ALL NBA Prediction Model (Logistic Regression Analysis)
Cooper Martin
2026-10-09
Research Question: Can we predict the ALL NBA teams using statistical modelling?
A prediction point must be determined, when are we making this prediction? Our prediction point will be after the buzzer sounds of the final NBA Regular season game.
What Data are we going to use? We are going to use the last 11 Seasons of NBA data. From and including the 2015/16 season up to and including the 2025/26 season.
What model are we going to use? Logistic Regression will be used to determine two outcomes:
- % chance a player will make an ALL NBA team
- Prediction of each ALL NBA team
Data Collection and Preparation
Data Collection
The dataset was manually compiled from Basketball-Reference.com from the 2015–16 to 2025–26 NBA seasons. Rather than using a single dataset, three tables were retrieved from each season:
Player Per Game Statistics: Individual player statistics for each season. Rows containing partial-season statistics were excluded to avoid duplicate player entries where a player appeared for multiple teams during the same season. Team Advanced Statistics: Advanced statistical measures for each NBA team during the regular season.
All-NBA Voting: Voting results used to identify players selected to an All-NBA Team.
Team Advanced Statistics: Team based Advanced statistics for each season. This was used to incorporate team level performance.
Each table was manually retrieved in CSV format from its respective season page. The tables were then organised into separate spreadsheets for further preparation before being imported into R.
Data Preparation
Before importing the data, two adjustments were made to the Team Advanced Statistics table in Excel. Firstly, a new column was made to mark if a team made the playoffs that season (yes = 1, no = 0). Basketball-reference.com mark these teams with a asterisks (*). After that was complete, any asterisks (*) from team names were removed. Finally, each team was assigned its three-letter abbreviation to match the team identifiers used in the Player Per Game Statistics table.
All (%) symbols were replaced with “_pct”
These adjustments were made to improve consistency between the tables and support the merging of player-level statistics with team-level data during the analysis.
1. Load Key Packages
Use Install_packages() if you do not have them installed
library(dplyr) ## Warning: package 'dplyr' was built under R version 4.5.3library(tidyr)
library(ggplot2)## Warning: package 'ggplot2' was built under R version 4.5.3library(gamlss)## Warning: package 'gamlss' was built under R version 4.5.3## Warning: package 'gamlss.dist' was built under R version 4.5.3library(readr)2. Collect & Import Data
