It's NBA Draft time. Which means we get to see the group think effect in full force.
So many mock drafts just with so little variance.
This blog post will attempt to provide something objective you can use to compare with the group think.
In the future I'd like to develop some systematic rankings that are smart about comparing international stats with college stats, and are able to consider multiple years of college stats, and adjust for injuries. Unfortunately, I have not yet done that, so my rankings will be focused on NCAA D1 players only, and will be based only on last year's stats.
Using individual player data from https://basketball.realgm.com/, team total statistics from https://www.sports-reference.com/cbb/seasons/2019-school-stats.html, and game data from https://kenpom.com/cbbga19.txt I've built a player ranking system.
An offensive simple box score (rawOSBS) value rating is calculated by taking Points Scored adjusted by the opportunity costs of Field Goal Attempts, Free Throw Attempts, and Turnovers, and then adding the value of creating additional possessions through offensive rebounds, and the value of assists determined through regression analysis. Opportunity Costs are calculated by determining the average points scored per possession for a school.
This raw offensive value is then adjusted by how efficient compared to average a player's team was overall in points per possession, what percentage of team possessions the player participated in, and how much better or worse than average a school's opponents were in allowing points per possession (adjusted for strength of schedule). This gives us an adjusted OSBS value.
A defensive simple box score (rawDSBS) value rating is then also calculated by looking at how a team performed relative to average in limiting opponent points per possession, what percentage of estimated defensive possessions a player was on the floor, calculating the value of a defensive rebound that ends an opponents possession (adjusted by average opponent offensive rebounding rate), the value of steal in both ending an opponents possession and starting a new offensive possession, the regressed calculated value of a block, and the opportunity costs created by fouling an opponent. Team performance is weighted more heavily here than in the offensive rating, especially since so much more of defensive value is not captured in box score stats like blocks, steals, and defensive rebounds.
This raw defensive value is then adjusted by how much better or worse than average a school's opponents were in scoring points per possession (adjusted for strength of schedule). This gives us an adjusted DSBS value.
OSBS and DSBS can measured per individual possession. They are then added together to give a player's simple box score rating (SBS/IP). Because different schools play different amount of games and at different paces, this is adjusted again by multiplying a players SBS/IP by the percentage of their schools possessions they played in (estimated by multiplying percentage of team minutes played by team possessions), and then multiplying again by the average number of total possessions played by D1 schools to give us a simple box score value rating (SBSVal). This number basically represents how many offensive points a player is worth plus how many points better than average on defense a player is worth, over the course of an average length season played at an average D1 pace against D1 average strength opponents.
After running this for all 4,811 D1 Players who RealGM recorded playing at least 1 minute, the top player is not surprising. Zion Williamson (Duke) comes in at #1. I ran a check on these numbers to make sure that, when added up for a team, it is representative of actual team strength. Top 10 teams by this measure are:
This looks like a pretty solid team ranking measure. Here are these teams final rankings in KenPom last year:
Gonzaga - 2
So many mock drafts just with so little variance.
This blog post will attempt to provide something objective you can use to compare with the group think.
In the future I'd like to develop some systematic rankings that are smart about comparing international stats with college stats, and are able to consider multiple years of college stats, and adjust for injuries. Unfortunately, I have not yet done that, so my rankings will be focused on NCAA D1 players only, and will be based only on last year's stats.
Using individual player data from https://basketball.realgm.com/, team total statistics from https://www.sports-reference.com/cbb/seasons/2019-school-stats.html, and game data from https://kenpom.com/cbbga19.txt I've built a player ranking system.
An offensive simple box score (rawOSBS) value rating is calculated by taking Points Scored adjusted by the opportunity costs of Field Goal Attempts, Free Throw Attempts, and Turnovers, and then adding the value of creating additional possessions through offensive rebounds, and the value of assists determined through regression analysis. Opportunity Costs are calculated by determining the average points scored per possession for a school.
This raw offensive value is then adjusted by how efficient compared to average a player's team was overall in points per possession, what percentage of team possessions the player participated in, and how much better or worse than average a school's opponents were in allowing points per possession (adjusted for strength of schedule). This gives us an adjusted OSBS value.
