High Floor and Ceiling Part 3
Now that we have a better understanding of what standard deviations, and Z-scores are, and how they can be applied to our stats here, we can start to work through how to functionally apply this and identify different types of players for different player pools.
One big issue you may have considered already is if a player plays for more than one team over the course of the season. It's not exactly uncommon for a player to have different levels of productivity and consistency when moving from team to team as these things can depend on teammates, how often they are utilized in the offensive scheme, etc. We will take a quick look at an example here, and then work towards breaking out our calculations by a combination of player and team in order to have more accurate numbers to work with.
As a quick example, we will take a look at Marcus Morris who played for both the Knicks and Clippers over the course of this season (when the stats were pulled), but only played a few games for the clippers so his standard deviation will likely be skewed.
import pandas as pddf = pd.read_excel(r"C:\Users\nfwya\OneDrive\Code Backup\nba\boxScores\20200215_Pandas.xlsx")dfM = df[df['Player'] == 'Marcus Morris Sr.']dfM.head(17)| Player | Match_Up | Game_Date | MIN | PTS | REB | AST | STL | BLK | TOV | FP | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 49 | Marcus Morris Sr. | NYK vs. POR | 01/01/2020 | 33 | 18 | 7 | 3 | 0 | 0 | 0 | 30.9 |
| 297 | Marcus Morris Sr. | NYK @ PHX | 01/03/2020 | 35 | 25 | 3 | 2 | 2 | 1 | 0 | 40.6 |
| 574 | Marcus Morris Sr. | NYK @ LAC | 01/05/2020 | 35 | 38 | 5 | 1 | 2 | 1 | 3 | 51.5 |
| 1949 | Marcus Morris Sr. | NYK vs. PHX | 01/16/2020 | 33 | 17 | 3 | 1 | 0 | 0 | 1 | 21.1 |
| 2217 | Marcus Morris Sr. | NYK vs. PHI | 01/18/2020 | 36 | 20 | 6 | 3 | 0 | 0 | 1 | 30.7 |
| 2476 | Marcus Morris Sr. | NYK @ CLE | 01/20/2020 | 30 | 19 | 3 | 0 | 1 | 0 | 0 | 25.6 |
| 2717 | Marcus Morris Sr. | NYK vs. LAL | 01/22/2020 | 37 | 20 | 6 | 1 | 0 | 0 | 4 | 24.7 |
| 2955 | Marcus Morris Sr. | NYK vs. TOR | 01/24/2020 | 37 | 21 | 10 | 2 | 0 | 0 | 1 | 35.0 |
| 3179 | Marcus Morris Sr. | NYK vs. BKN | 01/26/2020 | 34 | 21 | 4 | 0 | 1 | 1 | 1 | 30.8 |
| 3412 | Marcus Morris Sr. | NYK @ CHA | 01/28/2020 | 35 | 23 | 4 | 1 | 1 | 0 | 4 | 28.3 |
| 3527 | Marcus Morris Sr. | NYK vs. MEM | 01/29/2020 | 26 | 17 | 6 | 0 | 1 | 1 | 4 | 26.2 |
| 3894 | Marcus Morris Sr. | NYK @ IND | 02/01/2020 | 35 | 28 | 6 | 0 | 1 | 0 | 2 | 36.2 |
| 4125 | Marcus Morris Sr. | NYK @ CLE | 02/03/2020 | 36 | 26 | 2 | 1 | 0 | 0 | 1 | 28.9 |
| 4810 | Marcus Morris Sr. | LAC @ CLE | 02/09/2020 | 22 | 10 | 4 | 2 | 3 | 0 | 1 | 25.8 |
| 5042 | Marcus Morris Sr. | LAC @ PHI | 02/11/2020 | 35 | 13 | 5 | 1 | 0 | 1 | 1 | 22.5 |
| 5274 | Marcus Morris Sr. | LAC @ BOS | 02/13/2020 | 42 | 10 | 8 | 2 | 0 | 2 | 2 | 26.6 |
| 5453 | Marcus Morris Sr. | NYK @ SAS | 10/23/2019 | 39 | 26 | 4 | 1 | 3 | 0 | 1 | 40.3 |
As we can see here, Up until February 3, 2020 Marcus was playing with the Knicks, but starting on February 9th, he is a member of the Clippers. This data was pulled in mid February of that year so he had only played 3 games with the clippers at that time, and it seems his fantasy output was pretty consistent over those 3 games with them, so his standard deviation will likely be quite different than his with the Knicks.