As this data was collated manually, I will provide the best guide possible to follow.
Dataframe Names for each csv file
- All-NBA Voting = allnbavoting
- Team Advanced Statistics = teamadv
- Player Per Game Statistics = pergameraw
3. Data manipulation
View variable types
Ensure variable types are correct
glimpse(pergameraw)## Rows: 7,595
## Columns: 33
## $ Year <int> 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026…
## $ Rk <int> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 1…
## $ Player <chr> "Luka Dončić", "Shai Gilgeous-Alexander", "Anthony E…
## $ Age <int> 26, 27, 24, 29, 25, 34, 29, 30, 31, 31, 28, 37, 29, …
## $ Team <chr> "LAL", "OKC", "MIN", "BOS", "PHI", "LAC", "CLE", "DE…
## $ Pos <chr> "PG", "PG", "SG", "SF", "PG", "SF", "SG", "C", "PF",…
## $ G <int> 64, 68, 61, 71, 70, 65, 70, 65, 36, 38, 42, 43, 64, …
## $ GS <int> 64, 68, 60, 71, 70, 65, 70, 65, 36, 38, 42, 41, 64, …
## $ MP <dbl> 35.8, 33.2, 35.0, 34.4, 38.0, 32.1, 33.5, 34.8, 28.9…
## $ FG <dbl> 10.8, 10.8, 9.9, 10.4, 9.9, 9.8, 9.7, 9.9, 10.4, 9.0…
## $ FGA <dbl> 22.8, 19.4, 20.2, 21.7, 21.4, 19.4, 20.0, 17.4, 16.6…
## $ FG_pct <dbl> 0.476, 0.553, 0.489, 0.477, 0.462, 0.505, 0.483, 0.5…
## $ X3P <dbl> 4.0, 1.7, 3.4, 2.0, 3.1, 2.6, 3.2, 1.7, 0.4, 1.4, 2.…
## $ X3PA <dbl> 10.8, 4.4, 8.4, 5.7, 8.6, 6.8, 8.8, 4.5, 1.3, 4.2, 7…
## $ X3P_pct <dbl> 0.366, 0.386, 0.399, 0.347, 0.367, 0.387, 0.364, 0.3…
## $ X2P <dbl> 6.9, 9.1, 6.5, 8.4, 6.8, 7.1, 6.5, 8.2, 9.9, 7.6, 6.…
## $ X2PA <dbl> 11.9, 15.0, 11.8, 16.0, 12.9, 12.5, 11.2, 12.9, 15.4…
## $ X2P_pct <dbl> 0.575, 0.602, 0.554, 0.523, 0.525, 0.569, 0.577, 0.6…
## $ eFG_pct <dbl> 0.563, 0.597, 0.572, 0.522, 0.536, 0.573, 0.563, 0.6…
## $ FT <dbl> 7.9, 7.9, 5.7, 6.0, 5.3, 5.7, 5.3, 6.1, 6.4, 7.5, 5.…
## $ FTA <dbl> 10.1, 9.0, 7.2, 7.5, 6.0, 6.4, 6.1, 7.4, 9.9, 8.8, 6…
## $ FT_pct <dbl> 0.780, 0.879, 0.796, 0.795, 0.892, 0.892, 0.865, 0.8…
## $ ORB <dbl> 0.6, 0.6, 0.6, 1.1, 0.3, 1.1, 0.7, 3.0, 2.7, 2.0, 2.…
## $ DRB <dbl> 7.1, 3.7, 4.4, 5.8, 3.8, 5.3, 3.8, 9.9, 7.1, 5.7, 4.…
## $ TRB <dbl> 7.7, 4.3, 5.0, 6.9, 4.1, 6.4, 4.5, 12.9, 9.8, 7.7, 6…
## $ AST <dbl> 8.3, 6.6, 3.7, 5.1, 6.6, 3.6, 5.7, 10.7, 5.4, 3.9, 2…
## $ STL <dbl> 1.6, 1.4, 1.4, 1.0, 1.9, 1.9, 1.5, 1.4, 0.9, 0.6, 1.…
## $ BLK <dbl> 0.5, 0.8, 0.8, 0.4, 0.8, 0.4, 0.3, 0.8, 0.7, 1.2, 0.…
## $ TOV <dbl> 4.0, 2.2, 2.9, 3.6, 2.4, 2.0, 2.8, 3.7, 3.2, 2.9, 1.…
## $ PF <dbl> 2.4, 2.0, 1.9, 2.7, 2.2, 1.2, 2.3, 2.7, 2.4, 2.2, 1.