Toronto

I was watching the Wimbledon tournament and thought to myself that there's an opportunity there to learn a lot about game theory -- particularly from service strategy. I know there's been some academic research, but it seems to me that this is something that should really be looked at. Tennis has many advantages for research: it is a sport played between two individual players (so no teammates/coaching influence to consider); it has a binary result (point won or lost); a simulation would be relatively easy to write; a game state matrix would be easy to calculate.

There are some game-theoretic aspects to hoops play that would show up in the boxscore -- fouling strategy, shot selection, pace. I'm not quite sure what conclusions, if any, could be generalised from tennis, but I'm pretty sure we could learn something applicable to basketball.
_________________
ed

HoopStudies

Kurt · Joined: 10 Jan 2005 Posts: 30 Location: Los Angeles

With hockey, I know several teams for sure (including the Los Angeles Kings) keep more internal statistics. The one the coaches in LA used most was a +/- where coaches watched game film the next morning and gave a plus or minus to a player not just for scoring (or being on the ice when a goal is scored) but for making the correct play in the offense (the right pass or read) or other things that fit in with the team goal whether or not those efforts ultimately resulted in a score on that play.

The problem is that measure is still subjective to experts (and, obviously, would varry from system to system). The advantage for the coaches was to reward players who were doing what was asked. I'm not sure, because of the lower scoring, how you can build off of that in a non-subjective way, however.

HoopStudies · Posted: Tue Jul 26, 2005 11:37 am Post subject:

gkrndija · Joined: 20 Feb 2005 Posts: 64

Kevin Pelton · Posted: Tue Jul 26, 2005 1:06 pm Post subject:

According to this FO article, Indianapolis' offense doesn't really play all that fast.

That could be biased because they were a good team and running late in game, but still doesn't really show much effect of their no-huddle in terms of speed.

As theorized in the comments to that article:

"Of course, another way to keep the defense from substituting is to line up quickly but then let the play clock run down. I’ve seen Peyton Manning do that effectively."

gkrndija · Joined: 20 Feb 2005 Posts: 64

Kevin Pelton · Posted: Wed Jul 27, 2005 12:56 am Post subject:

One area where I do think football lends itself better to statistical analysis than basketball is coaching decision-making.

William Krasker at FootballCommentary.com has done some very interesting, very, very technical study of many of these decisions (go for it or kick? onside or kick it deep?) and some, notably a discussion of the value of timeouts, have basketball application.

kjb · Joined: 03 Jan 2005 Posts: 865 Location: Washington, DC

For the past couple years, I've been off and on playing online poker. The game is very much about stats (odds), but there's a ton of strategy and gamesmanship as well. Knowing the odds, being able to estimate them, is a huge help.

HoopStudies · Posted: Wed Jul 27, 2005 12:32 pm Post subject:

kjb · Joined: 03 Jan 2005 Posts: 865 Location: Washington, DC

In a massive tournament like the World Series of Poker has become, the winner has to get some luck. But, the same group of 15-20 professional players usually go pretty deep in these tournaments. If you look at the World Poker Tour events, for example, probably 4-5 guys made the final table (top 6) 2-3 times each. My view of poker is that it's basically a contest of decision-making abilities.

Dan Rosenbaum · Posted: Wed Jul 27, 2005 2:57 pm Post subject:

**Roland_Beech** · Joined: 14 Jan 2005 Posts: 43

there are some tools out there with poker, but the problem with hold 'em is that there isn't a lot to assess -- basically after the deal you know your two cards...then you add in the flops, it's not like on the tv shows where they know all the dealt cards and can calculate the hand odds as it goes along

moreover, with the online games you don't have any of the tells or behavior analysis other than the speed of response to a prompt, but I'm not sure how much you'd learn from that although I think that's the area I would focus on in building tools

on the football side I ~ahem~ have more than a few opinions since I've run the www.twominutewarning.com site since 1997. We've done various charting experiments in the past, the problem being frankly that you can't tell the detail of a lot of the action from a TV broadcast. We will be making a more serious charting attempt for this 2005 season, so again people with interests in game charting NFL action I would encourage to drop us a line...

There are a number of NFL teams that have substantial statistical analysis work going on, but they keep things very secret in pro football so you don't hear a lot about it through the normal channels.