A defensive simple box score (rawDSBS) value rating is then also calculated by looking at how a team performed relative to average in limiting opponent points per possession, what percentage of estimated defensive possessions a player was on the floor, calculating the value of a defensive rebound that ends an opponents possession (adjusted by average opponent offensive rebounding rate), the value of steal in both ending an opponents possession and starting a new offensive possession, the regressed calculated value of a block, and the opportunity costs created by fouling an opponent. Team performance is weighted more heavily here than in the offensive rating, especially since so much more of defensive value is not captured in box score stats like blocks, steals, and defensive rebounds.
This raw defensive value is then adjusted by how much better or worse than average a school's opponents were in scoring points per possession (adjusted for strength of schedule). This gives us an adjusted DSBS value.
OSBS and DSBS can measured per individual possession. They are then added together to give a player's simple box score rating (SBS/IP). Because different schools play different amount of games and at different paces, this is adjusted again by multiplying a players SBS/IP by the percentage of their schools possessions they played in (estimated by multiplying percentage of team minutes played by team possessions), and then multiplying again by the average number of total possessions played by D1 schools to give us a simple box score value rating (SBSVal). This number basically represents how many offensive points a player is worth plus how many points better than average on defense a player is worth, over the course of an average length season played at an average D1 pace against D1 average strength opponents.
After running this for all 4,811 D1 Players who RealGM recorded playing at least 1 minute, the top player is not surprising. Zion Williamson (Duke) comes in at #1. I ran a check on these numbers to make sure that, when added up for a team, it is representative of actual team strength. Top 10 teams by this measure are:
| Gonzaga | 4,710.76 |
| Duke | 4,573.72 |
| Virginia | 4,503.68 |
| Michigan St. | 4,499.23 |
| Tennessee | 4,202.39 |
| North Carolina | 4,174.30 |
| Purdue | 4,125.71 |
| Auburn | 3,901.85 |
| Iowa St. | 3,884.10 |
| Kentucky | 3,880.20 |
This looks like a pretty solid team ranking measure. Here are these teams final rankings in KenPom last year:
Gonzaga - 2
Duke - 4
Virginia - 1
Michigan St. - 3
Tennessee - 10
North Carolina - 7
Purdue - 9
Auburn - 11
Iowa St. - 15
Kentucky - 8
So given that these player ratings check out as being good collective representatives of team value, let's see who the top 60 draft eligible players are by this rating:
This might be a decent cheat sheet if you had a time machine and went back to draft players last year and you were only allowed to have them for one season. But what NBA teams are trying to do is project future value. All these guys are not the same age. So if you want to project forward, you have to make estimates as to how much the players will progress from the current value they are generating.
In order to calculate progression coefficients for each player, I ran data on the average value of NBA players at each age (combining WS, VORP, and my own NBA player rating) using data from basketball-reference.com going back to 1980. Using this data, I was able to create a best fit curve of progress for the average basketball player based on age (based on player age at February 1st of the season). As you can see, a player has a large increase in performance between 18 and 19 (64% increase) but that increase drops each year until it peaks at Age 28 and then begins to decline. Where a player is on this curve over the next several years can have a big impact on his value.
So given that these player ratings check out as being good collective representatives of team value, let's see who the top 60 draft eligible players are by this rating:
This might be a decent cheat sheet if you had a time machine and went back to draft players last year and you were only allowed to have them for one season. But what NBA teams are trying to do is project future value. All these guys are not the same age. So if you want to project forward, you have to make estimates as to how much the players will progress from the current value they are generating.
In order to calculate progression coefficients for each player, I ran data on the average value of NBA players at each age (combining WS, VORP, and my own NBA player rating) using data from basketball-reference.com going back to 1980. Using this data, I was able to create a best fit curve of progress for the average basketball player based on age (based on player age at February 1st of the season). As you can see, a player has a large increase in performance between 18 and 19 (64% increase) but that increase drops each year until it peaks at Age 28 and then begins to decline. Where a player is on this curve over the next several years can have a big impact on his value.