…
## $ PTS <dbl> 33.5, 31.1, 28.8, 28.7, 28.3, 27.9, 27.9, 27.7, 27.6…
## $ Awards <chr> "MVP-4CPOY-8ASNBA1", "MVP-1CPOY-1ASNBA1", "CPOY-3AS"…
## $ Player.additional <chr> "doncilu01", "gilgesh01", "edwaran01", "brownja02", …glimpse(allnbavoting)## Rows: 341
## Columns: 25
## $ Year <int> 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2…
## $ Team <chr> "1T", "1T", "1T", "1T", "1T", "2T", "2T", "2T", "2T", "2T", "…
## $ Pos <chr> "G", "C", "C", "G", "G", "F", "F", "G", "F", "G", "G", "G", "…
## $ Player <chr> "Shai Gilgeous-Alexander", "Nikola Jokić", "Victor Wembanyama…
## $ Age <int> 27, 30, 22, 26, 24, 29, 34, 29, 37, 29, 25, 28, 24, 22, 23, 2…
## $ Tm <chr> "OKC", "DEN", "SAS", "LAL", "DET", "BOS", "LAC", "CLE", "HOU"…
## $ Pts.Won <int> 500, 500, 498, 482, 414, 384, 277, 276, 241, 197, 168, 149, 1…
## $ Pts.Max <int> 500, 500, 500, 500, 500, 500, 500, 500, 500, 500, 500, 500, 5…
## $ Share <dbl> 1.000, 1.000, 0.996, 0.964, 0.828, 0.768, 0.554, 0.552, 0.482…
## $ X1st_Tm <int> 100, 100, 99, 91, 60, 44, 4, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
## $ X2nd_Tm <int> 0, 0, 1, 9, 38, 54, 81, 85, 72, 49, 36, 27, 19, 16, 8, 1, 1, …
## $ X3rd_Tm <int> 0, 0, 0, 0, 0, 2, 14, 11, 25, 50, 60, 68, 68, 73, 63, 23, 11,…
## $ G <int> 68, 65, 64, 64, 64, 71, 65, 70, 78, 74, 70, 75, 72, 70, 69, 6…
## $ MP <dbl> 33.2, 34.8, 29.2, 35.8, 33.9, 34.4, 32.1, 33.5, 36.4, 35.0, 3…
## $ PTS <dbl> 31.1, 27.7, 25.0, 33.5, 23.9, 28.7, 27.9, 27.9, 26.0, 26.0, 2…
## $ TRB <dbl> 4.3, 12.9, 11.5, 7.7, 5.5, 6.9, 6.4, 4.5, 5.5, 3.3, 4.1, 4.4,…
## $ AST <dbl> 6.6, 10.7, 3.1, 8.3, 9.9, 5.1, 3.6, 5.7, 4.8, 6.8, 6.6, 7.1, …
## $ STL <dbl> 1.4, 1.4, 1.0, 1.6, 1.4, 1.0, 1.9, 1.5, 0.8, 0.8, 1.9, 0.9, 1…
## $ BLK <dbl> 0.8, 0.8, 3.1, 0.5, 0.8, 0.4, 0.4, 0.3, 0.9, 0.1, 0.8, 0.4, 0…
## $ FG_pct <dbl> 0.553, 0.569, 0.512, 0.476, 0.461, 0.477, 0.505, 0.483, 0.520…
## $ X3P_pct <dbl> 0.386, 0.380, 0.349, 0.366, 0.342, 0.347, 0.387, 0.364, 0.413…
## $ FT_pct <dbl> 0.879, 0.831, 0.827, 0.780, 0.812, 0.795, 0.892, 0.865, 0.874…
## $ WS <dbl> 15.2, 14.9, 10.0, 9.5, 7.9, 6.9, 9.2, 8.2, 10.7, 8.8, 8.7, 9.…
## $ WS.48 <dbl> 0.323, 0.316, 0.257, 0.199, 0.174, 0.135, 0.212, 0.169, 0.180…
## $ playerid <chr> "gilgesh01", "jokicni01", "wembavi01", "doncilu01", "cunnica0…glimpse(teamadv)## Rows: 341
## Columns: 34
## $ Year <int> 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024, 20…
## $ Rk <int> 17, 9, 15, 25, 25, 22, 23, 23, 28, 30, 30, 10, 6, 4, 4, …
## $ Team <chr> "Washington Wizards", "Washington Wizards", "Washington …
## $ Abr <chr> "WAS", "WAS", "WAS", "WAS", "WAS", "WAS", "WAS", "WAS", …
## $ Age <dbl> 27.3, 26.0, 26.9, 26.5, 25.1, 26.6, 25.9, 26.2, 24.9, 23…
## $ W <int> 41, 49, 43, 32, 25, 34, 35, 35, 15, 18, 17, 40, 51, 48, …
## $ L <int> 41, 33, 39, 50, 47, 38, 47, 47, 67, 64, 65, 42, 31, 34, …
## $ PW <int> 40, 46, 43, 34, 26, 32, 32, 38, 20, 15, 16, 46, 52, 53, …
## $ PL <int> 42, 36, 39, 48, 46, 40, 50, 44, 62, 67, 66, 36, 30, 29, …
## $ MOV <dbl> -0.50, 1.80, 0.59, -2.90, -4.67, -1.83, -3.38, -1.21, -9…
## $ SOS <dbl> 0.00, -0.45, -0.06, -0.40, -0.57, -0.01, 0.15, 0.15, 0.0…
## $ SRS <dbl> -0.50, 1.36, 0.53, -3.30, -5.24, -1.85, -3.23, -1.06, -9…
## $ ORtg <dbl> 105.3, 111.2, 109.3, 111.1, 110.9, 111.2, 111.1, 114.4, …
## $ DRtg <dbl> 105.8, 109.3, 108.7, 113.9, 115.5, 113.0, 114.5, 115.6, …
## $ NRtg <dbl> -0.5, 1.9, 0.6, -2.8, -4.6, -1.8, -3.4, -1.2, -9.1, -12.