Now, we need to isolate the player's team and create a new column to hold that data for each player for every game they play. In order to do that, we will utilize a lambda function as we have done in the past to pull the team abbreviation out of the match up field. Luckily for us, there is very little logic we need to build into this function, as the player's team is always listed first in the match up field regardless of whether they were home or away that day.
So let's go ahead and define that lambda function, apply it to a new column titled 'Team', then we will go ahead and calculate the mean and standard deviation using both the Player and Team fields as our index values.
teamFxn = lambda x: x['Match_Up'].split(" ")[0]
df['Team'] = df.apply(teamFxn, axis=1)
dfStat = df.groupby(['Player', 'Team']).FP.agg({'mean', 'std'})
df = df.merge(dfStat, left_on=['Player', 'Team'], right_index=True)
df = df.sort_values('Game_Date')df.head()| Player | Match_Up | Game_Date | MIN | PTS | REB | AST | STL | BLK | TOV | FP | Team | mean | std | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | Anfernee Simons | POR @ NYK | 01/01/2020 | 26 | 3 | 3 | 3 | 1 | 0 | 1 | 13.1 | POR | 14.944444 | 7.667123 |
| 10 | CJ McCollum | POR @ NYK | 01/01/2020 | 32 | 17 | 4 | 4 | 0 | 0 | 0 | 27.8 | POR | 33.944231 | 10.935020 |
| 46 | Kyle Korver | MIL vs. MIN | 01/01/2020 | 18 | 8 | 4 | 2 | 0 | 1 | 3 | 15.8 | MIL | 12.036957 | 6.266112 |
| 47 | Kyle Kuzma | LAL vs. PHX | 01/01/2020 | 27 | 19 | 4 | 1 | 1 | 0 | 0 | 28.3 | LAL | 20.513636 | 9.673642 |
| 48 | LeBron James | LAL vs. PHX | 01/01/2020 | 38 | 31 | 13 | 12 | 2 | 1 | 5 | 68.6 | LAL | 51.711765 | 9.848099 |
Now we can go ahead and check back on our main man Marcus Morris and see if we calculated everything correctly, and if our hunch on his standard deviation being much smaller with the Clippers is true.
dfM = df[df['Player'] == 'Marcus Morris Sr.']
dfM.head(17)| Player | Match_Up | Game_Date | MIN | PTS | REB | AST | STL | BLK | TOV | FP | Team | mean | std | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 49 | Marcus Morris Sr. | NYK vs. POR | 01/01/2020 | 33 | 18 | 7 | 3 | 0 | 0 | 0 | 30.9 | NYK | 29.909302 | 10.926286 |
| 297 | Marcus Morris Sr. | NYK @ PHX | 01/03/2020 | 35 | 25 | 3 | 2 | 2 | 1 | 0 | 40.6 | NYK | 29.909302 | 10.926286 |
| 574 | Marcus Morris Sr. | NYK @ LAC | 01/05/2020 | 35 | 38 | 5 | 1 | 2 | 1 | 3 | 51.5 | NYK | 29.909302 | 10.926286 |
| 1949 | Marcus Morris Sr. | NYK vs. PHX | 01/16/2020 | 33 | 17 | 3 | 1 | 0 | 0 | 1 | 21.1 | NYK | 29.909302 | 10.926286 |
| 2217 | Marcus Morris Sr. | NYK vs. PHI | 01/18/2020 | 36 | 20 | 6 | 3 | 0 | 0 | 1 | 30.7 | NYK | 29.909302 | 10.926286 |
| 2476 | Marcus Morris Sr. | NYK @ CLE | 01/20/2020 | 30 | 19 | 3 | 0 | 1 | 0 | 0 | 25.6 | NYK | 29.909302 | 10.926286 |
| 2717 | Marcus Morris Sr. | NYK vs. LAL | 01/22/2020 | 37 | 20 | 6 | 1 | 0 | 0 | 4 | 24.7 | NYK | 29.909302 | 10.926286 |
| 2955 | Marcus Morris Sr. | NYK vs. TOR | 01/24/2020 | 37 | 21 | 10 | 2 | 0 | 0 | 1 | 35.0 | NYK | 29.909302 | 10.926286 |
| 3179 | Marcus Morris Sr. | NYK vs. BKN | 01/26/2020 | 34 | 21 | 4 | 0 | 1 | 1 | 1 | 30.8 | NYK | 29.909302 | 10.926286 |
| 3412 | Marcus Morris Sr. | NYK @ CHA | 01/28/2020 | 35 | 23 | 4 | 1 | 1 | 0 | 4 | 28.3 | NYK | 29.909302 | 10.926286 |
