Crunching the Numbers on Arsenals Perfect Centre Forward

Introduction

In this project, I will be crunching the numbers across Europes top 5 leagues and assessing the options for Arsenal in their pursuit of the perfect number 9.

It is an area where Arsenal actualy have quite a few options; Gabriel Jesus, Viktor Gyokeres Kai Havertz and even Mikel Merino have done solid jobs in the role, and each have produced impressive purple patches. That being said, through inconsistent form and injury issues, it would appear that Arsenal are yet to find that Number 9 that in my opinion would fire them to domination in England, and perhaps Europe.

Amongst their current options * Gabriel Jesus offers versatility, pressing intensity and outstanding link up play. On the other hand, he is often wasteful in front of goal and has suffered a series of unfortunate injuries over the last 2-3 years.

Kai Havertz excels in aerial and ground duels, has very intelligent off the ball movement and link up, and offers an incredible aerial threat, particularly from set pieces. However like Gabriel Jesus, his finishing has also been hot and cold and he has also suffered a series of injuries that have left him unavailable for extended periods.
Viktor Gyokeres is powerful, robust, and offers an exceptional work ethic with an eye for goal (He was Europes Top Scorer last season with an impressive 54 goals across all competitions). While there is still potential for him to grow and adapt, he has no doubt struggled to get to grips to the pace of the Premier League.
Mikel Merino has proved to be a more then capable back up in the role, but is a natural central midfield player and has talents are best served here.

Our Purpose is to identify the best centre forward who can fit Arsenals style of play using data from the Top 5 leagues in the 2024/25 season.

To address this topic, we will utilise a 4 tier, analytical approach: Tier 1 - Elite Players. World Class Centre Forward, who may be unattainable but it is still useful to compare. Tier 2 - Realistic Targets - Proven players who are attainable and between the ages of 23-27. Tier 3 - Value Options - Short Term solutions from smaler clubs (Age 25-27) Tier 4 - Young Prospects - High potential, developing strikers (Age 19-23)

For Each tier we will: * Develop a scoring system to evaluate candidates for Arsenal. * Apply a similiarity algorithm (Euclidean and Cosine) to find players who are a stylistic match. * Create charts and visualisations to compare some top candidates.

# First we set the working directory
# Load data from same directory as .Rmd file
data <- read.csv("FBREF_BigPlayers_2425.csv", sep=";", encoding="UTF-8")

knitr::opts_chunk$set(warning = FALSE, message = FALSE)

# Next we load the relevant libraries
library(tidyverse)  # Data manipulation
library(fmsb)       # Radar charts
library(lsa)        # Cosine similarity
library(knitr)      # Tables
library(DT)         # Interactive tables
library(scales)     # Color scales
library(ggplot2)

head(data)

##            Player         Squad Nation   Pos Age MP  Min Gls G.PK Ast  xG xAG
## 1 Abdoulie Ceesay     St. Pauli    GAM    FW  20  7   60   0    0   0 0.0 0.0
## 2      Adam Aznou Bayern Munich    MAR    DF  18  2   17   0    0   0 0.0 0.0
## 3   Adam Dźwigała     St. Pauli    POL DF,MF  28 16  373   0    0   0 0.4 0.0
## 4     Adam Hložek    Hoffenheim    CZE FW,MF  22 27 1871   8    8   4 5.6 3.5
## 5     Adrian Beck    Heidenheim    GER MF,FW  27 32 1598   4    4   1 3.3 1.2
## 6   Alassane Pléa      Gladbach    FRA MF,FW  31 30 1902  11   10   4 7.4 3.1
##   Gls.90 G.PK.90 Ast.90 xG.90 xAG.90 Competition Sh Sh.90 SoT.90 Passes.
## 1   0.00    0.00   0.00  0.00   0.03  Bundesliga  0  0.00   0.00    57.1
## 2   0.00    0.00   0.00  0.00   0.00  Bundesliga  0  0.00   0.00    76.5
## 3   0.00    0.00   0.00  0.09   0.00  Bundesliga  6  1.45   0.00    81.6
## 4   0.38    0.38   0.19  0.27   0.17  Bundesliga 59  2.84   1.15    68.5
## 5   0.23    0.23   0.06  0.18   0.07  Bundesliga 39  2.20   0.73    80.1
## 6   0.52    0.47   0.19  0.35   0.15  Bundesliga 41  1.94   0.57    72.1
##   ShortPasses. MediumPasses. LongPasses. A.xAG TklW.90 Blocks.90 Int.90
## 1         70.0           0.0         0.0   0.0    0.43      0.29   0.00
## 2         90.0          66.7         0.0   0.0    0.00      0.00   0.00
## 3         91.0          91.4        40.0   0.0    0.62      0.38   0.25
## 4         73.8          75.0        54.8   0.5    0.37      0.48   0.15
## 5         83.3          87.3        72.4  -0.2    0.53      0.97   0.28
## 6         83.1          74.2        63.2   0.9    0.10      0.37   0.07
##   Tkl.Int.90 Clr.90 Touches.90 Dribbles.90 Dribbles. SCA.90 GCA.90 Aerial.
## 1       0.43   0.14       3.71        0.00       0.0   1.48   0.00     0.0
## 2       0.00   0.50       9.00        0.00       0.0   0.00   0.00     0.0
## 3       1.50   1.75      17.50        0.06      50.0   1.21   0.00    76.5
## 4       0.96   0.74      28.19        1.22      51.6   2.74   0.48    50.6
## 5       1.41   0.53      30.06        1.25      53.3   3.16   0.23    59.2
## 6       0.30   0.23      32.63        0.57      39.5   3.41   0.66    31.9
##   Points.90  xGA OG PSxG PSxG.SoT PSxG... GA GA.90 Save. CS. SoTA.90 SoTA.GA
## 1      0.43  1.4  0    0        0       0  0     0     0   0       0       0
## 2      3.00  0.8  0    0        0       0  0     0     0   0       0       0
## 3      0.88  9.0  0    0        0       0  0     0     0   0       0       0
## 4      0.85 34.4  0    0        0       0  0     0     0   0       0       0
## 5      0.91 29.2  0    0        0       0  0     0     0   0       0       0
## 6      1.17 35.7  0    0        0       0  0     0     0   0       0       0
##   SoT.G G.xG PassesCompleted.90 PassesAttempted.90 ShortPassesCompleted.90
## 1  0.00  0.0               1.14               2.00                    1.00
## 2  0.00  0.0               6.50               8.50                    4.50
## 3  0.00 -0.4              10.81              13.25                    4.44
## 4  3.00  2.4              12.89              18.81                    6.56
## 5  3.25  0.7              18.16              22.66                    8.56
## 6  1.20  3.6              19.10              26.50                   10.30
##   MediumPassesCompleted.90 LongPassesCompleted.90 TotDistPasses.90
## 1                     0.00                   0.00            10.00
## 2                     2.00                   0.00            89.00
## 3                     5.31                   0.75           183.44
## 4                     4.67                   0.85           198.19
## 5                     7.28                   1.72           309.41
## 6                     6.33                   1.60           303.93
##   PrgDistPasses.90 xA.90 KP.90 FinalThirdPasses.90 PPA.90 CrsPA.90
## 1             0.29  0.00  0.14                0.00   0.00     0.00
## 2            22.00  0.00  0.00                0.50   0.00     0.00
## 3            75.50  0.01  0.00                0.69   0.06     0.00
## 4            48.19  0.12  0.74                1.37   0.70     0.15
## 5            70.56  0.06  0.66                1.66   0.56     0.12
## 6           100.97  0.11  0.83                2.67   1.17     0.27
##   PassesProgressive.90 xGA.90 Recov.90 Fls.90 Fld.90 AerialW.90   xGD xGD.90
## 1                 0.00   0.20     0.57   0.29   0.43       0.00  -1.4  -0.20
## 2                 0.50   0.40     0.50   0.00   0.00       0.00  -0.8  -0.40
## 3                 0.88   0.56     0.81   0.31   0.19       0.81  -8.6  -0.47
## 4                 2.52   1.27     2.44   0.56   1.11       1.48 -28.8  -1.00
## 5                 2.16   0.91     4.09   0.50   0.44       0.91 -25.9  -0.73
## 6                 3.87   1.19     2.20   0.57   1.10       0.50 -28.3  -0.84
##   MP_Squad
## 1       34
## 2       34
## 3       34
## 4       34
## 5       34
## 6       34

list.files("data")