…
## $ Pace <dbl> 98.5, 97.4, 96.6, 101.4, 102.7, 104.1, 97.0, 98.6, 102.7…
## $ FTr <dbl> 0.263, 0.254, 0.254, 0.266, 0.270, 0.288, 0.252, 0.258, …
## $ X3PAr <dbl> 0.282, 0.284, 0.310, 0.370, 0.358, 0.319, 0.356, 0.365, …
## $ TS_pct <dbl> 0.544, 0.564, 0.560, 0.567, 0.562, 0.569, 0.568, 0.585, …
## $ X <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, …
## $ eFG_pct <dbl> 0.511, 0.528, 0.525, 0.531, 0.523, 0.531, 0.532, 0.550, …
## $ TOV_pct <dbl> 13.1, 12.8, 13.3, 12.3, 12.2, 12.3, 12.1, 12.7, 12.2, 13…
## $ ORB_pct <dbl> 20.6, 24.1, 23.5, 21.3, 22.2, 21.3, 20.9, 22.6, 20.0, 22…
## $ FT.FGA <dbl> 0.192, 0.199, 0.196, 0.204, 0.213, 0.221, 0.197, 0.202, …
## $ X.1 <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, …
## $ eFG_pct.1 <dbl> 0.515, 0.524, 0.522, 0.546, 0.558, 0.539, 0.529, 0.540, …
## $ TOV_pct.1 <dbl> 14.6, 13.8, 13.6, 13.5, 13.9, 12.5, 10.7, 11.0, 12.0, 11…
## $ DRB_pct <dbl> 77.7, 75.5, 77.1, 74.1, 75.3, 77.6, 76.9, 76.1, 72.5, 71…
## $ FT.FGA.1 <dbl> 0.218, 0.213, 0.212, 0.199, 0.231, 0.217, 0.202, 0.194, …
## $ X.2 <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, …
## $ Arena <chr> "Verizon Center", "Verizon Center", "Capital One Arena",…
## $ Attend. <int> 725426, 697107, 739302, 716996, 532702, 19198, 637215, 7…
## $ Attend..G <int> 17693, 17003, 18032, 17448, 16647, 533, 15542, 17329, 16…
## $ Made_Playoffs <int> 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0,…Change variable types and remove useless rows
library(dplyr)
# Remove league average rows
pergameraw <- pergameraw |>
filter(Player != "League Average")
teamadv <- teamadv |>
filter(Team != "League Average")
# Correct variable type
pergameraw <- pergameraw |>
mutate(across(
c(Year, Age, G, GS, MP, FG, FGA, FG_pct, X3P, X3PA, X3P_pct,
X2P, X2PA, X2P_pct, eFG_pct, FT, FTA, FT_pct, ORB, DRB, TRB,
AST, STL, BLK, TOV, PF, PTS),
as.numeric
))
allnbavoting <- allnbavoting |>
mutate(across(
c(Year, Age, Pts.Won, Pts.Max, Share, X1st_Tm, X2nd_Tm,
X3rd_Tm, G, MP, PTS, TRB, AST, STL, BLK, FG_pct, X3P_pct,
FT_pct, WS, WS.48),
as.numeric
))
teamadv <- teamadv |>
mutate(across(
c(Year, Rk, Age, W, L, PW, PL, MOV, SOS, SRS, ORtg,
DRtg, NRtg, Pace, FTr, X3PAr, TS_pct, eFG_pct, TOV_pct, ORB_pct,
FT.FGA, eFG_pct.1, TOV_pct.1, DRB_pct, FT.FGA.1, Attend.,
Attend..G, Made_Playoffs),
as.numeric
))Standardizing labels between sheets
Some labels are different despite representing the same data.
