20 Jun 2005

# Football Commentary: Clock-Management Value of Timeouts

With the help of a lot of math, William Krasker looks to answer the question "How valuable is a timeout, anyway?" The basic summary is that the clock-management value of a timeout is not particularly large in most situations, but is still large enough to call into question some of the common uses of timeouts, such as avoiding delay of game or discussing strategy. For a more advanced look, click on the link.

Posted by: Aaron Schatz on 20 Jun 2005

20 comments, Last at 22 Jun 2005, 4:33pm by Larry

1
by Larry (not verified) :: Mon, 06/20/2005 - 2:40pm

This brings up a thought I had reading the Game in Review column on hardballtimes.com recently. Namely, that at any point in a game you can calculate the win probability of the two teams based on the game state. Graphs of these probabilities as a function of time would be fascinating. Will has all the tools to do it, though this analysis obviously acknowledges the model isn't fully detailed. But it would be interesting nonetheless. Then players could be evaluated based on their ability to improve their teams chance of winning the game. I realize DVOA is this sort of analysis, but not exactly.

2
by Aaron (not verified) :: Mon, 06/20/2005 - 2:53pm

Yep, we've got an article on this in the book too, by Jim Armstrong. Lots of comments lately seem to bring up articles from the book. I swear we are NOT planting these!

3
by Vern (not verified) :: Mon, 06/20/2005 - 3:58pm

From his examples, seems like teams err on the side of the near term benefit, at the expense of a potentially better long term benefit. I think this is because of the perceived emotional benefit and "sense of momentum" that is confered to the players by short term success. This also, could/should be considered bad coaching. Too many times you see teams "deflate" over a 5 yd penalty at a key situation. They let the symbolism of failure at one moment become too important, instead of realizing that it only marignally adversely impacted their chance of winning - meaning they should continue to play at a high level.

Also, I would LOVE to see a running "chance of success" graphic during football games, but since FO doesn't run the networks, how about in post game analysis. As in, for some crucial moment in the game, say at the beginning of a final drive or, on key third or fourth down conversion attempt - what's the league average for success (score or down conversion) in that situation?

I found it very counter-inuitive that the difference in percent succcess between 3-10 and 3-15 was so small. I also think "comback drives" are often over- and under-rated, in that they are primarily a function of time, distance, and other game state variables. For example I think a comeback drive that started on the 35 with 2:30 and 3 timeouts should not be grouped in the same difficultly level with a drive that started on the 17 with 1:24 and one timeout left. But only seeing the percentages would say for sure.

4
by B (not verified) :: Mon, 06/20/2005 - 4:02pm

Vern: Sadly the network's version of post-game strategy analysis boils down to "If it worked, it was a great descision. If it didn't work, the it was a bad descision."

5
by Harris L (not verified) :: Mon, 06/20/2005 - 4:33pm

Re #4: I remember seeing the horrific Sean Salsbury on ESPN. He was recapping the previous day's football. This happened to be the week where the Vikings ran a reverse to Randy Mos who then threw a pick in the endzone. Salsbury was pissed. This was terrible play calling, etc. Then, when the Steelers recap came on, he praised the Steelers for the ingenious playcall that was a HB pass for the deciding touchdown. It's ridiculous how fickle the media can be. Particularly Sena Salsbury.

6
by Larry (not verified) :: Mon, 06/20/2005 - 4:40pm

B,

Ah, but we're not limited to the networks anymore. I'm guessing a lot of people who come here find that they get a lot more information about how and why teams win and lose than they get from network, newspaper or ESPN coverage. I watch ESPN for the highlights and physical feats, I come here to figure out which teams/players are playing well and why.

Its only a matter of time before someone starts streaming that sort of info live on the internet on gameday. Its only a step away from things like ESPN GameCast. And a box that does picture-in-picture with the internet can't be that far away. Ahhh, the possibilities.

To keep on topic, though, I'm surprised Time Outs have such little value. I guess the point is, that while TOs are fairly valuable down by a few points late in the game, the chance of ending up in that scenario is actually a lot smaller than is assumed. And from what Vern said, if the 5yds and a TO are approximately equal in value, and the 5 yds does cause psychological damage, then using the TO is the RIGHT thing to do, though the point about convincing your team the 5 yds don't matter is a good one.

7
by B (not verified) :: Mon, 06/20/2005 - 4:59pm

I think with timeouts it's better to hang onto them because while the chance that they will be useful at the end of the game is small, it's better to have a timeout and not need one, than to need a timeout and not have one.

8
by MRH (not verified) :: Mon, 06/20/2005 - 5:35pm

According to Aaron Schatz at Football Outsiders, the likelihood of making a first down is about 31% on 3rd and 10, and 22% on 3rd and 15

I'm sorry Mr. Krasker didn't link to this info here. Where can I find the article with this (3 and 10 vs. 3 and 15) success comparision? I failed abysmally at searching for it.

Do you have a table showing all the 5 yd penalty differences for various down and distances? Offense (3rd and 10 vs 15 yds) and defense (3rd and 6 vs 2nd and 11)?

I'd guess that 3 and 1 vs. 3 and 6 is a bigger difference - so avoiding the delay of game penalty in that situation is more important than on 3rd and 10.

9
by Catfish (not verified) :: Mon, 06/20/2005 - 6:12pm

MRH:

My guess is that the author emailed Aaron and asked for the info, which Aaron then gave him. I don't remember any success rate table for different downs and distances.

Aaron, sorry to go off topic, but what is the release date for the book?