## [1] "FBREF_BigClubes_2324.csv"  "FBREF_BigClubes_2425.csv" 
## [3] "FBREF_BigPlayers_2324.csv" "FBREF_BigPlayers_2425.csv"

# Define the select_players function (complete)
select_players <- function(file, Encoding, position, competition, primary, player) {
  # 1. Read the file
  data <- read.csv(file, sep = ";", encoding = Encoding)
  
  # 2. Filter by competition
  if (length(competition) == 1 && competition == "ALL") {
    cat("We consider all players from all competitions\n")
    data_comp <- data
  } else {
    cat("We filter by competition\n")
    data_comp <- data %>%
      filter(Competition %in% competition)
  }
  
  # 3. Filter by position
  if (primary){
    cat("We keep players whose main position is:", position, "\n")
    data_players <- data_comp %>% 
      filter(substr(Pos, 1, 2) == position)
  } else {
    cat("We keep players whose position contains:", position, "\n")
    data_players <- data_comp %>%
      filter(grepl(position, Pos))
  }
  
  # 4. Filter by specific player if needed
  if (!is.na(player)){
    data_players <- data_players %>% filter(Player %in% player)
  }
  
  # 5. Return the filtered data
  return(data_players)
}

# NOW call the function with all the parameters
df_forwards <- select_players(
  file = "FBREF_BigPlayers_2425.csv",
  Encoding = "UTF-8",
  position = "FW", 
  competition = c("Premier League", "La Liga", "Serie A", "Bundesliga", "Ligue 1"),
  primary = TRUE,
  player = NA
)

## We filter by competition
## We keep players whose main position is: FW

cat("\nTotal Centre Forwards in Top 5 Leagues:", nrow(df_forwards), "\n")

## 
## Total Centre Forwards in Top 5 Leagues: 723

cat("(Excludes wingers and attacking midfielders)\n")

## (Excludes wingers and attacking midfielders)

head(df_forwards)

##                   Player         Squad Nation   Pos Age MP  Min Gls G.PK Ast
## 1        Abdoulie Ceesay     St. Pauli    GAM    FW  20  7   60   0    0   0
## 2            Adam Hložek    Hoffenheim    CZE FW,MF  22 27 1871   8    8   4
## 3 Alexander Bernhardsson Holstein Kiel    SWE FW,MF  25 19 1263   7    7   2
## 4         Andreas Albers     St. Pauli    DEN    FW  34 14   62   1    1   0
## 5     Andreas Skov Olsen     Wolfsburg    DEN FW,MF  24 12  535   1    1   1
## 6            Andrej Ilic  Union Berlin    SRB    FW  24 16  937   7    6   0
##    xG xAG Gls.90 G.PK.90 Ast.90 xG.90 xAG.90 Competition Sh Sh.90 SoT.90
## 1 0.0 0.0   0.00    0.00   0.00  0.00   0.03  Bundesliga  0  0.00   0.00
## 2 5.6 3.5   0.38    0.38   0.19  0.27   0.17  Bundesliga 59  2.84   1.15
## 3 4.1 1.3   0.50    0.50   0.14  0.29   0.10  Bundesliga 27  1.92   1.00
## 4 0.6 0.1   1.45    1.45   0.00  0.90   0.14  Bundesliga  3  4.35   1.45
## 5 0.8 1.9   0.17    0.17   0.17  0.14   0.32  Bundesliga  7  1.18   0.17
## 6 4.9 0.3   0.67    0.58   0.00  0.47   0.03  Bundesliga 27  2.59   1.06
##   Passes. ShortPasses. MediumPasses. LongPasses. A.xAG TklW.90 Blocks.90 Int.90
## 1    57.1         70.0           0.0         0.0   0.0    0.43      0.29   0.00
## 2    68.5         73.8          75.0        54.8   0.5    0.37      0.48   0.15
## 3    60.5         71.0          64.9        36.4   0.7    0.79      0.74   0.26
## 4    51.7         55.6          66.7        50.0  -0.1    0.07      0.00   0.00
## 5    80.4         85.7          87.6        55.0  -0.9    0.50      0.33   0.17
## 6    57.6         69.2          46.2        20.0  -0.3    0.12      0.44   0.31
##   Tkl.Int.90 Clr.90 Touches.90 Dribbles.90 Dribbles. SCA.90 GCA.90 Aerial.
## 1       0.43   0.14       3.71        0.00       0.0   1.48   0.00     0.0
## 2       0.96   0.74      28.19        1.22      51.6   2.74   0.48    50.6
## 3       1.79   1.32      27.37        1.16      41.5   2.57   0.50    35.8
## 4       0.07   0.07       3.14        0.14      66.7   2.95   0.00    45.5
## 5       0.83   1.00      27.42        0.33      28.6   5.38   0.17    50.0
## 6       0.44   0.38      18.88        0.06      14.3   1.44   0.19    48.1
##   Points.90  xGA OG PSxG PSxG.SoT PSxG... GA GA.90 Save. CS. SoTA.90 SoTA.GA
## 1      0.43  1.4  0    0        0       0  0     0     0   0       0       0
## 2      0.85 34.4  0    0        0       0  0     0     0   0       0       0
## 3      0.79 23.3  0    0        0       0  0     0     0   0       0       0
## 4      0.86  2.1  0    0        0       0  0     0     0   0       0       0
## 5      1.17  9.2  0    0        0       0  0     0     0   0       0       0
## 6      1.25 16.3  0    0        0       0  0     0     0   0       0       0
##   SoT.G G.xG PassesCompleted.90 PassesAttempted.90 ShortPassesCompleted.90
## 1  0.00  0.0               1.14               2.00                    1.00
## 2  3.00  2.4              12.89              18.81                    6.56
## 3  2.00  2.9              10.16              16.79                    4.89
## 4  1.00  0.4               1.07               2.07                    0.71
## 5  1.00  0.2              18.08              22.50                    8.50
## 6  1.83  2.1               6.88              11.94                    4.50
##   MediumPassesCompleted.90 LongPassesCompleted.90 TotDistPasses.90
## 1                     0.00                   0.00            10.00
## 2                     4.67                   0.85           198.19
## 3                     3.89                   0.84           166.53
## 4                     0.29                   0.07            16.64
## 5                     7.08                   1.83           306.42
## 6                     1.50                   0.06            76.44
##   PrgDistPasses.90 xA.90 KP.90 FinalThirdPasses.90 PPA.90 CrsPA.90
## 1             0.29  0.00  0.14                0.00   0.00     0.00
## 2            48.19  0.12  0.74                1.37   0.70     0.15
## 3            56.63  0.06  0.89                0.84   0.53     0.21
## 4             3.36  0.01  0.14                0.07   0.00     0.00
## 5            64.00  0.11  1.50                1.00   0.83     0.17
## 6            19.62  0.01  0.38                1.00   0.00     0.00
##   PassesProgressive.90 xGA.90 Recov.90 Fls.90 Fld.90 AerialW.90   xGD xGD.90
## 1                 0.00   0.20     0.57   0.29   0.43       0.00  -1.4  -0.20
## 2                 2.52   1.27     2.44   0.56   1.11       1.48 -28.8  -1.00
## 3                 1.74   1.23     2.37   1.11   0.79       1.00 -19.2  -0.94
## 4                 0.21   0.15     0.21   0.21   0.14       0.71  -1.5   0.75
## 5                 1.83   0.77     1.50   0.17   0.25       0.25  -8.4  -0.63
## 6                 0.81   1.02     1.00   0.88   0.50       3.25 -11.4  -0.55
##   MP_Squad
## 1       34
## 2       34
## 3       34
## 4       34
## 5       34
## 6       34

Next we filter by position, as we are not interested in defenders or midfielders in this study.

