Wednesday, January 31, 2018

Value Add: Original Position Adjustment (No Longer Used)

2,121 Most Valuable College Basketball Players in 2012

I believe our stat team (sorry they never want to be mentioned by name, but I’ve found researchers, data entry people, programmers and engineers are sometimes like that) has nailed how to account for perimeter defense and point guard play within the Value Add system. We just ran the program, and you can look up any team, such as the opponent drawn in the NCAA on Selection Sunday, to see where there players are ranked by clicking on the following four links (NOTE: on March 8, some updates were made to correct a few players on the following 4 links, mainly the 37 players who share a name with another Division I player - such as 53rd ranked Dion Waiters - who shares a name with a player on Jacksonville State.):

Air Force to Furman ranked players

Gardner Webb to Marist ranked players

Marquette to Syracuse ranked players

Temple to Youngstown State ranked players

The 20 most valuable players in the country are listed below, with an explanation under the table:


RnkPlayerTeamHtYrOffDefPG/PerTotNBA?
1Anthony DavisKentucky6'10"Fr7.06-5.31 12.361st
2Jae CrowderMarquette6'6"Sr6.11-4.45 10.562nd
3Draymond GreenMichigan St.6'7"Sr4.27-4.041.009.312nd
4Kevin JonesWest Virginia6'8"Sr7.86-1.38 9.242nd
5Cody ZellerIndiana6'11"Fr6.69-2.08 8.762013
6Thomas RobinsonKansas6'9"Jr4.54-3.87 8.401st
7Will BartonMemphis6'6"So5.79-2.37 8.162nd
8John ShurnaNorthwestern6'9"Sr7.21-0.91 8.12 
9Jared SullingerOhio St.6'9"So5.05-3.01 8.061st
10Jordan TaylorWisconsin6'1"Sr5.20-1.281.507.982nd
11T.J. McConnellDuquesne6'1"So3.40-3.011.57.91 
12C.J. McCollumLehigh6'3"Jr4.46-2.4417.902013
13Marcus DenmonMissouri6'3"Sr7.39-0.34 7.732nd
14Isaiah CanaanMurray St.6'0"Jr6.16-0.5417.702013
15John JenkinsVanderbilt6'4"Jr7.110.000.57.611st
16Damian LillardWeber St.6'2"Jr7.591.0017.591st
17Tyler ZellerNorth Carolina7'0"Sr5.32-2.22 7.541st
18Doug McDermottCreighton6'7"So7.420.00 7.422013
19Mike ScottVirginia6'8"Sr5.70-1.72 7.422nd
20Scott MachadoIona6'1"Sr5.23-0.1627.402nd



I’ve run the Offensive and Defensive Value Add numbers before, so the new adjustment is the next value of 0.5% to 2.0% (explanation below on why that seems to calculate correctly) for players who have to bring the ball up the court and/or guard the perimeter. In most cases Assist Ratios and Minutes Played made it pretty clear who the team relied on to get the ball up the court. However, occasionally we would have to look up a big player with a high assist rate, to make a judgment call that Draymond Green really does bring the ball up a good bit for Michigan State so gets some credit, but Henry Sims is leading Georgetown in assists by whipping the ball around from the high post, so the Hoyas guard credit goes to others.

While the initial Value Add equations are precise measurement of the points a player adds to his team’s score and takes off the opponent’s score, I believe the updated program with the Point Guard/Perimeter defense adjustment passes the eye test and yields a much more accurate result than the first show I took at accounting for guard play a couple of weeks ago. While it doesn’t figure into the calculation, I noted in the far right column if a player was projected to go in the NBA draft either in the 1st round, 2nd round, or in 2013. As you can see, 18 of the top 20 Value Add players are projected NBA draft picks.

I also looked at Ken Pomeroy’s top 10 players of the year, and nine of them are in my top 11, with only Mike Scott a little lower at #19.

T.J. McConnell of Duquesne was the unknown for me until he came in at #12, but in going to look him up he may be my next year’s Jae Crowder, a player who does so many things well that he is under appreciated. He is in the top 100 in shooting, hitting over 50% of his shots including 45% of three pointers, is in the top 100 in assist rate, and is 12th in steals despite just missing the top 100 in NOT fouling. Definitely a player to watch despite playing for a mediocre team.