teamadv <- teamadv |>
rename(TeamId = Abr)
pergameraw <- pergameraw |>
rename(TeamId = Team)
pergameraw <- pergameraw |>
rename(playerid = Player.additional)
allnbavoting <- allnbavoting |>
rename(TeamId = Tm)Assigning all players a team
If a player played for multiple teams, the team they played the most games for will be assigned to them
pergame <- pergameraw |>
filter(!TeamId %in% c("2TM", "3TM", "4TM")) |>
group_by(playerid, Year) |>
filter(row_number() == which.max(G)) |>
ungroup()Adding important columns
pergame <- pergame |>
left_join(allnbavoting |>
select(playerid, Year, Team, Share),
by = c('playerid', 'Year'))
pergame <- pergame |>
left_join(teamadv |>
select(TeamId, Year, Made_Playoffs, W, L),
by = c('Year', 'TeamId'))
pergame <- pergame |>
mutate(allnba = if_else(Team %in% c('1st', '1T', '2nd', '2T', '3rd', '3T'), 1, 0))summary(pergame)## Year Rk Player Age
## Min. :2016 Min. : 1 Length:5968 Min. :19.0
## 1st Qu.:2018 1st Qu.:136 Class :character 1st Qu.:23.0
## Median :2021 Median :272 Mode :character Median :25.0
## Mean :2021 Mean :273 Mean :25.9
## 3rd Qu.:2024 3rd Qu.:407 3rd Qu.:29.0
## Max. :2026 Max. :605 Max. :43.0
##
## TeamId Pos G GS
## Length:5968 Length:5968 Min. : 1.00 Min. : 0.0
## Class :character Class :character 1st Qu.:23.00 1st Qu.: 0.0
## Mode :character Mode :character Median :49.00 Median : 7.0
## Mean :45.17 Mean :21.5
## 3rd Qu.:68.00 3rd Qu.:40.0
## Max. :82.00 Max. :82.0
##
## MP FG FGA FG_pct
## Min. : 0.50 Min. : 0.000 Min. : 0.000 Min. :0.0000
## 1st Qu.:12.00 1st Qu.: 1.400 1st Qu.: 3.300 1st Qu.:0.4030
## Median :19.30 Median : 2.600 Median : 5.800 Median :0.4460
## Mean :19.48 Mean : 3.186 Mean : 6.958 Mean :0.4486
## 3rd Qu.:27.40 3rd Qu.: 4.400 3rd Qu.: 9.600 3rd Qu.:0.4980
## Max. :43.50 Max. :11.800 Max. :24.500 Max. :1.0000
## NA's :30
## X3P X3PA X3P_pct X2P
## Min. :0.0000 Min. : 0.000 Min. :0.0000 Min. : 0.000
## 1st Qu.:0.2000 1st Qu.: 0.800 1st Qu.:0.2780 1st Qu.: 0.900
## Median :0.7000 Median : 2.200 Median :0.3390 Median : 1.700
## Mean :0.9178 Mean : 2.612 Mean :0.3126 Mean : 2.269
## 3rd Qu.:1.4000 3rd Qu.: 3.900 3rd Qu.:0.3810 3rd Qu.: 3.200
## Max. :5.3000 Max. :13.200 Max. :1.0000 Max. :11.600
## NA's :379
## X2PA X2P_pct eFG_pct FT
## Min. : 0.000 Min. :0.0000 Min. :0.0000 Min. : 0.00
## 1st Qu.: 1.800 1st Qu.:0.4620 1st Qu.:0.4760 1st Qu.: 0.50
## Median : 3.300 Median :0.5140 Median :0.5210 Median : 0.90
## Mean : 4.348 Mean :0.5114 Mean :0.5115 Mean : 1.36
## 3rd Qu.: 6.000 3rd Qu.:0.5710 3rd Qu.:0.5640 3rd Qu.: 1.80
## Max. :19.200 Max. :1.0000 Max. :1.5000 Max. :10.20
## NA's :76 NA's :30
## FTA FT_pct ORB DRB
## Min. : 0.000 Min. :0.0000 Min. :0.0000 Min. : 0.000
## 1st Qu.: 0.600 1st Qu.:0.6820 1st Qu.:0.3000 1st Qu.: 1.400
## Median : 1.300 Median :0.7680 Median :0.6000 Median : 2.300
## Mean : 1.771 Mean :0.7465 Mean :0.8572 Mean : 2.687
## 3rd Qu.: 2.300 3rd Qu.:0.8340 3rd Qu.:1.1000 3rd Qu.: 3.600
## Max. :12.300 Max. :1.0000 Max. :5.4000 Max. :11.400
## NA's :308