## We now go ahead and load only Centre Forwards out of Europes top 5 leagues by setting primary = True. This eliminates players who play both CF and MF roles, to focus solely on out and out 9s.

Next, we keep the players whose main position is FW. We want the analysis to stay flexible, so we can include wingers who can play in the front 3 and offer a dynamic goal threat.

cat("\nTotal Centre Forwards in Top 5 Leagues:", nrow(df_forwards), "\n")

## 
## Total Centre Forwards in Top 5 Leagues: 723

cat("Excludes attacking midfielders)\n")

## Excludes attacking midfielders)

head(df_forwards)

##                   Player         Squad Nation   Pos Age MP  Min Gls G.PK Ast
## 1        Abdoulie Ceesay     St. Pauli    GAM    FW  20  7   60   0    0   0
## 2            Adam Hložek    Hoffenheim    CZE FW,MF  22 27 1871   8    8   4
## 3 Alexander Bernhardsson Holstein Kiel    SWE FW,MF  25 19 1263   7    7   2
## 4         Andreas Albers     St. Pauli    DEN    FW  34 14   62   1    1   0
## 5     Andreas Skov Olsen     Wolfsburg    DEN FW,MF  24 12  535   1    1   1
## 6            Andrej Ilic  Union Berlin    SRB    FW  24 16  937   7    6   0
##    xG xAG Gls.90 G.PK.90 Ast.90 xG.90 xAG.90 Competition Sh Sh.90 SoT.90
## 1 0.0 0.0   0.00    0.00   0.00  0.00   0.03  Bundesliga  0  0.00   0.00
## 2 5.6 3.5   0.38    0.38   0.19  0.27   0.17  Bundesliga 59  2.84   1.15
## 3 4.1 1.3   0.50    0.50   0.14  0.29   0.10  Bundesliga 27  1.92   1.00
## 4 0.6 0.1   1.45    1.45   0.00  0.90   0.14  Bundesliga  3  4.35   1.45
## 5 0.8 1.9   0.17    0.17   0.17  0.14   0.32  Bundesliga  7  1.18   0.17
## 6 4.9 0.3   0.67    0.58   0.00  0.47   0.03  Bundesliga 27  2.59   1.06
##   Passes. ShortPasses. MediumPasses. LongPasses. A.xAG TklW.90 Blocks.90 Int.90
## 1    57.1         70.0           0.0         0.0   0.0    0.43      0.29   0.00
## 2    68.5         73.8          75.0        54.8   0.5    0.37      0.48   0.15
## 3    60.5         71.0          64.9        36.4   0.7    0.79      0.74   0.26
## 4    51.7         55.6          66.7        50.0  -0.1    0.07      0.00   0.00
## 5    80.4         85.7          87.6        55.0  -0.9    0.50      0.33   0.17
## 6    57.6         69.2          46.2        20.0  -0.3    0.12      0.44   0.31
##   Tkl.Int.90 Clr.90 Touches.90 Dribbles.90 Dribbles. SCA.90 GCA.90 Aerial.
## 1       0.43   0.14       3.71        0.00       0.0   1.48   0.00     0.0
## 2       0.96   0.74      28.19        1.22      51.6   2.74   0.48    50.6
## 3       1.79   1.32      27.37        1.16      41.5   2.57   0.50    35.8
## 4       0.07   0.07       3.14        0.14      66.7   2.95   0.00    45.5
## 5       0.83   1.00      27.42        0.33      28.6   5.38   0.17    50.0
## 6       0.44   0.38      18.88        0.06      14.3   1.44   0.19    48.1
##   Points.90  xGA OG PSxG PSxG.SoT PSxG... GA GA.90 Save. CS. SoTA.90 SoTA.GA
## 1      0.43  1.4  0    0        0       0  0     0     0   0       0       0
## 2      0.85 34.4  0    0        0       0  0     0     0   0       0       0
## 3      0.79 23.3  0    0        0       0  0     0     0   0       0       0
## 4      0.86  2.1  0    0        0       0  0     0     0   0       0       0
## 5      1.17  9.2  0    0        0       0  0     0     0   0       0       0
## 6      1.25 16.3  0    0        0       0  0     0     0   0       0       0
##   SoT.G G.xG PassesCompleted.90 PassesAttempted.90 ShortPassesCompleted.90
## 1  0.00  0.0               1.14               2.00                    1.00
## 2  3.00  2.4              12.89              18.81                    6.56
## 3  2.00  2.9              10.16              16.79                    4.89
## 4  1.00  0.4               1.07               2.07                    0.71
## 5  1.00  0.2              18.08              22.50                    8.50
## 6  1.83  2.1               6.88              11.94                    4.50
##   MediumPassesCompleted.90 LongPassesCompleted.90 TotDistPasses.90
## 1                     0.00                   0.00            10.00
## 2                     4.67                   0.85           198.19
## 3                     3.89                   0.84           166.53
## 4                     0.29                   0.07            16.64
## 5                     7.08                   1.83           306.42
## 6                     1.50                   0.06            76.44
##   PrgDistPasses.90 xA.90 KP.90 FinalThirdPasses.90 PPA.90 CrsPA.90
## 1             0.29  0.00  0.14                0.00   0.00     0.00
## 2            48.19  0.12  0.74                1.37   0.70     0.15
## 3            56.63  0.06  0.89                0.84   0.53     0.21
## 4             3.36  0.01  0.14                0.07   0.00     0.00
## 5            64.00  0.11  1.50                1.00   0.83     0.17
## 6            19.62  0.01  0.38                1.00   0.00     0.00
##   PassesProgressive.90 xGA.90 Recov.90 Fls.90 Fld.90 AerialW.90   xGD xGD.90
## 1                 0.00   0.20     0.57   0.29   0.43       0.00  -1.4  -0.20
## 2                 2.52   1.27     2.44   0.56   1.11       1.48 -28.8  -1.00
## 3                 1.74   1.23     2.37   1.11   0.79       1.00 -19.2  -0.94
## 4                 0.21   0.15     0.21   0.21   0.14       0.71  -1.5   0.75
## 5                 1.83   0.77     1.50   0.17   0.25       0.25  -8.4  -0.63
## 6                 0.81   1.02     1.00   0.88   0.50       3.25 -11.4  -0.55
##   MP_Squad
## 1       34
## 2       34
## 3       34
## 4       34
## 5       34
## 6       34

unique(df_forwards$Competition)

## [1] "Bundesliga"     "La Liga"        "Ligue 1"        "Premier League"
## [5] "Serie A"

cat("Position distribution:\n")

## Position distribution:

table(df_forwards$Pos)

## 
##    FW FW,DF FW,MF 
##   375    24   324

cat("\nNote: We should only see 'FW' - Players that play in the front 3 and not AMs\n")

## 
## Note: We should only see 'FW' - Players that play in the front 3 and not AMs

Our next step is to filter the dataset to focus on our target sample, based on playing time and age criteria.

filter_players <- function(data, metrics, pct_min_minutes, age_max, age_min = 0){
  # Filter the data and select the metrics that define our sample
  data_filter <- data %>%
    filter(
      Min > round((pct_min_minutes * 90 * MP_Squad) / 100),
      Age <= age_max,
      Age >= age_min
    )%>%
    select(c("Player", "Squad", "Age", "Competition", all_of(metrics)))
  rownames(data_filter) <- 1:nrow(data_filter)
  return(data_filter)
}

Arsenal are a team that try to suffocate opponents high up the field with controlled, possession based football. All of the current centre forwards have offered different traits in playing the role, but undoubtedly, Arteta will be looking for someone who is:

A goal Threat We will look at Non penalty goals, xG, and shot volume/efficiency.
Great Link Up Play We will look at shot creating actions, xA, and Assists.
Pressing/Work Ethic High Intensity pressing, tackles won, balls recovered.
Other Traits Dribbling, first touch, aerial duels.