MATH BEHIND THE POINT GUARD/PERIMETER DEFENSE ADJUSTMENT

As far as the math behind the new calculation, good shooting guards tend to average 0.5% less value add than good front line players, due to often having to guard the perimeter rather than grab rebounds. Point guards average a bit lower than even shooting guards. Because Value Add is a very precise measurement (you can add the Value Adds on all players on a BCS team and get a very clear picture of where they rank in their conference), our stat team concluded that we could not adjust in a way that gives some teams more additional credit than others without distorting the overall team’s Value Add.

Therefore, all 345 teams were given an extra 2.5% to be split, with 0.5% being given to the better guards/perimeter defenders, and 1.0% to point guards. However, if a point guard averaged better than a 2-to-1 assist to turnover ratio then he received an extra 1.5%, and for better than 3-to-1 an extra 2.0%. Whatever the combination and number of players credited, the sum of the entire team’s players always adds up to exactly 2.5% since what we are doing is redistributed a little bit of the credit the front line is getting for rebounding or scoring in close to the guards who often break pressure or guard the perimeter without benefit of statistical credit. (You don’t always see all 2.5% for a given team, because sometimes a player receiving credit still does not make the sheet because he did not have a net positive.)

The players who do not look as good in this system as their true talent level are players from UNC and Kentucky, since with 8 and 6 projected NBA draft picks there simply aren’t enough possessions to go around. I still believe this is an accurate reflection of value, because UNC and Kentucky can replace any one star with another NBA-level player, while a star is more valuable to any of the other 343 teams since they will replace him with a much worse player.

So if you want to scout potential NCAA opponents, you could look at Alan’s s-curve. He has MU playing Davidson in the opening round, so if you go to the Air Force-Furman list and see that Davidson has no projected NBA picks but three top 300 players in point guard Nik Cochran and two big guys in De'Mon Brooks (6'7") and Jake Cohen (6'10"). A win there could put MU up against Florida, who on the same sheet we see has three projected NBA picks and three top 100 players.

Jerry Palm and Joe Lunardi today give us Belmont, who I have been saying could be the surprise Sweet 16 team since I did Value Add this past summary. That being said, I do believe we would match up pretty well since they rely on two great, but small, guards in 41st ranked Kerron Johnson (6'1") and 141st ranked Drew Hanlen (5'11"), and I believe DJO and Vander Blue would be up for that defensive challenge. Scott Saunders (6'10") and Ian Clark (6'3") make Belmont one of nine mid-Majors with four Top 400 players.

Hope this gives you a head start on your brackets!

Value Add: Original Defensive Piece

Top 25 returning defensive players - and how to calculate them

Since I posted the offensive ratings of the 2500+ returning players based on the % of points they added to their teams, I’ve been asked why I didn’t also calculate the defensive % of points good players take away from their opponents.

The fact that so few defensive stats are kept is cumbersome for someone who lives number crunching professionally.

Ironically, this is the topic of a talk I will be giving to a couple of colleges, “Winning With Political Microtargeting and Sports Sabermetrics.” The fact is that everywhere in my professional work from City Council Campaigns to a couple of successful Presidential campaigns, I have used Microtargeting to calculate every message and delivery method (door-to-door, mail, TV), and how many votes a dollar spent on each will get FOR my candidate (offense) or dissuade people from voting for our opponent (defense). Likewise in basketball, we keep offensive stats every time the team scores or fails to score, BUT ironically we only keep stats of 22% of the defensive contributions. To illustrate, below are the different ways a basketball possession can end, and how likely each is to happen per 100 possessions. (Parenthesis mean there is a stat that should be kept, but is not):

Possession ends with#PtsOff credit/blameDef credit/blame
3pts made9273PM (split credit if AST)(PtsAllowed)
2pt made2652FGM (split credit if AST)(PtsAllowed)
FT made1319FTM(PtsAllowed)
Steal90TOSTL
Turnover other than steal110TO(TOF)
blocked shot, def reb30FGA2/3rd BLK, 1/3 REB
other missed 2, def reb140FGA2/3rd (STOP), 1/3 REB
missed 3 pt, def reb110FGA2/3rd (STOP), 1/3 REB
missed FT, def reb40FTA2/3rd (STOP), 1/3 REB
100 possessions ended (2pts during possession))100100credit/blame for 100%credit/blame for 22%


(See note at bottom on why shooting percentages look higher than they are at first glance)

Please don't use this as an excuse to comment on politics, but comparing what is tracked is enlightening for me.