## TRB AST STL BLK
## Min. : 0.000 Min. : 0.000 Min. :0.0000 Min. :0.0000
## 1st Qu.: 1.800 1st Qu.: 0.700 1st Qu.:0.3000 1st Qu.:0.1000
## Median : 3.050 Median : 1.400 Median :0.6000 Median :0.3000
## Mean : 3.541 Mean : 1.958 Mean :0.6291 Mean :0.3936
## 3rd Qu.: 4.700 3rd Qu.: 2.600 3rd Qu.:0.9000 3rd Qu.:0.5000
## Max. :16.000 Max. :11.700 Max. :3.0000 Max. :3.8000
##
## TOV PF PTS Awards
## Min. :0.000 Min. :0.00 Min. : 0.000 Length:5968
## 1st Qu.:0.500 1st Qu.:1.10 1st Qu.: 3.900 Class :character
## Median :0.900 Median :1.70 Median : 7.000 Mode :character
## Mean :1.085 Mean :1.65 Mean : 8.646
## 3rd Qu.:1.500 3rd Qu.:2.20 3rd Qu.:11.900
## Max. :5.700 Max. :6.00 Max. :36.100
##
## playerid Team Share Made_Playoffs
## Length:5968 Length:5968 Min. :0.0020 Min. :0.0000
## Class :character Class :character 1st Qu.:0.0100 1st Qu.:0.0000
## Mode :character Mode :character Median :0.1120 Median :1.0000
## Mean :0.2903 Mean :0.5171
## 3rd Qu.:0.5160 3rd Qu.:1.0000
## Max. :1.0000 Max. :1.0000
## NA's :5627
## W L allnba
## Min. :10.00 Min. : 9.00 Min. :0.00000
## 1st Qu.:31.00 1st Qu.:32.00 1st Qu.:0.00000
## Median :42.00 Median :39.00 Median :0.00000
## Mean :39.53 Mean :40.56 Mean :0.02765
## 3rd Qu.:48.00 3rd Qu.:49.00 3rd Qu.:0.00000
## Max. :73.00 Max. :72.00 Max. :1.00000
## Visualise variable relationships
library(GGally)
ggpairplots <- ggpairs(data = pergame, columns = c("allnba",
"PTS",
"Share",
"Made_Playoffs",
"W",
"L",
"TRB",
"AST",
"STL",
"BLK",
"eFG_pct"))
ggpairplots
Split data into training and test datasets
We’ll train on the first 9 seasons (15/16 - 23/24), and test on the final 2 (24/25 & 25/26)
train <- pergame |>
filter(Year < 2025)
test <- pergame |>
filter(Year >= 2025)Fit 5 Different models with different reasoning
# Model 1: Basic PRA and wins
fit1 <- glm(
allnba ~ PTS + AST + TRB + W,
data = train,
family = binomial
)
# Model 2: Scoring contribution and efficiency
fit2 <- glm(
allnba ~ PTS + AST + FG_pct + X3P_pct + FT_pct + eFG_pct,
data = train,
family = binomial
)
# Model 3: Counting Stats performance
fit3 <- glm(
allnba ~ PTS + AST + TRB + STL + BLK + TOV + FG + X2P + X3P + FT,
data = train,
family = binomial
)
# Model 4: Playing time and experience
fit4 <- glm(
allnba ~ Age + G + GS + MP + PTS + AST + TRB,
data = train,
family = binomial
)
# Model 5: Comprehensive model (Subjectively Hand Picked)
fit5 <- glm(
allnba ~ G + GS + MP + PTS + AST + TRB +
STL + BLK + TOV + FT_pct + eFG_pct +
W + Made_Playoffs,
data = train,
family = binomial
)Create Probabilities and Predictions on Test Set
test <- test |>
mutate(
prob1 = predict(fit1, newdata = test, type = "response"),
prob2 = predict(fit2, newdata = test, type = "response"),
prob3 = predict(fit3, newdata = test, type = "response"),
prob4 = predict(fit4, newdata = test, type = "response"),
prob5 = predict(fit5, newdata = test, type = "response")
)
test <- test |>
mutate(
pred1 = if_else(prob1 >= 0.5, 1, 0),
pred2 = if_else(prob2 >= 0.5, 1, 0),
pred3 = if_else(prob3 >= 0.5, 1, 0),