# Key Metrics for Arsenals CF profile. 

list_metrics <-  c(
  "G.PK.90", # Non Pen goals per 90
  "xG.90", # xG per 90
  "Ast.90", # Assists per 90
  "xAG.90", # xA per 90
  "SCA.90", # Shot Creating Actions per 90
  "GCA.90", # Goal Creating actions per 90
  "TklW.90", # Tackles Won per 90
  "Recov.90", # Ball Recoveries per 90
  "Dribbles.90", # Successful dribbles per 90
  "Aerial.", # Aerial Duel percentage
  "Sh.90", # Shots per 90
  "SoT.90" # Shots on target per 90 
)

We will keep centre forwards who have played at least 50% of available minutes and who are below the age of 28.

df_forwards_filter <-  filter_players(
  data = df_forwards,
  metrics = list_metrics,
  pct_min_minutes = 50, 
  age_max = 28,
  age_min = 18
)

cat("Centre forwards meeting criteria:", nrow(df_forwards_filter), "\n")

## Centre forwards meeting criteria: 156

We can now filter to ensure that there are no duplicates in our data ser

duplicated_players <- df_forwards_filter[
  duplicated(df_forwards_filter$Player), ]$Player

if(length(duplicated_players) > 0) {
  cat("Duplicated players:", paste(duplicated_players, collapse = ", "), "\n")
}else {
  cat("No duplicated players found\n")
}

## No duplicated players found

To make our data easier to read, we will now rename the metrics.

df_forwards_rename <- df_forwards_filter %>%
  rename(
    'Non Penalty Goals/90' = 'G.PK.90',
    'Expected Goals/90' = 'xG.90',
    'Assists/90' = 'Ast.90',
    'Expected Assists/90' = 'xAG.90',
    'Shot Creating Actions/90' = 'SCA.90',
    'Goal Creating Actions/90' = 'GCA.90',
    'Tackles Won/90' = 'TklW.90',
    'Recoveries/90' = 'Recov.90',
    'Dribbles/90' = 'Dribbles.90',
    'Aerial Win %' = 'Aerial.',
    'Shots/90' = 'Sh.90',
    'Shots on Target/90' = 'SoT.90'
  )

head(df_forwards_rename
  )

##                Player          Squad Age Competition Non Penalty Goals/90
## 1         Adam Hložek     Hoffenheim  22  Bundesliga                 0.38
## 2 Benedict Hollerbach   Union Berlin  23  Bundesliga                 0.32
## 3      Benjamin Šeško     RB Leipzig  21  Bundesliga                 0.42
## 4         Deniz Undav      Stuttgart  28  Bundesliga                 0.47
## 5   Ermedin Demirović      Stuttgart  26  Bundesliga                 0.73
## 6        Hugo Ekitike Eint Frankfurt  22  Bundesliga                 0.49
##   Expected Goals/90 Assists/90 Expected Assists/90 Shot Creating Actions/90
## 1              0.27       0.19                0.17                     2.74
## 2              0.25       0.04                0.06                     2.80
## 3              0.38       0.19                0.08                     1.93
## 4              0.61       0.16                0.19                     3.02
## 5              0.69       0.05                0.09                     2.28
## 6              0.76       0.28                0.24                     3.55
##   Goal Creating Actions/90 Tackles Won/90 Recoveries/90 Dribbles/90
## 1                     0.48           0.37          2.44        1.22
## 2                     0.25           0.65          3.26        1.24
## 3                     0.34           0.09          1.88        1.18
## 4                     0.42           0.37          1.78        0.37
## 5                     0.34           0.21          0.97        0.15
## 6                     0.42           0.36          2.64        1.58
##   Aerial Win % Shots/90 Shots on Target/90
## 1         50.6     2.84               1.15
## 2         30.9     2.69               0.85
## 3         58.8     2.50               1.10
## 4         40.4     3.91               1.62
## 5         42.0     3.11               1.31
## 6         46.8     4.00               1.55

Now that we have our sample of players, and the relevent metrics that we are interested in, it is now time to created a weighting system that represents the needs of Arsenals number 9. One that reflects the CLubs priorities in the transfer market. We will weight this model slightly in favour of goalscoring ability, given the teams needs to find goals in that area.

scoring_calculate <- function(sample, metrics, weights){
  # Check that weights sum to 1
  if (sum(weights) != 1){
    stop("The sum of weights must be equal to 1")
  }
  
  # Normalize metrics (0-100 scale)
  data_scaled <- sample
  for (i in 1:length(metrics)){
    metric <- metrics[i]
    max_value <- max(sample[[metric]], na.rm = TRUE)
    min_value <- min(sample[[metric]], na.rm = TRUE)
    
    # Scale to 0-100
    data_scaled[[metric]] <- ((sample[[metric]] - min_value) / 
                               (max_value - min_value)) * 100
  }
  
  # Calculate weighted score
  data_scaled$Score <- 0
  for (i in 1:length(metrics)){
    data_scaled$Score <- data_scaled$Score + 
      (data_scaled[[metrics[i]]] * weights[i])
  }
  
  # Add ranking
  data_scaled <- data_scaled %>%
    arrange(desc(Score)) %>%
    mutate(Rank = row_number())
  
  return(data_scaled)
}

weights_arsenal_cf <- c(
  0.25, # Non penalty Goals/90 (Increased Weight - Main job is to score)
  0.18, # Expected Goals/90 (Increased Weight to show CF getting into positions)
  0.08, # Assists/90
  0.08, # Expected Assists/90
  0.10, # Shot Creating Actions/90
  0.06, # Goal Creating Actions/90
  0.08, # Tackles Won/90 (Pressing)
  0.04, # Recoveries/90
  0.04, # Dribbles/90
  0.06, # Aerial Win %
  0.02, # Shots/90
  0.01 # Shots on Target/90 
  )
# Verify weuights sum to 1
cat("Total Weight:", sum(weights_arsenal_cf), "\n\n")

## Total Weight: 1

# NExt, we calculate scores
data_final <- scoring_calculate(
  sample = df_forwards_filter,
  metrics = list_metrics,
  weights = weights_arsenal_cf
)

Now we can move on and take a look at the top performers up front so we can begin to create a shortlist of transfer options.

top_20 <- data_final %>%
  select(Rank, Player, Squad, Age, Competition, Score, 'G.PK.90', 'Aerial.') %>%
  head(20) %>%
  mutate(across(where(is.numeric), ~round(., 2)))
  kable(top_20, caption = "Top 20 Strikers for Arsenal")

