Saturday, February 17, 2007
A statistical thought on "high percentage throws" and turnovers
I played a winter league game yesterday. It was a rather galling experience because I felt the four best players on the field were all on my team, and yet we lost by two. It was especially galling because two other Purdue alums were on the other team, and I don't like giving those guys bragging rights. This game was a great example of a truism in league play - it's more important to have the better worst player than the better best player. In this case, we also had the three worst players, so there you go.
Anyway, at one point in the game, one of the league vets on my team was preaching short throws, patience, and the usual pablum. This was significantly directed at me, for (as I am wont to do in league play) I was letting it fly quite a bit. It seemed like an odd criticism to me, though, as I was actually having a pretty good day throwing long. I think the focus on me resulted from the combination of us having a lot of long turnovers, and me having a lot of long throws. At that point in the game I had had two long throwaways, and I had one more (in my defense, a terrible misread by the receiver) later on. But I also threw for five of the eight goals we had in the game.
Significantly, I had three short throw turnovers that I remember as well. (I probably had one or two other turnovers that I don't recall.) Even though I disagreed with my teammate's comment, his comment led me to think about how one's turnovers are distributed and what this says about your decision process. Let's assume for argument's sake that there is not a huge upwind-downwind effect which would motivate a field position game or drastically slant one's huck counts based on the offensive direction. Given that, and given that your team wants to employ a deep threat, what does the distribution of your turnovers say about your choices?
My instinct is that the proper choices in throws, as a team, would lead to a fairly even division between deep throw turnovers and short throw turnovers. That is, a histogram of team turnover frequency versus distance would be flat. High percentage throws are only higher percentage if they produce fewer turnovers.
This does NOT imply that you throw deep as often as you throw short, unless you are equally successful at both throws. Say you turn the disc over 10% of the time on short throws and 40% of the time on deep throws. (Yes, I know those numbers suck - if I was talking about club I would set the percentages lower.) In that case, assuming your deep throw percentage is not a lot lower than a teammate's, you should probably be throwing deep on about one fifth of your throws. This means you will turn the disc over 16% of the time you throw, with an 8%/8% split between short and deep.
Of course, your "true" percentage is unknowable, and you could end up throwing more or less deep throws depending on the situations you end up in. But I think this serves as a useful guide for both individual players who throw deep a lot, and team captains. As a hucker, take note if your huck turnovers exceed your other turnovers. There are exceptions to this (for example, Lucy's role on Purdue in 2005 was to huck it half the time she touched the disc, which was fine) but it is a red flag. And as a team, if your deep turnovers exceed your short turnovers, then unless field position demands that strategy, it's time to scale it back.
Anyway, at one point in the game, one of the league vets on my team was preaching short throws, patience, and the usual pablum. This was significantly directed at me, for (as I am wont to do in league play) I was letting it fly quite a bit. It seemed like an odd criticism to me, though, as I was actually having a pretty good day throwing long. I think the focus on me resulted from the combination of us having a lot of long turnovers, and me having a lot of long throws. At that point in the game I had had two long throwaways, and I had one more (in my defense, a terrible misread by the receiver) later on. But I also threw for five of the eight goals we had in the game.
Significantly, I had three short throw turnovers that I remember as well. (I probably had one or two other turnovers that I don't recall.) Even though I disagreed with my teammate's comment, his comment led me to think about how one's turnovers are distributed and what this says about your decision process. Let's assume for argument's sake that there is not a huge upwind-downwind effect which would motivate a field position game or drastically slant one's huck counts based on the offensive direction. Given that, and given that your team wants to employ a deep threat, what does the distribution of your turnovers say about your choices?
My instinct is that the proper choices in throws, as a team, would lead to a fairly even division between deep throw turnovers and short throw turnovers. That is, a histogram of team turnover frequency versus distance would be flat. High percentage throws are only higher percentage if they produce fewer turnovers.
This does NOT imply that you throw deep as often as you throw short, unless you are equally successful at both throws. Say you turn the disc over 10% of the time on short throws and 40% of the time on deep throws. (Yes, I know those numbers suck - if I was talking about club I would set the percentages lower.) In that case, assuming your deep throw percentage is not a lot lower than a teammate's, you should probably be throwing deep on about one fifth of your throws. This means you will turn the disc over 16% of the time you throw, with an 8%/8% split between short and deep.
Of course, your "true" percentage is unknowable, and you could end up throwing more or less deep throws depending on the situations you end up in. But I think this serves as a useful guide for both individual players who throw deep a lot, and team captains. As a hucker, take note if your huck turnovers exceed your other turnovers. There are exceptions to this (for example, Lucy's role on Purdue in 2005 was to huck it half the time she touched the disc, which was fine) but it is a red flag. And as a team, if your deep turnovers exceed your short turnovers, then unless field position demands that strategy, it's time to scale it back.
# posted by Tarr : 2/17/2007 02:11:00 PM
Comments:
<< Home
Some (old, old) stats:
First column of numbers is long throwaways, second column is non-long throwaways, third column is drops.