pred4 = if_else(prob4 >= 0.5, 1, 0),
pred5 = if_else(prob5 >= 0.5, 1, 0)
)Measuring accuracy of models against test set
accuracy <- tibble(
Model = paste("Model", 1:5),
Accuracy = c(
mean(test$pred1 == test$allnba, na.rm = TRUE),
mean(test$pred2 == test$allnba, na.rm = TRUE),
mean(test$pred3 == test$allnba, na.rm = TRUE),
mean(test$pred4 == test$allnba, na.rm = TRUE),
mean(test$pred5 == test$allnba, na.rm = TRUE)
)
) |>
mutate(Accuracy = round(Accuracy * 100, 2))
accuracy## # A tibble: 5 × 2
## Model Accuracy
## <chr> <dbl>
## 1 Model 1 98.9
## 2 Model 2 98.2
## 3 Model 3 97.9
## 4 Model 4 98.5
## 5 Model 5 98.9It is important to note, despite accuracy for each model being above 95%, this is significantly swayed by the fact only 15 players can make ALL NBA out of the ~500 players.
We will Compare the results of the 3 Models. The Highest scoring, Hand Picked Model (5) The Simplest Model (1) and the Lowest Scoring Model (3)
Top 30 in Probability
model1_26 <- test |> filter(Year == 2026) |> slice_max(prob1, n = 30) |> select(Player, prob1, allnba, Pos)
model3_26 <- test |> filter(Year == 2026) |> slice_max(prob3, n = 30) |> select(Player, prob3, allnba, Pos)
model5_26 <- test |> filter(Year == 2026) |> slice_max(prob5, n = 30) |> select(Player, prob5,allnba, Pos)Predicted teams for each model
# Model 1
Model_1_2026_1st <- model1_26 |> arrange(desc(prob1)) |> slice(1:5)
Model_1_2026_2nd <- model1_26 |> arrange(desc(prob1)) |> slice(6:10)
Model_1_2026_3rd <- model1_26 |> arrange(desc(prob1)) |> slice(11:15)
# Model 3
Model_3_2026_1st <- model3_26 |> arrange(desc(prob3)) |> slice(1:5)
Model_3_2026_2nd <- model3_26 |> arrange(desc(prob3)) |> slice(6:10)
Model_3_2026_3rd <- model3_26 |> arrange(desc(prob3)) |> slice(11:15)
# Model 5
Model_5_2026_1st <- model5_26 |> arrange(desc(prob5)) |> slice(1:5)
Model_5_2026_2nd <- model5_26 |> arrange(desc(prob5)) |> slice(6:10)
Model_5_2026_3rd <- model5_26 |> arrange(desc(prob5)) |> slice(11:15)Combine into one Dataframe
all_teams_2026 <- bind_rows(
Model_1_2026_1st |> mutate(Model = "Model 1", Year = 2026, Team = "1st"),
Model_1_2026_2nd |> mutate(Model = "Model 1", Year = 2026, Team = "2nd"),
Model_1_2026_3rd |> mutate(Model = "Model 1", Year = 2026, Team = "3rd"),
Model_3_2026_1st |> mutate(Model = "Model 3", Year = 2026, Team = "1st"),
Model_3_2026_2nd |> mutate(Model = "Model 3", Year = 2026, Team = "2nd"),
Model_3_2026_3rd |> mutate(Model = "Model 3", Year = 2026, Team = "3rd"),
Model_5_2026_1st |> mutate(Model = "Model 5", Year = 2026, Team = "1st"),
Model_5_2026_2nd |> mutate(Model = "Model 5", Year = 2026, Team = "2nd"),
Model_5_2026_3rd |> mutate(Model = "Model 5", Year = 2026, Team = "3rd")
) |>
select(Model, Year, Team, Player, starts_with("prob"), allnba)Model team predictions vs actual
knitr::kable(
all_teams_2026 |> select(Model, Team, Player, allnba),
caption = "Predicted All-NBA Teams for 2026"
)| Model | Team | Player | allnba |
|---|---|---|---|
| Model 1 | 1st | Nikola Jokić | 1 |
