Monday, August 24, 2009

Mixing it up on offense

There are several ways to mix it up on offense so as to keep the defense honest. You can run different stacks, you can vary the emphasis on the long game, you can run set plays and counters, you can try motion offense or you can be deliberate, and other variations. How much do higher level teams do this and what form does this take?


In our heyday, DoG really didn't do a lot of this other than shifting the names in the four-person play. Then again, the offense was built around giving the cutter options and taking what was given, so any mistakes or adjustments the defense made automatically gave a new look to the offense. What made it work (beyond talented players) was that we could operate successfully in a couple different modes, the jam-it-up-the-line mode as well as cut-deep-and-put-it-out-there mode.

Now, we will shift from vertical to horizontal stack if it feels right, but we won't flip back and forth. Our long cuts are a little more deliberate, too, as it's kinda tiring to run deep if you're not going to get it.

But all this is a long way to get to the point of this post, how to rotate the usage of players for maximum effectiveness. Braess's Paradox says that in certain circumstances, closing a major road will actually lead to improved travel times. With the road open, it's in each person's best interest to take the main road, but it will lead to an increased average travel time. If instead a certain percentage took other routes, the lesser travel on the main road would prevent clogging and those who took it would be much better off, more than enough to make up for the extra time the others take.

This guy decided to apply these principles of networking to basketball, though the target could just as easily be ultimate. For a single possession, it makes sense to go to your best option, but if you go to it all the time, it will get clogged. I discussed this a little bit a few years ago, taking the idea of usage vs efficiency from the seminal work on basktetball analytics, "Basketball on Paper" by Dean Oliver (who also chimes in on a discussion of the linked article at APBRmetrics. The idea is that a player's efficiency decreases with increased usage (defined as % of a team's touches he gets when he's on the floor) as he has to shoot more on less than perfect looks and there is more defensive pressure applied.

So how do offenses deal with this, especially in the case where there is a clear "best play" (I'm less interested in what you do when every option is a very strong one)? Do ho stacks place their lane cutters in the same lanes as always but run the play to different people? (I'm presuming that teams call some sort of four-person play or a set play on every point; does anyone ever run complete free flow?) Do teams instead run counters, plays that appear to be the same as always but in fact the first cut from the main cutter is just a decoy? (In this case, you lose a bit of the benefit as the main cutter still gets fatigued.) What is the right amount of going to your lesser options in order to keep things honest?

10 comments:

Match said...

A prime example that pops into my head when I read this is Team Canada in the WUGC 2008 gold medal game. For the first time in a long time, Furious was without Jeff Cruikshank and I can remember thinking that this could actually be good for Canada.

Shank was an obvious threat at the handler position and Sockeye knew it very well. However, with him gone, Mauro and Derek Alexander ran things and I think Sockeye failed to adjust well.

With this in mind, I think having multiple weapons at the handler slot can be an advantage. Most handlers have their preferred tools such as give and goes, flick hucks, scoobers, whatever. However, if the heart of the offense can fluctuate, it would seem difficult to adjust downfield. Any cutter is a potential deep threat, but if you don't necessarily know where the disc is coming from, it is hard to know when you're about to get taken deep.

nice post Jim

Rich said...

I'm curious to see what others have to say, but I would bet the answer to this is very different between the elite teams and the non-elite teams. On elite teams, the 2nd or 3rd options are still great people to run the offense through or huck to or whatever. On 2nd tier teams, the difference between running a play with your best thrower and best cutter and not doing so (in the interest of changing it up) is probably much much greater.

Anyway - I think ideally you get your best players involved doing what they do best early on. Then go to some variations of plays and or personnel (or both) in the middle of the game, giving yourself the option to get back to your big guns to close things out (and having them not be gassed when it comes time to do so)

parinella said...

Even elite teams have a hierarchy, though, with some players much more prominent than others. How does GOAT, for example, vary its use of John Hassell?

Or strategically, how much variation is planned out? AJ (I think it was) once commented here or there that his team tried to huck the first three points of each game.
How quickly do defenses react so that there is an equilibrium? They start off playing neutral (which isn't to say that they play every team the same, only that they have some expectation of each opponent), then adjust to cover the deep game or to force middle or to shut down the hot hand.
We've tended to stay with what is working until the defense shows that they can stop it, but probably the better strategy would be to switch over before they stop it, so they are always reacting to how you were playing a couple points ago.

AJ said...

It's been a while since we a had an honest-to-god nerdy strategy post. Nice work.

On Chain, for several years our offensive plan was AJ is first cut, Cricket is second cut, if we're not in the end zone Dylan scores the goal. We'd run that probably 85 percent of the time...of course we each had the option to go under or away...so it did have some built in variation...these days have many more weapons, so it gives us the luxury of attacking teams in different ways...that being said we pretty much run 4 person plays every time but we are able to mix things up by running the plays out of 3 different sets and are able to use more cutters in different spots.