| 3527 | Marcus Morris Sr. | NYK vs. MEM | 01/29/2020 | 26 | 17 | 6 | 0 | 1 | 1 | 4 | 26.2 | NYK | 29.909302 | 10.926286 |
| 3894 | Marcus Morris Sr. | NYK @ IND | 02/01/2020 | 35 | 28 | 6 | 0 | 1 | 0 | 2 | 36.2 | NYK | 29.909302 | 10.926286 |
| 4125 | Marcus Morris Sr. | NYK @ CLE | 02/03/2020 | 36 | 26 | 2 | 1 | 0 | 0 | 1 | 28.9 | NYK | 29.909302 | 10.926286 |
| 4810 | Marcus Morris Sr. | LAC @ CLE | 02/09/2020 | 22 | 10 | 4 | 2 | 3 | 0 | 1 | 25.8 | LAC | 24.966667 | 2.173323 |
| 5042 | Marcus Morris Sr. | LAC @ PHI | 02/11/2020 | 35 | 13 | 5 | 1 | 0 | 1 | 1 | 22.5 | LAC | 24.966667 | 2.173323 |
| 5274 | Marcus Morris Sr. | LAC @ BOS | 02/13/2020 | 42 | 10 | 8 | 2 | 0 | 2 | 2 | 26.6 | LAC | 24.966667 | 2.173323 |
| 5453 | Marcus Morris Sr. | NYK @ SAS | 10/23/2019 | 39 | 26 | 4 | 1 | 3 | 0 | 1 | 40.3 | NYK | 29.909302 | 10.926286 |
Sure enough, Marcus' numbers for the knicks and clippers do in fact differ, and his standard deviation is significantly lower with the Clippers as well. This brings up another good aspect to consider, how do we account for skewed deviations when working with smaller sample sizes? And is this a big enough problem for us to even worry about?
To start investigating this issue, we are going to create a new dataframe consisting of just the Player, Team, Mean, and Standard Deviation, and then drop all duplicates and sort by standard deviation. This will allow us to only have to see a single record for each player/team combination, and review the players with the lowest standard deviations. However we are not going to permanently drop the duplicates at this time, just create a temporary dataframe to take a quick look.
dfPlayers = df[['Player', 'Team', 'mean', 'std']]dfPlayers.drop_duplicates().sort_values('std').head(15)| Player | Team | mean | std | |
|---|---|---|---|---|
| 4774 | Glenn Robinson III | PHI | 10.950000 | 0.353553 |
| 4765 | Dewayne Dedmon | ATL | 31.066667 | 1.078579 |
| 4810 | Marcus Morris Sr. | LAC | 24.966667 | 2.173323 |
| 2322 | Allen Crabbe | MIN | 6.671429 | 3.526431 |
| 4902 | James Ennis III | ORL | 9.850000 | 3.747666 |
| 2584 | Anthony Tolliver | SAC | 5.550000 | 4.122499 |
| 2124 | Chris Chiozza | BKN | 6.200000 | 4.949747 |
| 5346 | Chris Chiozza | WAS | 14.375000 | 4.977880 |
| 208 | Theo Pinson | BKN | 11.204348 | 5.060675 |
| 168 | Matthew Dellavedova | CLE | 8.887500 | 5.075744 |
| 447 | Marco Belinelli | SAS | 10.885366 | 5.539114 |
| 513 | Yogi Ferrell | SAC | 10.096875 | 5.574914 |
| 318 | Troy Daniels | LAL | 7.640625 | 5.629580 |
| 480 | Semi Ojeleye | BOS | 8.165116 | 5.659627 |
| 112 | Dzanan Musa | BKN | 11.570833 | 5.664495 |
Taking a look at this list of players with the lowest standard deviations, we can pretty quickly notice that it is mostly comprised of players that either do not play in very many games, or players that we already know were traded over the course of the season. In order to keep track of this so we don't have to manually identify these outliers every time, we are going to introduce a new data field called Key. In this field we are going to concatenate the Player and Team fields, then run a quick count function over our full dataframe (not the temporary one where we dropped the duplicates) to see how many games that player played for that team to produce these numbers.