When Moneyball came out in 2003, the top Bush campaign guys went crazy over it. They loved the Sabermetrics behind it, and pretty soon they had built the most incredible Microtargeting model ever, calculating that the biggest bang for their campaign dollar would be getting specific people in specific Ohio localities who were predisposed to Bush to vote after sitting out the 2000 campaign. The fact that we don’t keep track of a player’s points allowed (PtsAllowed), forced turnovers other than steals (TOF) or defensive stops (STOP) would be like the Bush campaign saying, “I really don't want to calculate how many people will vote AGAINST Kerry, only how many will vote FOR Bush," or the Obama team deciding during their incredible Microtargeting of Virginia four years later that they really didn’t want to know why young Virginians might vote AGAINST McCain.

We simply have to do our best to calculate defense to have a whole picture of a player's overall statistical value to a team.

Here is what we can calculate from the 22% of individual defensive stats we do keep track of:

An average team blocks 9.2% of opponent’s 2-pointers takes away 2% of their points – double their blocks and they take away 4%. An average team steals the ball 9% of the time which takes away 9% of the opposition points. Double it and they take away 18% of the points. And finally a team that grabs the average 67.3% of opponents’ missed shots takes away 11% of their opponents points, so an individual player who gets one-fifth of that (13.5%) takes away just over 2% while he is on the floor. Here is the breakdown:




Adjust to ave. 1 pt. per tripPts+-kenpomConvert from Pomeroy
BLK-29.2%0.217
STL-99.4%0.957
REB (Defensive only)-1167.3%0.163
No individual stat for 30 STOPS or TOF-30  
48 trips scored 100 points52 


To understand the Pts+-, consider that the AVERAGE possession results in just over 1.0 points. In the 100 possession in the first table, the team allowed points on 48 of them, but it was a total of 100 points scored (9 treys, 26 other field goals, and 19 free throws on the 13 times they finished with a made free throw). Since you expect 1 point allowed per trip, if you instead give up a 3-point shot then you are penalized for a+2 (3 is 2 points more than average), while a regular field goal is +1, etc., but a defensive stop is -1 since they didn’t get their one-point average that trip.

I’ve put in a converter so that you can multiply the Blk% in Pomeroy by .217 to determine what % of points a player is erasing from the opponents score with his blocked shots, steals and defensive rebounds.

However, that still leaves us way short of a Defensive Efficiency Rating (DRtg) that would match the precise Offensive Efficiency Rating (ORat) developed by Dean Oliver in 2003 and incorporated into www.kenpom.com that same year.

Introducing TeamDef72%

While we lack the INDIVIDUAL defensive stats we need, we do have them at the team level. Basically starting with Pomeroy’s Team Defensive Rating, which is the number of Points Allowed per possession, and taking out the influence of the teams above or below average defensive rebounding, steals, and blocks, and we are left with one number that sums up a team’s Stops, Points Allowed and Turnovers Force. Here are the best and worst:

1. Florida State 94.4
2. Texas 94.9
3. San Diego State 95.2
4. North Carolina 95.9
5. Louisville 96.1
6. Ohio State 96.1
7. Duke 96.2
8. Utah State 96.4
9. Kansas 96.6
10. Purdue 96.9
61. Marquette 103.3 (42nd of 73 BCS teams)
343. Longwood 126.0
344. MD Baltimore County 126.1
345. Chicago State 127.3

What this figure does is to give every player on a team one-fifth of the Stops, Points Allowed and Turnovers Forced while he is on the court, which gives him this figure. We see that when we add that all together, even after adjusting for competition as built into Pomeroy’s defensive rating, MU was below average for a BCS school in all the things you do on defense that aren’t recorded.

With 72% of their rating taken care of, we can now add back in each player’s blocks, steals and defensive rebounds.

Once I ran the numbers (and I really have no idea where everyone will finish until I sort by Value), Jared Sullinger calculated as the best returning defensive player in the country. To walk through his numbers:

1. Ohio State’s “TeamDef72%” stat for all the things we don’t measure was a 96.1, meaning they were the 6th best team at all the defensive things you do for which we do not have a stat. We start by giving Sullinger that number as one of the five Buckeyes on the court making that happen.
2. Sullinger just missed being the top defensive rebounder in any BCS conference, and his 26.3% of rebounds grabbed after a miss can be multiplied by the factor on the table (.163)to convert this Pomeroy figure to save us some calculations in knowing his rebounding lower's opponents' scoring by 4.3%, so his defensive rating is lowered to 91.8.
3. He is not a great shot-blocker, only rejecting 2% of opponents two pointers, which when multiplied by the .217 on the converter tells us his shot blocking lowers opponents' scoring by 0.4% for a total for 91.4.
4. Finally, he does get some steals, enough to erase another 1.9% of opponents scoring. For steals, we use the Stl% that actually appears on Pomeroy's page.
5. This gives Sullinger an 89.5 Defensive Efficiency Rating, the flip side of the ORtg developed by Oliver in 2003. Not as precise as the ORtg because we are dividing up a lot of the team activity equally, but the same basis.
6. You may notice that Bernard James of Florida State was actually a more effective defender when on the floor, with a DRtg of 86.3, meaning teams are only likely to score 86.3 points per trip per possession with him on the floor. Just ask Notre Dame about James after he played 21 minutes despite almost passing out sick, blocking 3 shots and grabbing 9 rebounds in that game to help to hold ND to 31% shooting from the floor. However, like with Kemba Walker in the blog on offense, it was more valuable for Sullinger to be on the floor for 78.9% of the minutes than for James to play 52.7% of the minutes.
7. Once we multiply by minutes played, we see that over the course of the season Sullinger reduced the number of points scored by Ohio State opponents by -7.3%, the best total in the country, while James was just a little behind at -6.6% in 3rd place.

As will come as no surprise to all of us who watched last year, Marquette did not have the same kind of production on defense as they did on offense. Jae Crowder calculates as easily the best returning offensive and defensive player for Marquette, but he is the 24th best in the country on offense while the 160th best defensive player of the 2500+ returning.

The whole list to come, but here is how the Top 25 nationally calculate, followed by Marquette's returning players.

Value Add Original Offensive Calculations Post

How to determine the precise offensive value of each player to your team

This post outlines how to determine the exact offensive value of all 4,485 Division 1 players in the country and represent it with one concise “Value Add.” I believe it is the small final piece of the huge jigsaw puzzle built by four people much, much, much smarter than myself. I’ll start with the 10 most valuable offensive players in the country this year, and then go through how we know that is the case:



RankNameTeamORtgReplIn play%Poss%MinValue add
1Jordan TaylorWisconsin126.991.91.3827.490.69.46%
2Kemba WalkerConnecticut116.790.81.2931.492.48.29%
3Jimmer FredetteBrigham Young114.592.71.2336.488.57.56%
4Charles JenkinsHofstra123.596.11.2828.492.37.46%
5Noah DahlmanWofford130.297.31.3427.276.67.05%
6Reggie JacksonBoston College119.992.31.3027.185.46.93%
7Derrick WilliamsArizona122.792.71.3228.774.16.90%
8Ben HansbroughNotre Dame120.591.81.3125.087.16.81%
9Mickey McConnellSt. Mary's129.195.91.3521.392.16.80%
10Jared SullingerOhio St.120.492.51.3027.078.96.44%


This post is probably too detailed to be of interest to 95% of our readers, but I need it as a reference to refer back to in future posts to make it clear that is a precise scientific measurement of the value add of each player. I can’t decide to weight offensive rebounds a little more because I don’t like that a Wisconsin player was actually slightly more valuable than the next best three players, Kemba Walker, Kyrie Irving (7.80% in the 11 games he played, but drops due to injury) and Jimmer Fredette. The fact is Jordan Taylor’s presence this year meant Wisconsin scored 9.46% more points than they would have if he had not played.

History
In 1977, Bill James’ Baseball Abstract finally started to utilize stats to determine how good players are rather than just look at the top batting averages each Sunday. Then in 2003, Moneyball revealed how a GM (Oakland's Billy Beane) could beat opposing teams with four times the salary money and scouts if he simply knew how to weight players’ stats to make decisions. That same year, Dean Oliver’s Basketball on Paper revealed with astounding accuracy how stats could determine how many points basketball players and teams could produce, and Ken Pomeroy created www.kenpom.com to organize every shred of information on each player, with annual summaries now in the Basketball Prospectus.

For a basic understanding of how much value a player adds, you can read my easlier post:

Best Offensive Players in Big East

However, for simplicity sake I left out the more complicated parts of the calculation out of that post. The following are the actual steps you have to go through to know precisely how valuable a player is to your offense.

How many points does a player generate per offensive trip (ORtg)
Luckily Oliver and Pomeroy have done the hard work for us by weighing how often Kemba Walker helps the team by making a shot, drawing a foul, setting up a teammate with an assist, grabbing an offensive rebound, or how often he hurts the team by turning the ball over or missing a shot. Read Basketball On Paper for the in depth explanation of how we know UConn scores 116.7 points for every 100 possessions when Kemba is involved. That’s a very good total, but neither Kemba or Jimmer Fredette are in the Top 100 in the country based purely on ORtg because they play tired and have to create shots against double teams to hold down their ratio against some players who only shoot when open and get plenty of breathers.