| AGE | Average of APM | Best Fit | %Fact |
| 18 | 2.38 | 3.80 | |
| 19 | 7.46 | 6.24 | 0.64 |
| 20 | 10.14 | 8.24 | 0.32 |
| 21 | 11.18 | 9.86 | 0.20 |
| 22 | 9.56 | 11.15 | 0.13 |
| 23 | 9.67 | 12.15 | 0.09 |
| 24 | 11.34 | 12.90 | 0.06 |
| 25 | 13.35 | 13.44 | 0.04 |
| 26 | 14.44 | 13.80 | 0.03 |
| 27 | 15.02 | 14.01 | 0.02 |
| 28 | 15.66 | 14.09 | 0.01 |
| 29 | 15.04 | 14.06 | 0.00 |
| 30 | 14.44 | 13.93 | -0.01 |
| 31 | 13.41 | 13.71 | -0.02 |
| 32 | 12.88 | 13.42 | -0.02 |
| 33 | 12.37 | 13.05 | -0.03 |
| 34 | 12.16 | 12.60 | -0.03 |
| 35 | 11.96 | 12.07 | -0.04 |
| 36 | 11.24 | 11.45 | -0.05 |
| 37 | 10.11 | 10.72 | -0.06 |
| 38 | 8.84 | 9.87 | -0.08 |
| 39 | 9.92 | 8.87 | -0.10 |
| 40 | 8.46 | 7.70 | -0.13 |
| 41 | 7.06 | 6.32 | |
| 42 | 4.51 | 4.70 |
I was also able to calculate the average length of an NBA career over that time period - 5.45 years.
Given this, I decided on projecting out 5 years of value given the initial starting point of the value they generated in college last year and adding on the expected value increase over each of the next five years given a players age. The age for a given season is based on how old the player is on February 1st of that season. Doing this, I generated the following new Top 60 list among draft eligible players:
Given this, I decided on projecting out 5 years of value given the initial starting point of the value they generated in college last year and adding on the expected value increase over each of the next five years given a players age. The age for a given season is based on how old the player is on February 1st of that season. Doing this, I generated the following new Top 60 list among draft eligible players:
As you may be able to tell, this new list more closely resembles mock drafts you may have seen. The top 3 players here are the top 3 consensus picks in the draft (though Morant and Barrett are flipped). There appears to be an analytical backing to the group consensus at the very top. Below is a list of all the players included in and ranked by average position (MDA) in the 14 mock drafts I found online from popular analysts. You can use to compare against their five year value (5YV) and rank (5YVrnk*).
*Note: Five Year Value Rank on this chart is based on rank among all players, not just draft eligible players.
Some notes:
By my system, Darius Garland's mock draft position does not make much sense. He was decently ranked coming out of high school (14 on 247sports composite ranking), but has surprisingly shot up the rankings despite being hurt and not playing great. He only played in 5 games before his injury, and in those games he was not super efficient on offense (only 43rd among the 75 college players who appear in one of the 14 mock drafts I tracked in per possession terms - OSBS/IP), and even worse defensively (66th among 75 mock draft players).
Cam Reddish is another player who my system would suggest is overrated. He was not injured and still managed only anemic efficiency (dead last among the 75 mock draft players in OSBS/IP - and only shot 33.8% from 3 despite opposing defenses having all their focus on stopping Zion and R.J.)
Bol Bol, like Garland, was also hurt early in the season last year, but unlike Garland, actually performed quite well while he was on the court. His offensive efficiency (1.59 OSBS/IP) was much higher than Garland's (1.02 OSBS/IP) and Bol's defensive efficency (1.38 DSBS/IP) was actually the highest of any of the 75 mock draft players. If he can stay healthy and maintain that efficiency with a higher workload, he might be one of the steals of the draft.
Kevin Porter Jr. is another player whose mock draft stock does not make much sense to me. He was the 28th ranked player coming out of high school last year on 247sports composite, and then did not have a very good season at all. He played in 21 games but only managed an overall SBS/IP of 1.19, tied for the 3rd worst among the 75 mock players. His 5 year projected value ranks 1,180 of the 4,811 D1 players, not exactly 1st round material.
KZ Okpala and Luguentz Dort are two other stay-aways for me. They are the two players among the 75 mock draft players that actually had a lower overall efficiency rating than Kevin Porter.
Matisse Thybulle is an interesting case. His offensive numbers are very average but his defensive numbers are off the charts (1.36 DSBS/IP). Only Bol Bol had better defenisve efficiency among the 75 mock players. Thybulle, though, maintained this efficiency while playing in 15.5% of his team's possessions (Bol only played in 3.5% of his teams possessions). Thybulle has the highest Defensive Efficiency number in the country among all D1 players who played at least 5% of their team's possessions.
