The Value of Tanking
Posted: Thu Nov 11, 2010 3:02 pm
Now that we're officially a tanking team, this might not inspire the philosophical debate that it might have when I first set out to do this back in the spring (I then forgot about it) when we had visions of low playoff seeds dancing in our heads, but now that we're racing to the bottom it's good to know what we're racing toward, and the value of each additional loss along the way.
To that end, I've set about to get a rough approximation of the value of each pick, using data from every player selected between 1st and 14th overall between 1990-2004. I've used that date range because I'm looking at the win shares from their first five seasons in the league...the time frame in which the drafting team has control of the player's rights, before they bolt to some sunny locale (like New Jersey) and break our little hearts.
Why win shares? Because, while it's hardly a perfect metric no such thing exists, so bugger off. Win shares are however neat, easy to enter into a spreadsheet, and do a good enough job at telling the story when you have 75 data points (15 players x 5 years apiece) for each draft and you're only looking at the aggregate. So win shares it is.
This is looking at the first five seasons after a player is drafted, without consideration for injuries, going overseas, whatever. This is for a couple reasons: one, we're looking at the value provided by the player to their team(s), and two, there's no way to get around that stuff without adding a thick lather of subjectivity. Yeah, Amar'e has his win share total skewed a bit because he missed the vast majority of his fourth season with injury, but if one were to skip that season or discount it somehow, you'd then need to go player-by-player and do likewise, adding a judgment call each time. Then you'd have to determine whether Scrub X really was as injured as the team said they were back in the days when the Inactive List was the Injured List, or whether Star Y's bad season should be discounted because he played injured (like a champion!) from November on after that freak decapitation he suffered. No need to combat inaccuracy with more inaccuracy. And I'm lazy. So once I expand this to include playoff picks, Freddy Weis gets the big goose egg he so skilfully earned.
Don't get massively hung up on the win shares produced by individual draft slots...there's still a decent amount of variation from one spot to another, because basketball players are an unruly lot who run the gamut from great to atrocious, and each draft slot has at least one of each. Except for the 12th pick, which is where dreams go to die for some reason. What we're looking at is the trend and the points at which a serious dropoff in expected production can be seen.
And yeah, I know that there was a lockout during one of the seasons that skews the win shares for those players who suffered through it. Don't matter, because we're comparing the 1st pick to the 2nd to the 13th, and as such each group is affected. Also, bugger off.
So there are essentially two things that have to be looked at here. First is the average value, in win shares, of each player taken in each slot. But that's only a piece o' the puzzle...when you're tanking, it's the number of ping pong balls that you're chasing, with the hope of a draft pick only following once it's all said and done. Thus, I've created a version weighted for the lottery odds...that's the expected value, in win shares, of each lottery slot. This is not weighted by the past lottery results, just the odds; it doesn't matter whether the team with the fifth-most lotto balls has won the lottery a disproportionate number of times, because we're dealing with the actual odds of it happening.
Raw average win shares over five-year period:
There are a few things to take notice of here:
- The first overall pick is very, very tasty...about 30% better production, on average, than any other slot. This is something that is fairly well-acknowledged, but the contrast is still pretty stark.
- The second and third picks don't hold a candle to the first pick, but there's a sizable drop-off in expected value beyond there.
- From picks 6 through 14, there is considerable variance from slot to slot, to such an extent that if I were to fit a curve to this (which I would do, if I wasn't inept with such things, and which someone will hopefully do for me) it would look pretty random. And while it's tempting to assume that the 9th and 13th picks are imbued with some magical property, and the 6th and 12th picks suffering a gypsy curse, the reality is pretty simple: after the top five in most drafts, you're looking at a crapshoot. There might be one or two excellent players thereafter...or there might be none. The average drop in talent after the fifth pick is pretty significant, though; while a 4th/5th pick brought home around 22.5 win shares on average from '90-'04, the 6th through 14th picks were good for only 14.3.
And that's the first big lesson of tanking: if you have to suck, don't half-ass it. Aim to be very, very bad, because you really need that top-5 pick (and top-3 if you can help it).
Now on to the pre-lottery value table. The first thing that you'll notice is that there isn't nearly as much variance at the top of the chart, for two reasons: because there isn't a massive drop in lottery odds from the worst record to the second- and third-worst (-5.1% from 1st to 2nd, -4.3% from 2nd to 3rd), and because the top picks either have no chance or very little chance of falling into the Forbidden Zone known as the mid/late lottery. Anyhoo, here are the weighted totals for expected win shares, pre-lottery:
Expected win shares entering the lottery, based upon reverse order of finish:
So there isn't a huge difference between the worst record and second-worst, or the second-worst and third-worst. But there is enough of a difference to make it worthwhile...if it's late and the season and you're battling for ping pong balls, giving Mark Madsen the green light to shoot three-pointers is still a good idea.