Replacement vs. Defense
The fact is we don’t know how valuable Kemba was until we know how many points replacements would have scored if Kemba didn’t play this year. The “average” player going up against the “average” defense this year produced 101.3 points per 100 trips down the court. However, Kemba had to overcome much tougher defenses, as the average defense UConn faced allowed only 96.2 points per 100 trips, meaning UConn faced the 4th toughest defenses of the 345 Division 1 teams (only Michigan State, Georgetown and Penn State faced tougher defense).

Replacement Player
However, an “average” player does not replace a player when he doesn’t play. Teams typically have an 8-man rotation, so if a player didn’t play it would be the 9th or 10th best player on a roster that typically takes the player’s place, in combination with other players being less effective due to having to do more (more tired, more defensive attention). James first noted that “average” players have value because they are better than “replacement” players, and that many a championship has been lost for lack of an average player at a key position (e.g. how much better would Marquette have been with an average Big East Center to play with the Three Amigos).

Therefore to determine how many points a typical replacement player would have produced, the first step is to identify the average replacement player. To this end, I broke down every player in the country based purely on minutes and their level of conference. There are 345 teams, so I treated the 1725 players who averaged the most minutes as Starters (345 X 5), then the 1035 who played the next most as in the Rotation (typically 6th, 7th and 8th men), and finally any other player who played at least four minutes a game as a replacement (typically the 9th or 10th man who gets in the rotation when someone is out).



LevelTeamsStartersRotationReplacementSweats
BCS73365219128237
Mid-Major96480288178302
Other D1176880528414466
Total345172510357201005


It was the 128 “Replacement” players for BCS teams that give us the typical player who would replace a player who did not play. After quite a bit of math, I determined that multiplying the average defensive rating a player faced by 0.9435, we get the exact figure of how the 128 BCS Replacement players would have done in our player’s shoes. Against the tough defense played against Kemba, a typical BCS Replacement player would have produced on 90.8 points per 100 trips down the court.

While these 128 BCS Replacement players are 9th or 10th men in the BCS, they are actually producing almost as many points as your average starter in a lower D1 conference (1.03) or as a player in the rotation for a Mid-Major school (1.02). An actual Replacement player for a lower D1 school only produces 89% of the points of a BCS Replacement player, but to keep everyone on the same basis, we will compare every player to what a BCS Replacement player would have done in his shoes:



LevelBCSMid-MajorOther D1
Starters1.151.091.03
Rotation1.071.020.94
Replacement1.000.940.89


Percent better than a Replacement Player (% > Replace)
So when Kemba was involved in the play, UConn scored 29% more points than they would have if a Replacement Player had been involved instead, so we note that as 1.29 (116.7 / 90.8 = 1.29).

Take charge players (%Pos)
While the 1.29 figure is a precise measurement of the % of points Kemba adds when he is involved in the play, the next step is to measure how often he takes control vs. one of the four other players on the court having to make a play.

Note that Mickey McConnell of St. Mary’s was actually slightly more likely to produce points than even Kemba when he is involved in the play. However, McConnell is not the play maker that Kemba is. McConnell is only involved in the play 21.3% of St. Mary’s trips down the court – meaning teammates that aren’t as good have to make the play 78.7% of the time. Kemba can produce a play or set up a teammate many trips, meaning the play goes through him 31.4% of the time he is on the court and lesser teammates only have to make the play 68.6% of the time. The formula that shows the actual impact a player has while he is on the court then is:

(WheninPlay * %Pos) + (100 - %Pos) = OnCourtWith5AveragePlayers

Finally, %Min played
Kemba’s Offensive Rating would have been much higher if he could have rested more than an average of 3 minutes a game, but a very tired Kemba was still much more valuable than a fresh Donnell Beverly coming into take his place. So the additional minutes a player can stay on the floor and still add any value helps the team even though it lowers his ORtg.

Note that Noah Dahlman was as dominant a force as Kemba when he was on the court. He called for the ball and scrapped inside, hitting his typical 60% from the floor en route to a 21 point, 9 rebound game that almost knocked BYU out in the opening round.