Top 20 Strikers for Arsenal
Rank	Player	Squad	Age	Competition	Score	G.PK.90	Aerial.
1	Ousmane Dembélé	PSG	27	Ligue 1	75.61	97.20	59.44
2	Michael Olise	Bayern Munich	22	Bundesliga	63.79	42.99	71.03
3	Raphinha	Barcelona	27	La Liga	60.28	47.66	89.15
4	Bradley Barcola	PSG	21	Ligue 1	59.93	54.21	99.11
5	Kylian Mbappé	Real Madrid	25	La Liga	57.22	69.16	59.44
6	Leroy Sané	Bayern Munich	28	Bundesliga	55.20	56.07	79.94
7	Bukayo Saka	Arsenal	22	Premier League	54.38	24.30	49.48
8	Mateo Retegui	Atalanta	25	Serie A	54.00	73.83	63.74
9	Désiré Doué	PSG	19	Ligue 1	53.94	28.97	65.97
10	Hugo Ekitike	Eint Frankfurt	22	Bundesliga	53.39	45.79	69.54
11	Rayan Cherki	Lyon	20	Ligue 1	51.51	32.71	49.48
12	Patrik Schick	Leverkusen	28	Bundesliga	48.19	100.00	66.86
13	Mason Greenwood	Marseille	22	Ligue 1	47.73	42.06	79.20
14	Nick Woltemade	Stuttgart	22	Bundesliga	47.56	51.40	64.04
15	Vinicius Júnior	Real Madrid	24	La Liga	47.28	33.64	18.57
16	Luis Díaz	Liverpool	27	Premier League	46.90	45.79	37.89
17	Alexander Isak	Newcastle Utd	24	Premier League	46.00	57.94	47.70
18	Serhou Guirassy	Dortmund	28	Bundesliga	45.88	57.94	78.60
19	Riccardo Orsolini	Bologna	27	Serie A	44.82	54.21	81.43
20	Erling Haaland	Manchester City	24	Premier League	43.98	57.94	79.20

Tier Analysis

We will now separate our pool of talent into tiers, based on age, club, profile and realism of making the signing.

#Elite Clubs and rivals would indicate that the chances of signing are much lower. 
elite_clubs <- c(
  "Manchester City", "Real Madrid", "Barcelona", "Bayern Munich", "Tottenham", "Liverpool", "Chelsea", "Manchester United", "PSG"
)

#Tier 1: Elite Benchmarks (age <= 27, top scorers regardless of team)
tier1_elite <- data_final %>%
  filter(Age <= 27) %>%
  arrange(desc('G.PK.90')) %>%
  head(10) %>%
  mutate(Tier = "Tier1: Eliter Benchmark")

#Tier 2: Realistic Targets (Age 23-27), not at the elite clubs or rivals) - MAIN FOCUS AREA
tier2_realistic <-  data_final %>%
  filter(
    Age >= 23, Age <= 27,
    !Squad %in% elite_clubs
  ) %>%
  mutate(Tier = "Tier 2: Realistic Target")

#Tier 3A: Value Opportunities (Age 24-27, experienced players at smaller clubs)
tier3a_value <- data_final %>%
  filter(
    Age >= 24, Age <= 27,
    !Squad %in% elite_clubs
  ) %>%
  arrange(desc(Score)) %>%
  head(10) %>%
  mutate(Tier = "Tier 3A: Value Options")

#Tier 3B: Young Prospects (age 19-23, Pure Centre Forwards only as at a young age there is little data to suggest they can be a success in any front 3 position
tier3b_prospects <-  data_final %>%
  filter(Age >= 19, Age <= 23) %>%
  arrange(desc(Score)) %>%
  head(10) %>%
  mutate(Tier = "Tier 3B: Young Prospects")
cat("Tier 3B Young Prospects - These are Centre Forwards, not wingers:\n")

## Tier 3B Young Prospects - These are Centre Forwards, not wingers:

Tier 1 - Ranking the elite benchmarks of the worlds best strikers - These players are perhaps unattainable, but it is still useful information for comparison (and perhaps an unexpected opportunity will arise

tier1_display <- tier1_elite %>%
  select(Player, Squad, Age, `G.PK.90`, `xG.90`, `Aerial.`, Score) %>%
  mutate(across(where(is.numeric), ~round(., 2)))

kable(tier1_display, caption = "Tier 1: Elite Benchmark Centre Forwards")

Tier 1: Elite Benchmark Centre Forwards
Player	Squad	Age	G.PK.90	xG.90	Aerial.	Score
Ousmane Dembélé	PSG	27	97.20	100.00	59.44	75.61
Michael Olise	Bayern Munich	22	42.99	42.86	71.03	63.79
Raphinha	Barcelona	27	47.66	70.24	89.15	60.28
Bradley Barcola	PSG	21	54.21	63.10	99.11	59.93
Kylian Mbappé	Real Madrid	25	69.16	92.86	59.44	57.22
Bukayo Saka	Arsenal	22	24.30	40.48	49.48	54.38
Mateo Retegui	Atalanta	25	73.83	82.14	63.74	54.00
Désiré Doué	PSG	19	28.97	29.76	65.97	53.94
Hugo Ekitike	Eint Frankfurt	22	45.79	88.10	69.54	53.39
Rayan Cherki	Lyon	20	32.71	23.81	49.48	51.51

Of course, a lot of these players are probbaly quite unrealistic. These players (particularly the ones from Europes elite clubs) would command enormous transfer fees and their clubs will likely not be willing to let them go. That being said, Arsenal should remain alert to these players situations, as market opportunities can arise unexpectedly. If these players run into contract negotiation issues, or if any of these players feel ready for a new challenge, then Arsenal should be ready to strike as these represent some of the worlds best players who would guarantee goals in Arsenasl team.

After puting a shortlist together of our elite prospects, we now put together our top targets out of our most realistic options.

tier2_display <- tier2_realistic %>%
  select(Player, Squad, Age, Competition, 'G.PK.90', 'Ast.90', 'Aerial.', Score) %>%
  mutate(across(where(is.numeric), ~round(., 2)))

kable(tier2_display, caption = "Tier 2: Realistic Target Centre Forwards")