10
by sippican (not verified) :: Mon, 06/20/2005 - 11:28pm

Now let's see-

S(t,d,T1,T2,1) = max{ p2pt R( t−1, d+2, T1, T2, 2) + (1−p2pt) R( t−1, d, T1, T2, 2) ,
p1pt R( t−1, d+1, T1, T2, 2) + (1−p1pt) R( t−1, d, T1, T2, 2) }.

Hmm. Well, I guess that clears that up. All those years I spent LOOKING AT THE GAME, and forming opinions...wasted.

Let me give you another formula that worked for me.

Patriots, minus Drew Bledsoe uselessly burning timeouts one minute into the third quarter because he can't read the defense, pulling his chinstrap off with a disgusted look, like it's someone else's job to know what to do at the line of scrimmage, walking to the sidelines, plus Mo Lewis to the ribcage, minus Bill Parcell's mindless devotion to aging veterans, plus Bill Belichick, times Tom Brady equals Lombardi Trophy cubed.

11
by Israel (not verified) :: Tue, 06/21/2005 - 2:11am

Re: #5
Conventional wisdom that day - Salisbury notwithstanding - was that Bettis had the option to run if his receiver (Jerame Tuman) was not totally open, while Moss did not seem to have that option.

12
by Harris L (not verified) :: Tue, 06/21/2005 - 6:02am

Re #11: The option to run!?!? Of course Randy had the option to run, but it was taken away from him. It was the same play call. You know what, even if your right, and there was some logic, Salsbury sure as heck didn't make mention of it. He just yells and gaurentees things without conveying either logic nor reasoning.

13
by Aaron (not verified) :: Tue, 06/21/2005 - 9:34am

Book will be out early August, hopefully August 1. And yeah, I just provided William with the info. I may make the percentage chance of getting first downs table available at some point.

14
by Dennis (not verified) :: Tue, 06/21/2005 - 3:09pm

Re #12: IIRC, the difference between the two plays (aside from their outcomes) was that the Vikings had been moving the ball very well whereas the Steelers weren't. It doesn't make a lot of sense to call a trick play when your regular offense is already getting the job done. A play like that is a better call when your offense is struggling because you have less to lose if it goes bad.

15
by zip (not verified) :: Tue, 06/21/2005 - 6:06pm

Re #10:

Ooh, the classic "belittling statistical anaysis due to its complexity" comment. Well done.

16
by Starshatterer (not verified) :: Wed, 06/22/2005 - 10:14am

Regarding the Randy Moss and Jerome Bettis option passes...

From a play-calling standpoint, they were probably about even. In terms of fooling the defense, an end-around to Moss is about as believable as a sweep to Bettis. Which is to say, Moss is better at catching the ball downfield and Bettis is better at pounding the ball between the tackles, but you still have to respect either one running the ball outside.

(#14 )--
I don't have enough examples to really analyze, but is there really a strong correlation either way between overall offensive success and trick-play success?

The difference in outcomes came down to two factors. First, Bettis's run outside drew the defenders up, which opened up the receiver, so the play worked as drawn. Second, after Moss's run outside failed to draw the defense out of coverage, Moss followed up with a poor decision to throw, rather than take his chances running or toss it out of bounds.

Ego may have played a role (Moss is the biggest star in the league in his own mind, double coverage should be no impediment to his throwing the decisive touchdown pass), which is what makes that particular play design questionable to me. You could have chosen another end to run the ball, then throw it to Moss pretending to block downfield. If it works, Moss still has a reason to TiVo SportsCenter, and any other receiver on the Vikings would have been more willing to go to Plan B (eat the ball or throw it away if the coverage is still in place).

17
by B (not verified) :: Wed, 06/22/2005 - 10:50am

Who's going to fall for a play that has Moss pretending to block? A better trick would have him pretend to walk off the field.

18
by zip (not verified) :: Wed, 06/22/2005 - 11:18am

#16
There's two reasons I can think of to run that play with Moss, but neither of them are very good.

1) Moss gets the most attention from the defence. Since the play requires the defenders dropping coverage to play the run, they may be more likely to do so if Moss has the ball over, say, Nate Burleson. Although Burleson (having seen him return punts) is probably a more dangerous runner. Like I said, these weren't very good reasons.

2) The WR actually has to throw the ball. Did Moss play QB in High School? Can the other WRs throw reasonably well?

also...

3) In the red zone, isn't the jump ball thrown to Moss from Culpepper disturbingly effective anyway? Screw that end-around option pass business. Pittsburgh, on the other hand, maybe doesn't have a guy like Moss to throw to?

19
by Starshatterer (not verified) :: Wed, 06/22/2005 - 3:56pm

B (#17 )--

If Moss is unconvincing enough as a blocker, it looks like a normal end-around. ;-)

Zip (#18 )--

1. I'm sure that was the logic: Moss with the ball would draw the defense.

2. According to a few sources, Moss played wide receiver, safety, punter, place kicker, kick returner, and punt returner in football, in addition to playing other sports in high school. No QB. He had attempted half a dozen passes (4/6, no TDs or INTs, 69 yards) as a pro in prior seasons, which made him a reasonable choice to throw, and also reasonable that the defense stayed in coverage.

3. Pittsburgh liked to throw the jump ball to Burress. Not quite as effective as Culpepper to Moss, but still a tactic you had to defend. In any case, I'm reasonably sure Tice & co. thought the element of surprise would work in their favor, compared with the almost-always-double-covered jump ball to Moss.

20
by Larry (not verified) :: Wed, 06/22/2005 - 4:33pm

Hey, shouldn't we be discussing the value of Time Outs here? Or at least quantatative analysis of game strategy? Or anything but Randy Moss?