I think the Goat example is more interesting because they run so much of their offense through Hassell...Seems like bombing it to him on one of the first three possessions would make them even more dangerous.

As a coach, I've been scripting the first 5 offensive plays for each game for a few years now. In those plays, I go out of my way to go against our tendencies in 2 of the 5 plays(handlers downfield in the play, running counter-plays etc.)After the first five, I tend to either run the plays that were successful, or just fall into running our strongest plays.

My stat-keeping has really gone to hell in recent years, so while my sense is that the early game "anti-tendency" plays are generally more successful on average, I don't have the numbers to back it up.

parinella said...

AJ,

The Huddle has mostly taken the place of the strategic post.

Did you consider running other plays to save your legs? Were you calling the line?

Now, in your alternate sets, do the primary cutters position themselves in the same positions, so that a smart defender (say, an O player who broke his throwing hand) could know, "AJ's on the side in the ho stack, he's not primary"?

I tried to run two counter plays in our semis at Nats last year and the timing got messed up from good pulls. So based on a sample of two, I'm leery. But I suppose y'all have time to practice such things. And you have handlers who can go downfield.

b-lo said...

Back in Pike's heyday, we would cycle different people into the same pull play. We'd change up who was #2 if the other team starting getting down hard on a particular thrower (i can't explain why they didn't just get down hard on all our handlers). We'd also have a few different faces slot into the 3 cut. We ran counters to the extent if a team started provided deep help, the poached player would come under. nothing mind boggling.

We'd generally start with our "best option" personnel-wise early and switch away after a few points, and try end with our big guns at the end.

We didn't feel the quality of opposing D-teams warranted more than 1 pull play.

Interestingly, in college, we had more pull plays. Maybe that was because we had fewer good players or because we thought it would be cool to have a lot of plays.

Lou had an interesting write up in The Huddle saying Sockeye ran 15 pull plays with 40 in the playbook. I have a hard time imagining what those 15 plays would be.

Sport Management Steven said...

Jim,

Very nice post. I don't want to draw out any haters, but I always seem to enjoy your strategy stuff. You've either delved into some good literature on the topics, or you're a very insightful student of the game itself. (Probably a bit of both)

As for the huddle, I love it and I enjoy it as a read, but by no means should it be the end portal for strategy. Outside of some key writers, there is a lot lacking in the actual evidence/strategy part.

In terms of this article.. it seems like you're discussing two major issues that are bandied about in sport management/bio mechanics classes and by sport intelligence officers at the national team level- Strategy and the value of unpredictable in relation to reaction time.

I tried to summarize what smarter people have taught me in the following article.

http://canadaultimate.blogspot.com/2008/01/trainor-teaches-basics-of-team-strategy.html

Essentially, most ultimate teams tend to choose the GO strategy on Offense.Using Goat as an example, they probably say Mr Hassell is unstoppable, and operates on a higher efficiency than anything else we can use, so he gets the call early and often. Only when a team stops him does GOAT look elsewhere, if your posters are correct.

On the defensive side, teams in ultimate tend to go with the DENY strategy when they know the opponent or feel over matched. If they think they are better, they sometimes go with the Exploit/Go strategies.

Personally, I think the better stats you collect (video, not the pen and paper sideline joke) the better you'll really know how your team operates and what type of strategy you should use.

In terms of being unpredictable, it has a significant advantage in that it slows your opponents reaction time. This is why you'll notice complementary players in the NFL (Santonio Holmes) and NBA (Robert Horry) have the ability to make key plays when it matters most.

However, gains from creating such chaos only if you can carry out the strategy with high efficiency.

Bill Mill said...

A relevant paper is "The Price of Anarchy in Basketball": http://arxiv.org/abs/0908.1801

Its basic thesis is that Basketball players have "efficiency curves", which is to say that their scoring ability exhibits deminishing returns as the offense goes to them more often. Therefore, there's a team-optimal way to use a star player that may be different from each player's optimal strategy (the one where they all shoot enough that they shoot the same percentage).

Anyway, the point is, the paper suggests that underusing your star a bit may make him more effective (if you allow that ultimate has a similar "skill curve", which is a whole topic onto itself...)

parinella said...

Bill,
That's the first paper I linked to (well, it's the author's blog about the paper), and I said as much.

Steven,
I've seen each of the GO/DENY/EXPLOIT/PREVENT from both O and D. I think it often comes down to how big of a mismatch there is individually in choosing what to do.

Bill Mill said...

Doh! Sorry, I read the article right after it came out and the comment thread just yesterday... my bad.