Here, though, the groupings are a little bit different than before weighting for the lottery. The top three are bunched relatively tightly, fourth (the first slot where falling out of the top five isn't seriously unlikely, at 16%) is a wee bit behind with fifth a bit behind that, and there isn't a tonne of separation between sixth and tenth (which average 18.2 expected win shares), when the odds of vaulting into the top three become quite a bit lower and the pickins if one fails to do so become just a bit more slim. And after that, the lepers of the lottery from 11th on down, which average just 12.2 expected win shares.
So once we've smoothed things out with the lottery odds, it becomes pretty clear that:
- You can be pretty happy with your tanking effort, and have a good expectation of value, if you finish with one of the three worst records in the league. After that, things will start to get a little more dicey. On average over the last ten seasons, the third-worst team in the league has had a record of 23-59; the fourth-worst team averaged 25-57, but dipped as low as 22-60. Records beyond that can bunch up as well, meaning that a successful homestand near the end of the year can potentially slice off almost one-third of your expected draft value as you plummet from third-worst to sixth-worst. So never, ever take for granted that you'll out-lose the competition...in a down year, you're really going to have to turn up the suck to get the most value out of those empty seats and disenchanted players.
- If you go after the playoffs and fail, or have an unexpectedly good (by which I mean mediocre) season and finish in the late-lottery, chances are that you are boned. Diamonds do appear in the rough, but you're much more likely to get bench filler or an outright bust. I suspect that this is magnified now that 18 year olds aren't available in the draft...while Tracy McGrady, Amar'e Stoudemire and Dirk Nowitzki fell to 9 because of the question marks attached, all three would now get more seasoning before entering the draft, and thus would be more likely to assuage the concerns that caused them to drop.
- There's just no such thing as too many losses.
If people want, I can make the spreadsheet publicly-available...once someone tells me how to do that with Google Documents while keeping it read-only. Because if I allow you hooligans to edit it, ever player will be named Little Ozzy and the average win share will be PENIS!
To that end, I've set about to get a rough approximation of the value of each pick, using data from every player selected between 1st and 14th overall between 1990-2004. I've used that date range because I'm looking at the win shares from their first five seasons in the league...the time frame in which the drafting team has control of the player's rights, before they bolt to some sunny locale (like New Jersey) and break our little hearts.
Why win shares? Because, while it's hardly a perfect metric no such thing exists, so bugger off. Win shares are however neat, easy to enter into a spreadsheet, and do a good enough job at telling the story when you have 75 data points (15 players x 5 years apiece) for each draft and you're only looking at the aggregate. So win shares it is.
This is looking at the first five seasons after a player is drafted, without consideration for injuries, going overseas, whatever. This is for a couple reasons: one, we're looking at the value provided by the player to their team(s), and two, there's no way to get around that stuff without adding a thick lather of subjectivity. Yeah, Amar'e has his win share total skewed a bit because he missed the vast majority of his fourth season with injury, but if one were to skip that season or discount it somehow, you'd then need to go player-by-player and do likewise, adding a judgment call each time. Then you'd have to determine whether Scrub X really was as injured as the team said they were back in the days when the Inactive List was the Injured List, or whether Star Y's bad season should be discounted because he played injured (like a champion!) from November on after that freak decapitation he suffered. No need to combat inaccuracy with more inaccuracy. And I'm lazy. So once I expand this to include playoff picks, Freddy Weis gets the big goose egg he so skilfully earned.
Don't get massively hung up on the win shares produced by individual draft slots...there's still a decent amount of variation from one spot to another, because basketball players are an unruly lot who run the gamut from great to atrocious, and each draft slot has at least one of each. Except for the 12th pick, which is where dreams go to die for some reason. What we're looking at is the trend and the points at which a serious dropoff in expected production can be seen.
And yeah, I know that there was a lockout during one of the seasons that skews the win shares for those players who suffered through it. Don't matter, because we're comparing the 1st pick to the 2nd to the 13th, and as such each group is affected. Also, bugger off.
So there are essentially two things that have to be looked at here. First is the average value, in win shares, of each player taken in each slot. But that's only a piece o' the puzzle...when you're tanking, it's the number of ping pong balls that you're chasing, with the hope of a draft pick only following once it's all said and done. Thus, I've created a version weighted for the lottery odds...that's the expected value, in win shares, of each lottery slot. This is not weighted by the past lottery results, just the odds; it doesn't matter whether the team with the fifth-most lotto balls has won the lottery a disproportionate number of times, because we're dealing with the actual odds of it happening.