However, Dahlman wasn’t quite as dominant as Kemba because as a 6-foot-6 inch center, he needed about 10 minutes of rest a game to maintain his very high level of play. Those extra 7 minutes a game Kemba gave UConn made him more valuable than Dahlman overall, and the formula that measures the final factor - % of minutes played, is as follows:

(((OnCourtWith5AveragePlayers – 100)*(%Min * 0.01))/100) = Value Add

The percentage that comes from this formula is the % of points a player has added to his team’s results over the course of the season. The formula does not distinguish between a player being injured vs. not player well enough to be on the court. For example, Kyrie Irving played 68.95% of Duke’s minutes in the 11 games he was healthy, and had a Value Add of 7.8% to Duke during those 11 games. That makes him one of only six players in the country with a Value Add of 7% or higher. However, because of the 26 games he was out with an injury, Irving only played 20.5% of Duke’s overall minutes, and therefore his actual value add for the season was 2.32%, good for 365th best of the 4,485 players.

Bell Curve of Talent
Looking at the breakdown of Value Adds, we see that we are looking at the end of the bell curve we see in almost all sports. There are always very few players who are much, much more valuable than average, then a little group that are much better, etc., until we get down to most players who are about average. As you can see, most Division 1 players (the 1489 plus the 997 in the last two categories) have a 0% because they do not add offensive points. Here is the number of players that fall into each Value Add range:




Players% in rangeValue Add Ranges and Notes
50.11%7% or higher Value Add: 1 in 1000 add more than 7% to their team's total scoring
300.67%5 to 7% Value Add: Fewer than 1% of all players can add >5% to team's scoring
1733.86%3 to 5% Value Add: Still in the top 5% of all players if you add 3%
75316.79%1 to 3% Value Add: Just over 1 in 5 players are in one of top categories
103823.14%Some Value Add, but less than 1%: Another 23% add some scoring
148933.20%0% Value Add: 33% do not help offense, but still play due to defensive ability,etc.
99722.23%0% Value Add, less than 4 min a game: 22% play fewer than 4 minutes a game


Precise vs. Potential/Good
Of course, we can only have precise measurements of how good a player was in the past, not his potential for future greatness. The two often correspond fairly closely since having great talent makes it more likely you produce great results. For example 20 of the top 30 players in Value Add this year are projected to go in the NBA Draft while obviously a great college player like Dahlman isn’t likely to play center in the NBA at 6-foot-6. I have to admit I didn’t know who Charles Jenkins of Hofstra was until the formula showed he was the 4th most valuable offensive player in the country this year, and then I looked and saw who is projected to be the 28th college player drafted.

There are others like Yancy Gates who is expected to be picked next year, but did not realize his potential last year as the 315th most valuable player in the country.

Defensive Value
Finally, keep in mind that this is a precise measurement of OFFENSIVE value add. The fact that Derrick Williams and Jared Sullinger are also dominant defensive players inside means they might have been the two most valuable overall players in the country (and they are projected to be the 2nd pick in the NBA draft this year and next respectively). We could add to the formula that Sullinger dominates the defensive glass by grabbing an incredible 26.2% of all opponents’ misses (Williams also great at 21.7%) and other defensive stats, but that would take us away from our precise measurement and start making the process subjective.

“You can’t measure everything!”
Of course, the counter argument is that there are still things that are not measured – the nice pick that helped a team, or throwing a pass to a guy 30 feet from the basket with one second on the shot clock so he gets blamed for the missed shot. This is true, however in the scope of things, these small factors are like saying, “you didn’t credit the player for getting hit by a pitch four times this year,” when in fact that was probably balanced out by a few times of hitting into a double play in the clutch.

The fact is that what Oliver laid out in 2003 has proven to be amazingly accurate every year, just like the Oakland As success on a tiny budget defied logic until author Michael Lewis wrote Moneyball and let Beane's secrets out so that the Yankees and Red Sox could start using the same statistics.

Explanation and History of Value Add

For an explanation of the Value Add basketball rankings, you can always click on this post from the bottom of the www.valueaddbasketball.com database.  I must thank Sports IllustratedESPNNBC Sports, Fox Sports and all of the other outlets who have covered Value Add.  A variation of this formula measures how likely a player is to be successful in the NBA, but the basic explanation of how and why it works follows.