dfPlayers['Key'] = dfPlayers['Player'] + dfPlayers['Team']
dfPlayerTeam = dfPlayers.groupby('Key').count()C:\Users\nfwya\Anaconda3\lib\site-packages\ipykernel_launcher.py:1: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
"""Entry point for launching an IPython kernel.dfPlayerTeam.reset_index(inplace=True)
dfPlayerTeam = dfPlayerTeam[['Key', 'Player']]dfPlayerTeam| Key | Player | |
|---|---|---|
| 0 | Aaron GordonORL | 49 |
| 1 | Aaron HolidayIND | 49 |
| 2 | Abdel NaderOKC | 34 |
| 3 | Al HorfordPHI | 50 |
| 4 | Al-Farouq AminuORL | 17 |
| ... | ... | ... |
| 364 | Will BartonDEN | 48 |
| 365 | Willie Cauley-SteinDAL | 7 |
| 366 | Willie Cauley-SteinGSW | 41 |
| 367 | Yogi FerrellSAC | 32 |
| 368 | Zach LaVineCHI | 55 |
369 rows × 2 columns
Now that we have the number of games each player played for their respective team at the time this data was pulled, we can go ahead and add this column to our main dataframe for analysis.
dfPlayers = dfPlayers.merge(dfPlayerTeam, on=['Key'])dfPlayers.columns = ['Player', 'Team', 'mean', 'std', 'Key', 'games']dfPlayers.drop_duplicates().head()| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 0 | Anfernee Simons | POR | 14.944444 | 7.667123 | Anfernee SimonsPOR | 54 |
| 54 | CJ McCollum | POR | 33.944231 | 10.935020 | CJ McCollumPOR | 52 |
| 106 | Kyle Korver | MIL | 12.036957 | 6.266112 | Kyle KorverMIL | 46 |
| 152 | Kyle Kuzma | LAL | 20.513636 | 9.673642 | Kyle KuzmaLAL | 44 |
| 196 | LeBron James | LAL | 51.711765 | 9.848099 | LeBron JamesLAL | 51 |
After that, we can now see how many games a player has played to formulate these metrics. Let's go ahead and stick with our Marcus Morris example to see how that works.
dfPlayers.drop_duplicates().loc[dfPlayers['Player'] == 'Marcus Morris Sr.']| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 247 | Marcus Morris Sr. | NYK | 29.909302 | 10.926286 | Marcus Morris Sr.NYK | 43 |
| 14049 | Marcus Morris Sr. | LAC | 24.966667 | 2.173323 | Marcus Morris Sr.LAC | 3 |
Now that we have the number of games as a metric, we can easily see that we shouldn't put much faith in the small standard deviation for Marcus with the clippers since he only has 3 games under his belt contributing to it. We can probably feel confident about his average since it is fairly close to his previous average with the Knicks, as long as there hasn't been a major injury or anything elevating his playtime in these 3 games (but it's the Clippers so let's be real. PG and Kawhi only play when they feel like it).
Now that we have the metrics we need, we're going to go ahead and create a new permanent dataframe, with the duplicates dropped for good this time, and we are also going to filter out player that have played less than 20 games for their respective team. I arbitrarily picked 20 games here, this is a variable that will depend on how many games have been played in the season, and your comfort with smaller sample sizes.
df20 = dfPlayers[dfPlayers['games'] >= 20].drop_duplicates()
df20.describe()| mean | std | games | |
|---|---|---|---|
| count | 317.000000 | 317.000000 | 317.000000 |
| mean | 23.719149 | 9.421841 | 43.952681 |
| std | 10.065573 | 1.846430 | 8.967291 |
| min | 7.640625 | 5.060675 | 20.000000 |
| 25% | 15.954545 | 8.100564 | 38.000000 |
| 50% | 21.855769 | 9.519882 | 46.000000 |
| 75% | 29.602439 | 10.471087 | 52.000000 |
| max | 57.629167 | 15.366266 | 57.000000 |
As we can see here, filtering out players with less than 20 games leaves us with a much more expected set of standard deviations. The lowest being 5.06, and the average being 9.4 . Let's go ahead and take a look at the 15 players with the lowest standard deviations in this filtered subset.