Grant Williams and Bruno Fernando are two players who I think will provide strong value to whoever drafts them. Williams is #4 among all players in projected 5 year value and Fernando is #10, but both players are going end of 1st/beginning of 2nd round in most mock drafts.
Shamorie Ponds and Dedric Lawson are two players I think are being over corrected for their upperclassman status. They both just finished their Junior years, but were extremely effective and efficient basketball players despite the relatively large loads they carried for their teams. Ponds is 22nd in projected 5 year value but most mocks have him middle to late 2nd round. Lawson is 13th in projected 5 year value but is only even included on 2 of the 14 mock drafts I'm tracking.
Matt Raferty of Furman (#9), Juwan Morgan of Indiana (#31), Jordan Murphy of Minnesota (#40), and Chris Clemons of Campbell (#50) are the 4 players who appear in the top 60 of my 5 year projected value rankings but nevertheless did not appear in any of the mock drafts I tracked. They might not get drafted, but if they get fair shots in Summer League or in the G-League, expect them to eventually make some noise in the NBA.
Spreadsheet with all of the data can be found here.
*Note: Five Year Value Rank on this chart is based on rank among all players, not just draft eligible players.
Some notes:
By my system, Darius Garland's mock draft position does not make much sense. He was decently ranked coming out of high school (14 on 247sports composite ranking), but has surprisingly shot up the rankings despite being hurt and not playing great. He only played in 5 games before his injury, and in those games he was not super efficient on offense (only 43rd among the 75 college players who appear in one of the 14 mock drafts I tracked in per possession terms - OSBS/IP), and even worse defensively (66th among 75 mock draft players).
Cam Reddish is another player who my system would suggest is overrated. He was not injured and still managed only anemic efficiency (dead last among the 75 mock draft players in OSBS/IP - and only shot 33.8% from 3 despite opposing defenses having all their focus on stopping Zion and R.J.)
Bol Bol, like Garland, was also hurt early in the season last year, but unlike Garland, actually performed quite well while he was on the court. His offensive efficiency (1.59 OSBS/IP) was much higher than Garland's (1.02 OSBS/IP) and Bol's defensive efficency (1.38 DSBS/IP) was actually the highest of any of the 75 mock draft players. If he can stay healthy and maintain that efficiency with a higher workload, he might be one of the steals of the draft.
Kevin Porter Jr. is another player whose mock draft stock does not make much sense to me. He was the 28th ranked player coming out of high school last year on 247sports composite, and then did not have a very good season at all. He played in 21 games but only managed an overall SBS/IP of 1.19, tied for the 3rd worst among the 75 mock players. His 5 year projected value ranks 1,180 of the 4,811 D1 players, not exactly 1st round material.
KZ Okpala and Luguentz Dort are two other stay-aways for me. They are the two players among the 75 mock draft players that actually had a lower overall efficiency rating than Kevin Porter.
Matisse Thybulle is an interesting case. His offensive numbers are very average but his defensive numbers are off the charts (1.36 DSBS/IP). Only Bol Bol had better defenisve efficiency among the 75 mock players. Thybulle, though, maintained this efficiency while playing in 15.5% of his team's possessions (Bol only played in 3.5% of his teams possessions). Thybulle has the highest Defensive Efficiency number in the country among all D1 players who played at least 5% of their team's possessions.
Grant Williams and Bruno Fernando are two players who I think will provide strong value to whoever drafts them. Williams is #4 among all players in projected 5 year value and Fernando is #10, but both players are going end of 1st/beginning of 2nd round in most mock drafts.
Shamorie Ponds and Dedric Lawson are two players I think are being over corrected for their upperclassman status. They both just finished their Junior years, but were extremely effective and efficient basketball players despite the relatively large loads they carried for their teams. Ponds is 22nd in projected 5 year value but most mocks have him middle to late 2nd round. Lawson is 13th in projected 5 year value but is only even included on 2 of the 14 mock drafts I'm tracking.
Matt Raferty of Furman (#9), Juwan Morgan of Indiana (#31), Jordan Murphy of Minnesota (#40), and Chris Clemons of Campbell (#50) are the 4 players who appear in the top 60 of my 5 year projected value rankings but nevertheless did not appear in any of the mock drafts I tracked. They might not get drafted, but if they get fair shots in Summer League or in the G-League, expect them to eventually make some noise in the NBA.
Spreadsheet with all of the data can be found here.
Comments
Post a Comment
Practice Respect!