TeamA 85 304 78
TeamB 36 162 49
TeamC 49 264 73
TeamD 76 422 123
TeamE 18 44 13
TeamA is Big Brother (DoG predecessor) in spring '92, TeamB is Big Brother in fall '92, TeamC is Earth Atomizer (mid-level Nats team) in fall '91, TeamD is Earth Atomizer in spring '91, TeamE is DoG in Nats and Worlds finals from '94-'99.
Earth ran a conservative offense, BB an aggressive, DoG in between. (Vintage DoG _would_ huck it, but we also looked to jam it up the line if that's what was open.)
So, those old stats have more of a ratio of 3 or 5 to 1, short to long throwaways, neglecting drops. I should point out that long passes were dropped a higher percentage of the time since they were more often contested. Overall drop rate was about 1.5%, but that was probably about 5% on contested passes and 0.5% on uncontested passes (which is to say that most dropped passes were on far from perfect throws, about 5/6 if the above estimates are correct).
The ultivillage ratio is about 5 to 1 the other way.
First column of numbers is long throwaways, second column is non-long throwaways, third column is drops.
TeamA 85 304 78
TeamB 36 162 49
TeamC 49 264 73
TeamD 76 422 123
TeamE 18 44 13
TeamA is Big Brother (DoG predecessor) in spring '92, TeamB is Big Brother in fall '92, TeamC is Earth Atomizer (mid-level Nats team) in fall '91, TeamD is Earth Atomizer in spring '91, TeamE is DoG in Nats and Worlds finals from '94-'99.
Earth ran a conservative offense, BB an aggressive, DoG in between. (Vintage DoG _would_ huck it, but we also looked to jam it up the line if that's what was open.)
So, those old stats have more of a ratio of 3 or 5 to 1, short to long throwaways, neglecting drops. I should point out that long passes were dropped a higher percentage of the time since they were more often contested. Overall drop rate was about 1.5%, but that was probably about 5% on contested passes and 0.5% on uncontested passes (which is to say that most dropped passes were on far from perfect throws, about 5/6 if the above estimates are correct).
The ultivillage ratio is about 5 to 1 the other way.
Thanks Jim, that's really interesting stuff. I guess I'm surprised by the distribution's slant toward short turnovers. Even if I give 3/4 of the drops to the long throws, none of those teams even reach a 1:2 long:short turnover ratio.
I think that it's possible the argument only goes the other way - that is, unless you are in a wind-influenced field position battle, turnovers should not be slanted more to long throws than to short. A team can still logically have more short turnovers than long as long as they are throwing long often enough to keep the defense honest. Then again, maybe not - maybe DoG ought to be throwing it up to Colin or blasting it out the back to Zip a bit more often.
Admittedly, my whole argument in this post is just intuitive and heuristic, as oppose to being based on a derivation of the approach that leads to the maximum likelihood of scoring.
I think that it's possible the argument only goes the other way - that is, unless you are in a wind-influenced field position battle, turnovers should not be slanted more to long throws than to short. A team can still logically have more short turnovers than long as long as they are throwing long often enough to keep the defense honest. Then again, maybe not - maybe DoG ought to be throwing it up to Colin or blasting it out the back to Zip a bit more often.
Admittedly, my whole argument in this post is just intuitive and heuristic, as oppose to being based on a derivation of the approach that leads to the maximum likelihood of scoring.
Not only that, but very few _individuals_ had less than a 2:1 short:long turnover ratio. Even the guy who was 4 for 16 on his hucks had 27 shorter turnovers as well.
Also, those teams completed about 93-95% of non-long and 60% of long passes (not counting drops). And those teams completed about 1 huck for every 5 goals. And no individual hucked on more than 7% of his passes. So weigh these last two bits of info when making comparisons.
Also, those teams completed about 93-95% of non-long and 60% of long passes (not counting drops). And those teams completed about 1 huck for every 5 goals. And no individual hucked on more than 7% of his passes. So weigh these last two bits of info when making comparisons.
Obviously, it's very hard to look at that data and draw solid conclusions about the decision processes on those teams. Again, more of a thought experiment. I remember you mentioning in 2004/early 2005 that you thought Pike had the best balance of agressiveness and conservativeness, which implied that you thought DoG was a bit too conservative, at least at the time.
The ultimate question here is, if one of those fairly conservative teams became incrementally less selective about which deep throws to take, or became more agressive about cutting deep, would they score more goals or fewer? Or, to put it a little less generally but more directly, would the increased number of deep turnovers be more or less than the reduced number of short turnovers?
The flipped version of this question ("...incrementally MORE selective...") would apply to HnH teams.
Post a Comment
The ultimate question here is, if one of those fairly conservative teams became incrementally less selective about which deep throws to take, or became more agressive about cutting deep, would they score more goals or fewer? Or, to put it a little less generally but more directly, would the increased number of deep turnovers be more or less than the reduced number of short turnovers?
The flipped version of this question ("...incrementally MORE selective...") would apply to HnH teams.
<< Home