df20.sort_values('std').head(15)| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 8222 | Theo Pinson | BKN | 11.204348 | 5.060675 | Theo PinsonBKN | 23 |
| 7957 | Matthew Dellavedova | CLE | 8.887500 | 5.075744 | Matthew DellavedovaCLE | 40 |
| 12068 | Marco Belinelli | SAS | 10.885366 | 5.539114 | Marco BelinelliSAS | 41 |
| 12440 | Yogi Ferrell | SAC | 10.096875 | 5.574914 | Yogi FerrellSAC | 32 |
| 10097 | Troy Daniels | LAL | 7.640625 | 5.629580 | Troy DanielsLAL | 32 |
| 12221 | Semi Ojeleye | BOS | 8.165116 | 5.659627 | Semi OjeleyeBOS | 43 |
| 5919 | Dzanan Musa | BKN | 11.570833 | 5.664495 | Dzanan MusaBKN | 24 |
| 8551 | Terrance Ferguson | OKC | 10.105000 | 5.686419 | Terrance FergusonOKC | 40 |
| 4667 | Justin Jackson | DAL | 11.520930 | 5.969588 | Justin JacksonDAL | 43 |
| 9820 | Mike Scott | PHI | 11.566038 | 6.128965 | Mike ScottPHI | 53 |
| 5754 | Georges Niang | UTA | 10.667442 | 6.136996 | Georges NiangUTA | 43 |
| 12949 | Ryan Arcidiacono | CHI | 10.051064 | 6.157494 | Ryan ArcidiaconoCHI | 47 |
| 12595 | Noah Vonleh | MIN | 12.916000 | 6.225129 | Noah VonlehMIN | 25 |
| 14098 | Jeff Green | UTA | 13.306667 | 6.254154 | Jeff GreenUTA | 30 |
| 106 | Kyle Korver | MIL | 12.036957 | 6.266112 | Kyle KorverMIL | 46 |
Looking at this subset of players, it's not too hard to figure out why the standard deviations are so low. They just don't play a ton, and when they do they aren't a primary option, so they typically get the same looks and the same minutes every game that they do play. Another thing that may be beneficial for creating player pools would be filtering out players with an average below a certain threshold just like we did for players below the number of games threshold.
df2020 = df20[df20['mean'] >= 20]
df2020.sort_values('std').head(15)| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 976 | Terrence Ross | ORL | 22.060377 | 6.974330 | Terrence RossORL | 53 |
| 5031 | Danilo Gallinari | OKC | 30.508511 | 7.040313 | Danilo GallinariOKC | 47 |
| 1992 | Evan Fournier | ORL | 28.357407 | 7.425265 | Evan FournierORL | 54 |
| 3823 | Collin Sexton | CLE | 28.822222 | 7.483710 | Collin SextonCLE | 54 |
| 5375 | Dorian Finney-Smith | DAL | 21.092727 | 7.626364 | Dorian Finney-SmithDAL | 55 |
| 3380 | Cedi Osman | CLE | 20.314815 | 7.631771 | Cedi OsmanCLE | 54 |
| 1760 | Donte DiVincenzo | MIL | 22.754167 | 7.777024 | Donte DiVincenzoMIL | 48 |
| 7703 | Maxi Kleber | DAL | 20.112963 | 7.798283 | Maxi KleberDAL | 54 |
| 5873 | Gary Harris | DEN | 20.600000 | 7.953532 | Gary HarrisDEN | 46 |
| 12399 | Willie Cauley-Stein | GSW | 23.634146 | 8.024388 | Willie Cauley-SteinGSW | 41 |
| 3985 | Bojan Bogdanovic | UTA | 28.577358 | 8.046156 | Bojan BogdanovicUTA | 53 |
| 3471 | Joe Harris | BKN | 23.548077 | 8.067254 | Joe HarrisBKN | 52 |
| 11908 | Garrett Temple | BKN | 20.454167 | 8.076469 | Garrett TempleBKN | 48 |
| 5706 | Glenn Robinson III | GSW | 23.847917 | 8.100564 | Glenn Robinson IIIGSW | 48 |
| 5988 | Ivica Zubac | LAC | 20.785455 | 8.119374 | Ivica ZubacLAC | 55 |
These are going to be the players with the lowest standard deviations after we filter out all players with a per game average of less than 20 fpts. These would be good filler players to take a look at if they are cheap and have a favorable matchup because you can be pretty confident with their floor.