Tier 2: Realistic Target Centre Forwards
Player	Squad	Age	Competition	G.PK.90	Ast.90	Aerial.	Score
Mateo Retegui	Atalanta	25	Serie A	73.83	51.72	63.74	54.00
Alexander Isak	Newcastle Utd	24	Premier League	57.94	34.48	47.70	46.00
Riccardo Orsolini	Bologna	27	Serie A	54.21	32.76	81.43	44.82
Christian Pulisic	Milan	25	Serie A	27.10	56.90	30.91	43.80
Evann Guessand	Nice	23	Ligue 1	39.25	48.28	62.26	43.76
Ermedin Demirović	Stuttgart	26	Bundesliga	68.22	8.62	62.41	43.39
Marcus Thuram	Inter	26	Serie A	51.40	27.59	84.70	43.13
Julián Álvarez	Atlético Madrid	24	La Liga	43.93	24.14	43.83	42.77
Bryan Mbeumo	Brentford	24	Premier League	37.38	31.03	46.81	42.70
Rafael Leão	Milan	25	Serie A	28.97	53.45	83.95	42.64
Jonathan Burkardt	Mainz 05	24	Bundesliga	63.55	15.52	28.23	42.54
Moise Kean	Fiorentina	24	Serie A	56.07	17.24	76.52	41.59
Antoine Semenyo	Bournemouth	24	Premier League	28.97	24.14	69.09	40.98
Breel Embolo	Monaco	27	Ligue 1	27.10	34.48	74.29	40.66
Yoane Wissa	Brentford	27	Premier League	55.14	20.69	67.61	40.53
Ritsu Doan	Freiburg	26	Bundesliga	28.97	37.93	45.77	39.09
Jarrod Bowen	West Ham	27	Premier League	33.64	41.38	29.72	38.51
Lautaro Martínez	Inter	26	Serie A	39.25	18.97	71.17	38.37
Kaoru Mitoma	Brighton	27	Premier League	32.71	24.14	72.96	37.50
Harvey Barnes	Newcastle Utd	26	Premier League	42.99	36.21	42.50	37.43
Marcus Tavernier	Bournemouth	25	Premier League	13.08	39.66	65.53	37.41
Mohamed Amoura	Wolfsburg	24	Bundesliga	27.10	56.90	44.13	37.11
Valentín Castellanos	Lazio	25	Serie A	28.04	18.97	78.90	35.73
Jonathan David	Lille	24	Ligue 1	32.71	31.03	41.01	35.45
Dan Ndoye	Bologna	23	Serie A	23.36	29.31	59.44	34.63
Lassine Sinayoko	Auxerre	24	Ligue 1	14.95	53.45	44.13	34.38
Anthony Gordon	Newcastle Utd	23	Premier League	16.82	31.03	74.29	33.99
Kai Havertz	Arsenal	25	Premier League	40.19	24.14	66.42	33.87
Dušan Vlahović	Juventus	24	Serie A	28.04	34.48	71.47	33.80
Zuriko Davitashvili	Saint-Étienne	23	Ligue 1	21.50	44.83	34.18	33.79
Dodi Lukebakio	Sevilla	26	La Liga	27.10	10.34	68.65	33.37
Mathias Pereira Lage	Brest	27	Ligue 1	10.28	68.97	68.50	33.11
Jean-Philippe Mateta	Crystal Palace	27	Premier League	38.32	12.07	55.13	32.63
Mohammed Kudus	West Ham	23	Premier League	15.89	17.24	32.10	32.57
Morgan Guilavogui	St. Pauli	26	Bundesliga	28.04	17.24	69.54	32.55
Evanilson	Bournemouth	24	Premier League	36.45	6.90	60.03	32.07
Robin Hack	Gladbach	25	Bundesliga	15.89	50.00	61.37	31.69
Gabriel Martinelli	Arsenal	23	Premier League	28.97	27.59	45.17	31.58
Keito Nakamura	Reims	24	Ligue 1	34.58	12.07	49.48	31.34
Shuto Machino	Holstein Kiel	24	Bundesliga	38.32	15.52	58.10	31.11
Issa Soumaré	Le Havre	23	Ligue 1	22.43	32.76	89.15	30.27
Jonas Wind	Wolfsburg	25	Bundesliga	34.58	24.14	73.55	30.18
Gabriel Strefezza	Como	27	Serie A	19.63	24.14	43.68	29.73
Iliman Ndiaye	Everton	24	Premier League	24.30	0.00	35.66	29.69
Benedict Hollerbach	Union Berlin	23	Bundesliga	29.91	6.90	45.91	29.47
Loïs Openda	RB Leipzig	24	Bundesliga	30.84	31.03	56.46	29.28
Jørgen Strand Larsen	Wolves	24	Premier League	45.79	24.14	59.58	29.21
Marvin Pieringer	Heidenheim	24	Bundesliga	14.95	27.59	55.13	28.33
Nikola Krstović	Lecce	24	Serie A	24.30	25.86	61.81	28.23
Phillip Tietz	Augsburg	27	Bundesliga	29.91	18.97	71.17	27.61
Dennis Man	Parma	25	Serie A	16.82	31.03	51.41	27.60
Jorge de Frutos	Rayo Vallecano	27	La Liga	20.56	18.97	57.21	27.29
Artem Dovbyk	Roma	27	Serie A	34.58	12.07	61.22	27.02
Patrick Cutrone	Como	26	Serie A	28.97	31.03	38.34	26.61
Hugo Duro	Valencia	24	La Liga	39.25	13.79	54.83	26.44
Carlos Vicente	Alavés	25	La Liga	11.21	25.86	53.79	26.07
Farid El Melali	Angers	27	Ligue 1	9.35	24.14	32.99	25.97
Juan Cruz	Leganés	24	La Liga	18.69	27.59	65.53	25.44
Callum Hudson-Odoi	Nott’ham Forest	23	Premier League	19.63	13.79	34.32	25.14
Javi Puado	Espanyol	26	La Liga	19.63	20.69	34.32	24.96
Bryan Gil	Girona	23	La Liga	14.95	27.59	0.00	24.66
Esteban Lepaul	Angers	24	Ligue 1	46.73	0.00	63.45	24.66
Gustav Isaksen	Lazio	23	Serie A	14.95	13.79	52.15	24.54
Lorenzo Lucca	Udinese	23	Serie A	39.25	6.90	69.09	24.45
Lucas Beltrán	Fiorentina	23	Serie A	13.08	31.03	55.87	24.44
Nicolás González	Juventus	26	Serie A	14.02	17.24	82.76	24.28
Tete Morente	Lecce	27	Serie A	12.15	15.52	91.68	24.09
Samuel Essende	Augsburg	26	Bundesliga	36.45	18.97	65.08	24.07
Viktor Tsyhankov	Girona	26	La Liga	8.41	41.38	49.48	24.04
Mikel Oyarzabal	Real Sociedad	27	La Liga	18.69	20.69	45.32	23.93
Gorka Guruzeta	Athletic Club	27	La Liga	31.78	17.24	50.97	23.86
Oladapo Afolayan	St. Pauli	26	Bundesliga	14.95	8.62	44.13	23.65
Santiago Pierotti	Lecce	23	Serie A	17.76	15.52	66.12	23.43
Roberto Piccoli	Cagliari	23	Serie A	24.30	5.17	63.60	23.18
Isaac Romero	Sevilla	24	La Liga	15.89	13.79	53.94	21.44
Andrea Pinamonti	Genoa	25	Serie A	29.91	5.17	63.60	20.91
Dany Mota	Monza	26	Serie A	19.63	15.52	68.05	20.61
Junior Adamu	Freiburg	23	Bundesliga	11.21	20.69	48.89	20.32
Amin Sarr	Hellas Verona	23	Serie A	17.76	8.62	61.96	18.86
Johannes Eggestein	St. Pauli	26	Bundesliga	9.35	32.76	36.85	17.85
Jack Harrison	Everton	27	Premier League	3.74	0.00	18.57	16.65
Miguel	Leganés	24	La Liga	0.00	27.59	34.47	14.02
Alessandro Zanoli	Genoa	23	Serie A	4.67	0.00	74.29	13.32

##Tier 3A

This tier looks at strikers at smaller clubs who offer an immediate impact but at a more affordable cost.

tier3a_display <- tier3a_value %>%
  select(Player, Squad, Age, Competition, 'G.PK.90', Score) %>%
  mutate(Score = round(Score, 2), 'G.PK.90' = round(`G.PK.90`, 2))

kable(tier3a_display, caption = "Tier 3A: Value Option Strikers")

Tier 3A: Value Option Strikers
Player	Squad	Age	Competition	G.PK.90	Score
Mateo Retegui	Atalanta	25	Serie A	73.83	54.00
Alexander Isak	Newcastle Utd	24	Premier League	57.94	46.00
Riccardo Orsolini	Bologna	27	Serie A	54.21	44.82
Christian Pulisic	Milan	25	Serie A	27.10	43.80
Ermedin Demirović	Stuttgart	26	Bundesliga	68.22	43.39
Marcus Thuram	Inter	26	Serie A	51.40	43.13
Julián Álvarez	Atlético Madrid	24	La Liga	43.93	42.77
Bryan Mbeumo	Brentford	24	Premier League	37.38	42.70
Rafael Leão	Milan	25	Serie A	28.97	42.64
Jonathan Burkardt	Mainz 05	24	Bundesliga	63.55	42.54

Tier 3B - Young Prospects

For the purpose of a detailed pipeline of talent, it is also good to look at younger options, ones with large upsides of potential. If none of our ‘prime’ targets are available, then it is shrewd business to assess the market for younger, lesser known options.

tier3b_display <- tier3b_prospects %>%
  select(Player, Squad, Age, Competition, `G.PK.90`, `Aerial.`, Score) %>%
  mutate(across(where(is.numeric), ~round(., 2)))

kable(tier3b_display, caption = "Tier 3B: Young Prospects")

Tier 3B: Young Prospects
Player	Squad	Age	Competition	G.PK.90	Aerial.	Score
Michael Olise	Bayern Munich	22	Bundesliga	42.99	71.03	63.79
Bradley Barcola	PSG	21	Ligue 1	54.21	99.11	59.93
Bukayo Saka	Arsenal	22	Premier League	24.30	49.48	54.38
Désiré Doué	PSG	19	Ligue 1	28.97	65.97	53.94
Hugo Ekitike	Eint Frankfurt	22	Bundesliga	45.79	69.54	53.39
Rayan Cherki	Lyon	20	Ligue 1	32.71	49.48	51.51
Mason Greenwood	Marseille	22	Ligue 1	42.06	79.20	47.73
Nick Woltemade	Stuttgart	22	Bundesliga	51.40	64.04	47.56
Evann Guessand	Nice	23	Ligue 1	39.25	62.26	43.76
Maghnes Akliouche	Monaco	22	Ligue 1	14.02	26.60	43.13

Similiarity Algorithms

We can now use similiarity algorithms to find the strikers who match Arsenals desired profile.