Anthony Davis added 7.29% to Kentucky’s scoring with his offense and took away -5.06% from opponent’s scoring, so his total impact on the score was 12.35%, the highest Value Add in the country in 2012. This means that if Kentucky would have lost a game 69-70 with a typical fourth or fifth man off the bench playing instead of Davis, then with Davis they win 74-66.
When you first pull up the database at www.valueaddbasketball.com, you will see the ratings of all of the basketball players for this season.  You can search by team, conference or a player’s name to see where specific players rank, or just scroll through more than 3000 players for each season.
You can also click on the other tabs to see players from the 2006 season through the projections for how good players should be in 2014, including new recruits.  This is probably enough explanation for most readers, but if you want more details, the ratings consist of three components:
1. The Offensive component is explained in this post, as we can measure with great precision how many points a player ad to his team’s score.
2. The Defensive component is not quite as precise but measures a player’s ability to block shots, steal the ball, grab defensive rebounds and be part of a defense that prevents opponents from scoring in other ways.  In both cases, a players rating is measured against exactly how good each of an opponents’ offense and defense is, so the same player should have basically the same rating whether he plays for the national champion or the worst Division 1 team.
3. The Point Guard/Perimeter Defensive Rating (PG/Per) redistributes a small percent of the credit from post players who do not turn the ball over as much because they do not have to dribble as often and who grab more defensive rebounds because they do not have to play defense on the perimeter.  After extensive study, this figure was determined the most accurate way to fairly adjust ratings based on position, as explained in this post.
NBA Indicators Developed by Rob Lowe
Much like Ken Pomeroy at www.kenpom.com, Rob Lowe (who is much smarter than me) developed methods for measuring tempo-free statistics as calculated in Dean Oliver’s book, “Basketball on Paper.”  While Oliver, and in turn Pomeroy, determined how to measure a dozen things a player does to help his team win, www.valueaddbasketball.cominstead breaks down all ways a player helps his team into on figure that determines his overall value.
Lowe built on Value Add and other sources to develop calculations that measure how likely a player is to be successful in the NBA, and officials from several NBA teams met with me prior to the 2012 draft to review these valuations.  These evaluations are only available to select NBA teams.
Others have built on the Value Add system since it was developed.  Sports Illustrated used it to go back further in history and conclude that Duke’s JJ Reddick was the best Value Add guard since modern stats were introduced and that research was picked up by ESPN. Basketball Prospectus then started to add to the equation with some additional calculations.
Big Apple Buckets out of New York was instrumental in the further development of a variation to calculate the value of players in the Ivy League and other low major conferences, and we have received follow-up questions from Ivy League schools and officials.  These ratings can be found by clicking on the “Low-Major” link in the bottom right corner.
In addition, these rankings are written about regularly on sites for teams like KentuckyArizonaNC StateBaylor and League sites for the Big Ten, Summit, Horizon and Patriot Leagues as well as the original www.crackedsidewalks.com page that focuses on Marquette and the Big East/Catholic 7.
One credibility test for the system occurred after an engineer developed the program to run the calculations, and a Marquette player, Jae Crowder, showed up as an All-American.  Statisticians never like to have a favorite when running numbers and the fact that I covered Marquette and Crowder’s Value Add was so high even though he was not even an All-Conference player nor listed in the top 100 NBA prospects made me nervous.  However, several NBA team officials met with me and were very complementary that my calculations were “finding” several of the same sleeper prospects they had found through intense scouting, including Crowder.
I did later meet Crowder after his last game at Marquette and was put at ease months later when the AP sports writers named him one of the top 10 players as a second-team All-American and then the Dallas Mavericks drafted him and he was named an All-Star in Summer League and made the Mavs.  While the numbers were backed up, I will give Crowder and the other players who backed up their Value Add ratings with great runs on the court assists for the success of Value Add.

I have also taken questions on Value Add on the radio, and spoken on the statistics behind politics and sports.

Sunday, January 28, 2018

Value Add 5.0 Adjustment for Top 31 Teams of this Century

00The key advance in the 5.0 Version of Value Add Basketball to adjust for the fact that a player on a great team gets fewer touches. I long noted that some of this is offset because while a player who goes to Grambling will shoot a lot more than if he went to Kentucky, he also would face much tougher defenses, many more double teams, and he would need to handle the ball much much more than he would lose efficiency due to wearing down (even Kobe Bryant was a below average player when asked to handle the ball close to 40 percent of the time.

Separate adjustments are made for the level of defense played, but the adjustment calculated that normal adjustment from the Value Add Version 3.0 POINTS PER HUNDRED POSSESSIONS stats that coincided with www.kenpom.com was .670 for the average 67 trips per game this century. Instead, the best teams in the country would multiply by .850 and the worst team at just over .500.

This worked well for ALMOST all of the 6,000 teams to take the court since the 2001-02 season measured in the historic database, but the old bell curve indicated the occasional exceptional team deserved even beyond the 0.850.