Let's also take a look at the players with the highest per game average and see how much different their standard deviations are.
df2020.sort_values('mean', ascending=False).head(15)| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 2220 | Giannis Antetokounmpo | MIL | 57.629167 | 10.903698 | Giannis AntetokounmpoMIL | 48 |
| 10935 | James Harden | HOU | 57.403922 | 14.117804 | James HardenHOU | 51 |
| 9542 | Luka Doncic | DAL | 53.890698 | 12.238519 | Luka DoncicDAL | 43 |
| 3000 | Anthony Davis | LAL | 52.032609 | 13.066106 | Anthony DavisLAL | 46 |
| 196 | LeBron James | LAL | 51.711765 | 9.848099 | LeBron JamesLAL | 51 |
| 13659 | Karl-Anthony Towns | MIN | 49.260000 | 11.960086 | Karl-Anthony TownsMIN | 35 |
| 9999 | Russell Westbrook | HOU | 49.133333 | 11.790963 | Russell WestbrookHOU | 45 |
| 4528 | Kawhi Leonard | LAC | 48.547619 | 9.790494 | Kawhi LeonardLAC | 42 |
| 4758 | Andre Drummond | DET | 48.367347 | 12.757749 | Andre DrummondDET | 49 |
| 9894 | Trae Young | ATL | 47.630000 | 15.366266 | Trae YoungATL | 50 |
| 2946 | Damian Lillard | POR | 47.620370 | 12.798178 | Damian LillardPOR | 54 |
| 13516 | Kyrie Irving | BKN | 46.130000 | 14.420712 | Kyrie IrvingBKN | 20 |
| 8699 | Nikola Jokic | DEN | 45.792727 | 13.050612 | Nikola JokicDEN | 55 |
| 10394 | Joel Embiid | PHI | 45.474359 | 11.946123 | Joel EmbiidPHI | 39 |
| 1243 | Bradley Beal | WAS | 44.806522 | 11.874144 | Bradley BealWAS | 46 |
The players that make up the highest average fpts list is no surprise, however some of the standard deviations may be alittle surprising. While most of them are with a few points of each other, this can help drive which high dollar players to roll with in gpp's vs cash games depending on their matchups.
Now, for strictly gpp purposes, lets' take a look at the players with the HIGHEST standard devations to see the high risk plays.
df2020.sort_values('std', ascending=False).head(15)| Player | Team | mean | std | Key | games | |
|---|---|---|---|---|---|---|
| 9894 | Trae Young | ATL | 47.630000 | 15.366266 | Trae YoungATL | 50 |
| 13516 | Kyrie Irving | BKN | 46.130000 | 14.420712 | Kyrie IrvingBKN | 20 |
| 10935 | James Harden | HOU | 57.403922 | 14.117804 | James HardenHOU | 51 |
| 8245 | Paul George | LAC | 37.467647 | 13.513511 | Paul GeorgeLAC | 34 |
| 13483 | D'Angelo Russell | GSW | 37.981818 | 13.449918 | D'Angelo RussellGSW | 33 |
| 11014 | Jabari Parker | ATL | 28.350000 | 13.209039 | Jabari ParkerATL | 32 |
| 2268 | Hassan Whiteside | POR | 42.776471 | 13.098131 | Hassan WhitesidePOR | 51 |
| 3000 | Anthony Davis | LAL | 52.032609 | 13.066106 | Anthony DavisLAL | 46 |
| 8699 | Nikola Jokic | DEN | 45.792727 | 13.050612 | Nikola JokicDEN | 55 |
| 2591 | Jordan McRae | WAS | 24.307143 | 12.865974 | Jordan McRaeWAS | 28 |
| 2872 | Gorgui Dieng | MIN | 20.515556 | 12.837462 | Gorgui DiengMIN | 45 |
| 3577 | Christian Wood | DET | 22.750980 | 12.827180 | Christian WoodDET | 51 |
| 2946 | Damian Lillard | POR | 47.620370 | 12.798178 | Damian LillardPOR | 54 |
| 4758 | Andre Drummond | DET | 48.367347 | 12.757749 | Andre DrummondDET | 49 |
| 10433 | Clint Capela | HOU | 38.753846 | 12.661937 | Clint CapelaHOU | 39 |
This grouping of players is a nice mix of high performers and middle of the pack guys, but that is good since it demonstrates there are valuable high risk plays at all levels of salary. These picks will be very matchup dependent, and likely lineup dependent for guys like jabari parker and gorgui dieng. But it is always good to know which players have the biggest difference in impact based on those scenarios to factor into your lineups.
Like every other metric and predictive measure, there is no one correct metric to use. Merely more tools to put into your toolbox to use when appropriate. Now you can add standard deviation to that list and begin taking a look at what types of players are consistent, and factor in how likely they are to be scoring within a certain threshold given a favorable/unfavorable matchup.