We are going to use the top scoring player in Tier 2 as the reference point for comparison

similiarity_tool <- function(sample, player, metrics, metrics_rename, distance, n){
  data_scaled <- sample
  
  # Scale each metric
  for (metric in metrics) {
    max_value <- max(sample[[metric]], na.rm = TRUE)
    min_value <- min(sample[[metric]], na.rm = TRUE)
    data_scaled[[metric]] <- (sample[[metric]] - min_value) / (max_value - min_value)
  }
  
  # Select only metrics for distance calculation
  data_for_dist <- data_scaled[, metrics]
  rownames(data_for_dist) <- sample$Player
  
  # Calculate distance matrix
  if(distance == "euclidean"){
    mat_dist <- as.matrix(dist(data_for_dist, method = "euclidean"))
  } else if(distance == "cosine"){
    mat_dist <- as.matrix(1 - cosine(t(as.matrix(data_for_dist))))
  } else {
    stop("Distance method must be 'euclidean' or 'cosine'")
  }
  
  # Extract the similarity for our target player
  if(!(player %in% rownames(mat_dist))){
    stop(paste("Player", player, "not found in sample"))
  }
  
  player_sim <- mat_dist[, player]
  df_sim <- data.frame(
    Player = names(player_sim),
    Distance = as.numeric(player_sim)
  )
  
  # Drop the Player Themselves (distance = 0)
  df_sim <- df_sim[df_sim$Player != player, ]
  
  # Convert the distances to similarity percentage
  d95 <- quantile(df_sim$Distance, 0.95, na.rm = TRUE)
  df_sim$Similarity <- (1 - (df_sim$Distance / d95)) * 100
  
  # Order by distance (most similar first)
  df_sim <- df_sim[order(df_sim$Distance), ]
  
  # Take top n
  final_df <- df_sim[1:n, c("Player", "Similarity")]
  
  # Merge with original data for context
  data_clean <- sample %>%
    select(Player, Age, Squad, Competition, all_of(metrics))
  
  # Rename the metric columns
  colnames(data_clean)[colnames(data_clean) %in% metrics] <- metrics_rename
  
  final_df <- merge(
    x = final_df, y = data_clean, 
    by = "Player", all.x = TRUE
  )
  
  final_df <- final_df[order(-final_df$Similarity), ]
  rownames(final_df) <- 1:n
  return(final_df)
}

It is very easy to see why Ousmane Dembele is the current Ballon D’or holder. After a historic season of firing PSG to their first ever Champions League and treble, Dembele ranks highest in our Tier 1 forwards. Stylistically, physically, output and age would all be perfectly aligned for Arsenal if they were to be able to sign a player of Dembeles quality. This is unlikely, therefore we will use him as a benchmark when targeting other, ore realistic elite forwards.

metrics_rename <- c(
  "npg/90", "xG/90", "Ast/90", "xAG/90", "SCA/90", "GCA/90", "tklw/90", "Recov/90", "Drib/90", "Aerial%", "Sh/90", "SoT/90"
)
  
  # Use the top ranked striker (Dembele) as the gold standard benchmark
  
reference_player <- tier1_elite$Player[1]
  
  cat("Using", reference_player, "as reference player\n")

## Using Ousmane Dembélé as reference player

cat("(Elite Tier 1 - The Gold Standard\n")

## (Elite Tier 1 - The Gold Standard

cat("\nFinding realistic targets who play most similiarly to Dembele\n\n")

## 
## Finding realistic targets who play most similiarly to Dembele

# Euclidean Distance (Similiar Absolute Output)
sim_euclidean <- similiarity_tool(
  sample = data_final, 
  player = reference_player,
  metrics = list_metrics,
  metrics_rename = metrics_rename,
  distance = "euclidean",
  n = 15
)

kable(sim_euclidean[, 1:6],
      caption = paste("Top 15 Similiar Centre Forwards to", reference_player, "(Euclidean Distance)"), digits = 2)

Top 15 Similiar Centre Forwards to Ousmane Dembélé (Euclidean Distance)
Player	Similarity	Age	Squad	Competition	npg/90
Kylian Mbappé	60.22	25	Real Madrid	La Liga	69.16
Leroy Sané	52.94	28	Bayern Munich	Bundesliga	56.07
Raphinha	51.53	27	Barcelona	La Liga	47.66
Hugo Ekitike	51.41	22	Eint Frankfurt	Bundesliga	45.79
Bradley Barcola	50.89	21	PSG	Ligue 1	54.21
Mateo Retegui	46.11	25	Atalanta	Serie A	73.83
Nick Woltemade	45.54	22	Stuttgart	Bundesliga	51.40
Alexander Isak	43.29	24	Newcastle Utd	Premier League	57.94
Deniz Undav	42.80	28	Stuttgart	Bundesliga	43.93
Mason Greenwood	42.25	22	Marseille	Ligue 1	42.06
Michael Olise	38.85	22	Bayern Munich	Bundesliga	42.99
Julián Álvarez	38.10	24	Atlético Madrid	La Liga	43.93
Erling Haaland	36.61	24	Manchester City	Premier League	57.94
Riccardo Orsolini	36.57	27	Bologna	Serie A	54.21
Rafael Leão	36.12	25	Milan	Serie A	28.97

Next, we can take a look at the cosine similiarity comparisons, which measures similiarity between tactical profiles, regardless of total output volume (matches strengths with strengths and weaknesses with weaknesses)

# Cosine Distance (similiar style/profile)

sim_cosine <- similiarity_tool(
  sample = data_final,
  player = reference_player,
  metrics = list_metrics, 
  metrics_rename = metrics_rename,
  distance = "cosine",
  n = 15
)

kable(sim_cosine[, 1:6],
      caption = paste("Top 15 similiar Strikers to", reference_player, "(Cosine Similiarity)"),
      digits = 2)

Top 15 similiar Strikers to Ousmane Dembélé (Cosine Similiarity)
Player	Similarity	Age	Squad	Competition	npg/90
Harvey Barnes	91.47	26	Newcastle Utd	Premier League	42.99
Alexander Isak	89.64	24	Newcastle Utd	Premier League	57.94
Nick Woltemade	89.24	22	Stuttgart	Bundesliga	51.40
Kylian Mbappé	87.46	25	Real Madrid	La Liga	69.16
Julián Álvarez	87.24	24	Atlético Madrid	La Liga	43.93
Leroy Sané	85.23	28	Bayern Munich	Bundesliga	56.07
Deniz Undav	84.97	28	Stuttgart	Bundesliga	43.93
Nicolas Jackson	84.96	23	Chelsea	Premier League	38.32
Hugo Ekitike	84.10	22	Eint Frankfurt	Bundesliga	45.79
Mateo Retegui	82.68	25	Atalanta	Serie A	73.83
Jonathan David	81.58	24	Lille	Ligue 1	32.71
Loïs Openda	80.25	24	RB Leipzig	Bundesliga	30.84
Raphinha	80.03	27	Barcelona	La Liga	47.66
Shuto Machino	79.63	24	Holstein Kiel	Bundesliga	38.32
Bradley Barcola	79.24	21	PSG	Ligue 1	54.21