| Model 1 | 1st | Luka Dončić | 1 |
| Model 1 | 1st | Shai Gilgeous-Alexander | 1 |
| Model 1 | 1st | Victor Wembanyama | 1 |
| Model 1 | 1st | Jaylen Brown | 1 |
| Model 1 | 2nd | Cade Cunningham | 1 |
| Model 1 | 2nd | Jayson Tatum | 0 |
| Model 1 | 2nd | Jalen Johnson | 1 |
| Model 1 | 2nd | Donovan Mitchell | 1 |
| Model 1 | 2nd | Jamal Murray | 1 |
| Model 1 | 3rd | Karl-Anthony Towns | 0 |
| Model 1 | 3rd | Jalen Duren | 1 |
| Model 1 | 3rd | Anthony Edwards | 0 |
| Model 1 | 3rd | Kevin Durant | 1 |
| Model 1 | 3rd | Joel Embiid | 0 |
| Model 3 | 1st | Nikola Jokić | 1 |
| Model 3 | 1st | Luka Dončić | 1 |
| Model 3 | 1st | Shai Gilgeous-Alexander | 1 |
| Model 3 | 1st | Tyrese Maxey | 1 |
| Model 3 | 1st | Victor Wembanyama | 1 |
| Model 3 | 2nd | Kawhi Leonard | 1 |
| Model 3 | 2nd | Giannis Antetokounmpo | 0 |
| Model 3 | 2nd | Jalen Johnson | 1 |
| Model 3 | 2nd | Jayson Tatum | 0 |
| Model 3 | 2nd | Donovan Mitchell | 1 |
| Model 3 | 3rd | Lauri Markkanen | 0 |
| Model 3 | 3rd | Cade Cunningham | 1 |
| Model 3 | 3rd | James Harden | 0 |
| Model 3 | 3rd | Anthony Edwards | 0 |
| Model 3 | 3rd | Jamal Murray | 1 |
| Model 5 | 1st | Nikola Jokić | 1 |
| Model 5 | 1st | Luka Dončić | 1 |
| Model 5 | 1st | Shai Gilgeous-Alexander | 1 |
| Model 5 | 1st | Tyrese Maxey | 1 |
| Model 5 | 1st | Donovan Mitchell | 1 |
| Model 5 | 2nd | Jaylen Brown | 1 |
| Model 5 | 2nd | Jamal Murray | 1 |
| Model 5 | 2nd | Cade Cunningham | 1 |
| Model 5 | 2nd | Jalen Johnson | 1 |
| Model 5 | 2nd | Kevin Durant | 1 |
| Model 5 | 3rd | Jalen Duren | 1 |
| Model 5 | 3rd | Kawhi Leonard | 1 |
| Model 5 | 3rd | Jalen Brunson | 1 |
| Model 5 | 3rd | Alperen Şengün | 0 |
| Model 5 | 3rd | Karl-Anthony Towns | 0 |
Final Performance 2026
all_teams_2026 |>
group_by(Model) |>
summarise(All_NBA_Count = sum(allnba == 1, na.rm = TRUE))## # A tibble: 3 × 2
## Model All_NBA_Count
## <chr> <int>
## 1 Model 1 11
## 2 Model 3 10
## 3 Model 5 13Conclusions
Model 5 performed best, correctly predicting 13 of 15 All-NBA selections in 2026, compared with 11 for Model 1 and 10 for Model 3. This suggests that combining playing time, individual performance, efficiency and team success may be more effective than relying on basic statistics or a broader set of counting statistics. However, testing on a single season limits the reliability of this conclusion.
Potential methodological improvements
- Expand testing: Evaluate multiple unseen seasons to assess consistency across different NBA seasons.
- Refine predictors: Use cross-validation and regularisation to identify the most useful variables and reduce redundancy.
Future Applications
The model could be updated throughout the NBA season to track changing All-NBA predictions. The methodology could also be adapted to forecast other awards, compare predictions with media voting and identify potential snubs. With further testing and refinement, it could become a useful tool for sports analytics and data-driven content creation.