To make this adjustment, when a team's AdjEM rating at www.kenpom.com is 30+, then the following formula is run to determine the adjustment made to players on the team between their Value Add 3.0 rating (impact on team AdjEM) and Value Add 5.0 rating (points per game above replacement):

((AdjEM + 100) x 0.0066) = Adjustment

As of January 28, 2018, these are the 31 super teams of the century requiring this adjustment.

21st CentSuper Teams (30+AdjEM)yearConfW-LAdjEMValue Add Adjust
1Kentucky2015SEC38-136.910.917
2Kansas2008B1237-335.210.906
3Duke2002ACC31-434.190.899
4Wisconsin2015B1036-433.720.896
5Ohio St.2011B1034-333.470.894
6Duke2010ACC35-533.290.893
7Louisville2013BE35-532.920.891
8North Carolina2005ACC33-432.770.890
9Villanova2018BE20-132.690.889
10Illinois2005B1037-232.680.889
11Kentucky2012SEC38-232.590.888
12Duke2015ACC35-432.480.888
13Arizona2015P1234-432.360.887
14Duke2004ACC31-632.330.887
15Virginia2018ACC20-132.270.886
16Gonzaga2017WCC37-232.050.885
17Villanova2016BE35-532.010.884
18Kansas2010B1233-331.850.883
19Memphis2008CUSA38-231.510.881
20Purdue2018B1021-231.410.880
21North Carolina2007ACC31-731.370.880
22Florida2013SEC29-831.180.879
23North Carolina2009ACC34-431.140.879
24Florida2007SEC35-530.810.876
25Villanova2015BE33-330.650.875
26Louisville2014Amer31-630.410.874
27North Carolina2008ACC36-330.220.872
28Cincinnati2002CUSA31-430.190.872
29Arizona2014P1233-530.110.872
30Ohio St.2012B1031-830.070.871
31Virginia2015ACC30-430.060.871

Monday, January 15, 2018

Oklahoma Surges to #4 Behind Trae Young; Ohio State Star Deserves 1st Team All-American Status So Far


On this MLK Day, Villanova receives 63 of 65 votes in the new AP Top 25 on the strength of the two best players in the Big East listed in the post above. However, Oklahoma surges to No. 4 behind two more dominant performances against ranked teams (TCU and Texas Tech) by Trae Young. While the No. 1 spot in the AP poll has rotated between teams, Value Add Basketball calculates the Player of the Year race so far is a blowout - with Young ahead of where Seth Curry was at Davidson and rivaling Anthony Davis as the best college player of the century in all-time ratings that go back to the year before Carmelo Anthony and Dwyane Wade took Syracuse and Marquette to the Final Four.

Here are the Top 25 players, ranked by how many points per game they improve their team over what a replacement player would do. The fastest riser in this ranking is Keita Bates-Diop, who has been the most valuable player on the court in 11 straight games according to www.kenpom.com and as of today should give the Ohio State Buckeyes a 1st Team All-American. You can click here to sort all 4,090 players this season by conference or the two teams you are watching in a game, or simply look up names to see where your favorite players rank. Email or call 404.606.3163 with any questions.

Rank Player Ht Team VA5D Class
1Trae Young6'2"Oklahoma12.93Fr
2Deandre Ayton7'1"Arizona10.98Fr
3Jevon Carter6'2"West Virginia10.84Sr
4Keita Bates-Diop6'7"Ohio St.10.78Jr
5Jock Landale6'11"Saint Mary's10.55Sr
6Jalen Brunson6'2"Villanova10.43Jr
7Yante Maten6'8"Georgia10.27Sr
8Marvin Bagley6'11"Duke10.12Fr
9Gary Clark6'8"Cincinnati9.9Sr
10Luke Maye6'8"North Carolina9.79Jr
11Devonte' Graham6'2"Kansas9.57Sr
12Dean Wade6'10"Kansas St.9.52Jr
13Mikal Bridges6'7"Villanova9.51Jr
14Keenan Evans6'3"Texas Tech9.21Sr
15Dewan Huell6'11"Miami FL9.19So
16Tookie Brown5'11"Georgia Southern9.07Jr
17Devon Hall6'5"Virginia9.05Sr
18Donte Grantham6'8"Clemson9.02Sr
19Garrison Mathews6'5"Lipscomb8.96Jr
20CJ Massinburg6'3"Buffalo8.65Jr
21Bonzie Colson6'6"Notre Dame8.48Sr
22Kenrich Williams6'7"TCU8.43Sr
23Nick King6'7"Middle Tennessee8.41Sr
24Dakota Mathias6'4"Purdue8.39Sr
25Juwan Morgan6'7"Indiana8.29Jr