Post by eric on May 3, 2018 18:14:52 GMT
Scoring efficiency is important. But how do we define it? Some field goals are worth two points, some are worth three points, so field goal percentage is out. And what about free throws? What about them, smart guy? What about them, smart guy?
One way of doing this is to take all points scored by whatever which way and divide it by the total number of possessions spent scoring them, or 'true shot attempts'. To do this we measure empirically the ratio between free throw attempts and possessions. If every free throw attempt occurred during a two shot situation, this would be exactly .5, but sometimes there are three shot fouls (which drive the value down towards .33), and sometimes there are and-ones (which drive it up towards 1). [N.B. There are no technical fouls in the software. Not bad.]
The best way to do this is to go through the play by play and count how many free throw situations of each kind there are, but we don't have PBP. What I'm trying out instead is taking every game box score, adding up field goal attempts and turnovers and subtracting a proportion of missed field goals based on rebounding percentage, then seeing what modifier to free throw attempts brings the difference in team possessions in each box score as close to zero as possible.
Sexually.
While that's ongoing, we can look at how different potential values will impact player evaluations. Let's take the most recent season (2001) and sort the top eight players in points scored by points per true shot attempt using the 4.0 value of .555:
Neon is way ahead, everyone else is pretty granular. It turns out that we'll have the same top three throughout this exercise so let's look at just four through eight. Here's how they look for values of .555, .5, .44, and .4:
Note that while all the numbers increase (necessarily as all their denominators decrease), some increases increase more than others. Players who play 'in the crease' so to speak, your big fellas who take more free throw attempts as a proportion of field goal attempts, see the most increase.
.
We should also look at the averages league wide. In 4.0 the positional averages were 1.06 for perimeter players and .96 for post players. Here's how they look again with fta mods of .555, .5, .44, and .4 for 2001 (the most recent season):
Again, big players climb relative to other positions the lower the modifier is. But the main takeaway here is that no matter what the mod is, everyone is pretty much at the average. There's no huge disparity like we had in 4.0. A player just needs to score efficiently.
I'll continue trying to nail down what the number actually is, whether we get PBPs or not, but so long as you use the same number across positions you can be reasonably sure you're evaluating relative differences in players correctly regardless of shot profile. Yes DeAndre Ayton goes from 8th most efficient to 4th, but relative to Brain Winter he only goes from -.02 to +.01, easily within the error bar due to random noise from measuring a single season (.04ish), and either way they're both phenomenally above average.
One way of doing this is to take all points scored by whatever which way and divide it by the total number of possessions spent scoring them, or 'true shot attempts'. To do this we measure empirically the ratio between free throw attempts and possessions. If every free throw attempt occurred during a two shot situation, this would be exactly .5, but sometimes there are three shot fouls (which drive the value down towards .33), and sometimes there are and-ones (which drive it up towards 1). [N.B. There are no technical fouls in the software. Not bad.]
The best way to do this is to go through the play by play and count how many free throw situations of each kind there are, but we don't have PBP. What I'm trying out instead is taking every game box score, adding up field goal attempts and turnovers and subtracting a proportion of missed field goals based on rebounding percentage, then seeing what modifier to free throw attempts brings the difference in team possessions in each box score as close to zero as possible.
Sexually.
While that's ongoing, we can look at how different potential values will impact player evaluations. Let's take the most recent season (2001) and sort the top eight players in points scored by points per true shot attempt using the 4.0 value of .555:
1.225 Neon Boudeaux
1.184 Firsto Picko
1.164 George Mikan
1.156 Brain Winter
1.144 Cameron Reddish
1.140 Elvis Delle Donne
1.138 Gary Bossert
1.137 DeAndre AytonNeon is way ahead, everyone else is pretty granular. It turns out that we'll have the same top three throughout this exercise so let's look at just four through eight. Here's how they look for values of .555, .5, .44, and .4:
1.156 Brain Winter 1.169 Brain Winter 1.187 DeAndre Ayton 1.206 DeAndre Ayton
1.144 Cameron Reddish 1.162 Cameron Reddish 1.184 Brain Winter 1.196 Elvis Delle Donne
1.140 Elvis D. Donne 1.161 DeAndre Ayton 1.182 Cameron Reddish 1.196 Cameron Reddish
1.138 Gary Bossert 1.159 Elvis D. Donne 1.181 Elvis D. Donne 1.194 Brain Winter
1.137 DeAndre Ayton 1.154 Gary Bossert 1.172 Gary Bossert 1.184 Gary BossertNote that while all the numbers increase (necessarily as all their denominators decrease), some increases increase more than others. Players who play 'in the crease' so to speak, your big fellas who take more free throw attempts as a proportion of field goal attempts, see the most increase.
.
We should also look at the averages league wide. In 4.0 the positional averages were 1.06 for perimeter players and .96 for post players. Here's how they look again with fta mods of .555, .5, .44, and .4 for 2001 (the most recent season):
0.555 0.5 0.44 0.4 position
1.03 1.04 1.05 1.06 point guard
1.06 1.08 1.09 1.10 shooting guard
1.04 1.05 1.07 1.08 small forward
1.05 1.07 1.09 1.11 power forward
1.02 1.04 1.07 1.08 center
1.04 1.06 1.07 1.09 leagueAgain, big players climb relative to other positions the lower the modifier is. But the main takeaway here is that no matter what the mod is, everyone is pretty much at the average. There's no huge disparity like we had in 4.0. A player just needs to score efficiently.
I'll continue trying to nail down what the number actually is, whether we get PBPs or not, but so long as you use the same number across positions you can be reasonably sure you're evaluating relative differences in players correctly regardless of shot profile. Yes DeAndre Ayton goes from 8th most efficient to 4th, but relative to Brain Winter he only goes from -.02 to +.01, easily within the error bar due to random noise from measuring a single season (.04ish), and either way they're both phenomenally above average.