Raw average win shares over five-year period:
1st pick: 37.1 win shares.
2nd pick: 28.7.
3rd pick: 28.1.
4th pick: 20.5.
5th pick: 24.4.
6th pick: 14.1.
7th pick: 16.0.
8th pick: 13.5.
9th pick: 21.4.
10th pick: 16.7.
11th pick: 12.5.
12th pick: 8.9.
13th pick: 17.5.
14th pick: 7.9.
There are a few things to take notice of here:
- The first overall pick is very, very tasty...about 30% better production, on average, than any other slot. This is something that is fairly well-acknowledged, but the contrast is still pretty stark.
- The second and third picks don't hold a candle to the first pick, but there's a sizable drop-off in expected value beyond there.
- From picks 6 through 14, there is considerable variance from slot to slot, to such an extent that if I were to fit a curve to this (which I would do, if I wasn't inept with such things, and which someone will hopefully do for me) it would look pretty random. And while it's tempting to assume that the 9th and 13th picks are imbued with some magical property, and the 6th and 12th picks suffering a gypsy curse, the reality is pretty simple: after the top five in most drafts, you're looking at a crapshoot. There might be one or two excellent players thereafter...or there might be none. The average drop in talent after the fifth pick is pretty significant, though; while a 4th/5th pick brought home around 22.5 win shares on average from '90-'04, the 6th through 14th picks were good for only 14.3.
And that's the first big lesson of tanking: if you have to suck, don't half-ass it. Aim to be very, very bad, because you really need that top-5 pick (and top-3 if you can help it).
Now on to the pre-lottery value table. The first thing that you'll notice is that there isn't nearly as much variance at the top of the chart, for two reasons: because there isn't a massive drop in lottery odds from the worst record to the second- and third-worst (-5.1% from 1st to 2nd, -4.3% from 2nd to 3rd), and because the top picks either have no chance or very little chance of falling into the Forbidden Zone known as the mid/late lottery. Anyhoo, here are the weighted totals for expected win shares, pre-lottery:
Expected win shares entering the lottery, based upon reverse order of finish:
1st: 27.7 win shares.
2nd: 27.1.
3rd: 26.3.
4th: 24.8.
5th: 21.9.
6th: 18.3.
7th: 17.7.
8th: 16.6.
9th: 21.4.
10th: 16.9.
11th: 12.8.
12th: 9.8.
13th: 17.6.
14th: 8.4.
So there isn't a huge difference between the worst record and second-worst, or the second-worst and third-worst. But there is enough of a difference to make it worthwhile...if it's late and the season and you're battling for ping pong balls, giving Mark Madsen the green light to shoot three-pointers is still a good idea.
Here, though, the groupings are a little bit different than before weighting for the lottery. The top three are bunched relatively tightly, fourth (the first slot where falling out of the top five isn't seriously unlikely, at 16%) is a wee bit behind with fifth a bit behind that, and there isn't a tonne of separation between sixth and tenth (which average 18.2 expected win shares), when the odds of vaulting into the top three become quite a bit lower and the pickins if one fails to do so become just a bit more slim. And after that, the lepers of the lottery from 11th on down, which average just 12.2 expected win shares.
So once we've smoothed things out with the lottery odds, it becomes pretty clear that:
- You can be pretty happy with your tanking effort, and have a good expectation of value, if you finish with one of the three worst records in the league. After that, things will start to get a little more dicey. On average over the last ten seasons, the third-worst team in the league has had a record of 23-59; the fourth-worst team averaged 25-57, but dipped as low as 22-60. Records beyond that can bunch up as well, meaning that a successful homestand near the end of the year can potentially slice off almost one-third of your expected draft value as you plummet from third-worst to sixth-worst. So never, ever take for granted that you'll out-lose the competition...in a down year, you're really going to have to turn up the suck to get the most value out of those empty seats and disenchanted players.
- If you go after the playoffs and fail, or have an unexpectedly good (by which I mean mediocre) season and finish in the late-lottery, chances are that you are boned. Diamonds do appear in the rough, but you're much more likely to get bench filler or an outright bust. I suspect that this is magnified now that 18 year olds aren't available in the draft...while Tracy McGrady, Amar'e Stoudemire and Dirk Nowitzki fell to 9 because of the question marks attached, all three would now get more seasoning before entering the draft, and thus would be more likely to assuage the concerns that caused them to drop.
- There's just no such thing as too many losses.
If people want, I can make the spreadsheet publicly-available...once someone tells me how to do that with Google Documents while keeping it read-only. Because if I allow you hooligans to edit it, ever player will be named Little Ozzy and the average win share will be PENIS!