From this, we can generate a side by side view of the top 10 players based on both cosine and euclidean similiarities to Dembele.

comparison <- data_frame(
  Rank = 1:10,
  `Euclidean (Output)` = sim_euclidean$Player[1:10],
  `Cosine (Style)` = sim_cosine$Player[1:10]
)

kable(comparison, caption = "Euclidean vs Cosine Top 10 Comparison")

Euclidean vs Cosine Top 10 Comparison
Rank	Euclidean (Output)	Cosine (Style)
1	Kylian Mbappé	Harvey Barnes
2	Leroy Sané	Alexander Isak
3	Raphinha	Nick Woltemade
4	Hugo Ekitike	Kylian Mbappé
5	Bradley Barcola	Julián Álvarez
6	Mateo Retegui	Leroy Sané
7	Nick Woltemade	Deniz Undav
8	Alexander Isak	Nicolas Jackson
9	Deniz Undav	Hugo Ekitike
10	Mason Greenwood	Mateo Retegui

We can now visualize our top targets to compare profiles. Based on our scoring and similiarity analysis, we select our most interesting targets for a deeper dive. From this shortlist, we can use our football knowledge to assess current situations to decide upon the final shortlist.

Mbappe and Raphinha are completely untouchable by their clubs barring a dramatic change in availability, it is impossible to sign them.
Isak, Ekitike and Woltemade have all recently signed for Liverpool and Newcastle respectively.
Barnes, while clearly impressive output has mainly produced his best attacking output when Newcastle have been hit with many injuries, himself struggling for form and fitness at times.

# Use the absolute Benchmark (Dembele) as our gold standard
elite_reference <- tier1_elite$Player[1]

# Top Scorer from Tier 2 (Best Realistic Target)
top_tier2 <- tier2_realistic$Player[1]

# Most Similiar Player (Euclidean). Of course, we assume that Mbappe is completely untouchable by Real Madrid, so we default to option 2

most_similar_eucl <- sim_euclidean$Player [2]

# Most similiar player (cosine - ensuring a different result from euclidean for more in depth comparison). We also want to exclude Barnes, Isak, Woltemade and Mbappe for the reasons explained above
cosine_candidates <- sim_cosine$Player[!sim_cosine$Player %in% most_similar_eucl]
most_similar_cos <- cosine_candidates[5]

#Top Young Prospect
top_prospect <- tier3b_prospects$Player[1]

# Combine these players for radar (include elite benchmark and realistic options)
players_radar <- unique(c(elite_reference, top_tier2, most_similar_eucl, most_similar_cos, top_prospect))

cat("Centre Forwards selected for radar comparison:\n\n")

## Centre Forwards selected for radar comparison:

cat("  1.", elite_reference, "(Tier 1 - World's Best)\n\n")

##   1. Ousmane Dembélé (Tier 1 - World's Best)

cat("REALISTIC ALTERNATIVES:\n")

## REALISTIC ALTERNATIVES:

for(i in 2:length(players_radar)){
cat(" ", i, ".", players_radar[i], "\n")
}

##   2 . Mateo Retegui 
##   3 . Leroy Sané 
##   4 . Julián Álvarez 
##   5 . Michael Olise

So there we have it; a 5 man shortlist of fantastic strikers for Arsenal to consider; Dembele (Who of course we know would be perfect but unrealistic), Retegui, Sane, Alvarez and Olise.

Now we calculate the 5th and 95th percentile for each metric to create the radar boundaries.

min_max_df <- rbind(
  apply(data_final[, list_metrics], 2, 
        function(x) quantile(x, probs = 0.95, na.rm = TRUE)),
  apply(data_final[, list_metrics], 2,
        function(x) quantile(x, probs = 0.05, na.rm = TRUE))
)
rownames(min_max_df) <- c("p95", "p5")
min_max_df

##       G.PK.90    xG.90   Ast.90    xAG.90    SCA.90    GCA.90   TklW.90
## p95 57.943925 77.67857 62.50000 64.673913 75.188324 57.024793 71.487603
## p5   9.345794 13.09524  3.87931  6.521739  5.178908  8.264463  9.297521
##      Recov.90 Dribbles.90  Aerial.    Sh.90   SoT.90
## p95 67.533937   70.421245 83.06092 70.29478 58.36820
## p5   7.522624    3.205128 28.04606 16.66667 10.46025

Now we can prepare our radar charts

# Filter Data for Selected Players
df_forwards_radar <- data_final[data_final$Player %in% players_radar, ]

# Ensure values are within [p5, p95] boundaries

for (metric in list_metrics) {
  for (p in players_radar) {
    value_c <- df_forwards_radar[df_forwards_radar$Player == p, metric]
    
    if(length(value_c) > 0 && !is.na(value_c)){
      if(value_c < min_max_df["p5", metric]){
        df_forwards_radar[df_forwards_radar$Player == p, metric] <- min_max_df["p5", metric]
      } else if (value_c > min_max_df["p95", metric]){
        df_forwards_radar[df_forwards_radar$Player == p, metric] <- min_max_df["p95", metric]
      }
    }
  }
}

#Create final radar dataframe

df_forwards_radar <- as.data.frame(df_forwards_radar)
rownames(df_forwards_radar) <- df_forwards_radar$Player
         df_final_plot <-  rbind(
           min_max_df, df_forwards_radar[, list_metrics]
         )
         df_final_plot

##                   G.PK.90    xG.90   Ast.90    xAG.90    SCA.90    GCA.90
## p95             57.943925 77.67857 62.50000 64.673913 75.188324 57.024793
## p5               9.345794 13.09524  3.87931  6.521739  5.178908  8.264463
## Ousmane Dembélé 57.943925 77.67857 53.44828 64.673913 75.188324 57.024793
## Michael Olise   42.990654 42.85714 62.50000 64.673913 75.188324 57.024793
## Leroy Sané      56.074766 64.28571 46.55172 58.695652 56.120527 37.190083
## Mateo Retegui   57.943925 77.67857 51.72414 32.608696 31.450094 40.495868
## Julián Álvarez  43.925234 58.33333 24.13793 34.782609 50.282486 41.322314
##                   TklW.90  Recov.90 Dribbles.90  Aerial.    Sh.90   SoT.90
## p95             71.487603 67.533937   70.421245 83.06092 70.29478 58.36820
## p5               9.297521  7.522624    3.205128 28.04606 16.66667 10.46025
## Ousmane Dembélé 11.570248 12.443439   47.619048 59.43536 70.29478 58.36820
## Michael Olise   53.719008 52.036199   70.421245 71.02526 62.13152 51.46444
## Leroy Sané      38.842975 43.891403   42.124542 79.94056 70.29478 58.36820
## Mateo Retegui   18.181818 25.565611    6.593407 63.74443 70.29478 43.93305
## Julián Álvarez  35.537190 28.054299   28.205128 43.83358 43.76417 47.28033

We are now ready to create our radar chart

# Define radar chart function
create_radarchart <- function(data, color = color,
                              vlabels = colnames(data), vlcex = 0.7,
                              caxislabels = NULL, title = NULL){
  radarchart(
    data, axistype = 1,
    
    # Polygon
    pcol = color, pfcol = scales::alpha(color, 0.5),
    plwd = 2, plty = 1,
    cglcol = "grey", cglty = 1, cglwd = 8,
    
    # Axis
    axislabcol = "grey30",
    
    #Labels
    vlcex = vlcex, vlabels = vlabels,
    caxislabels = caxislabels, title = title
  )
}
# Metric Names for radar chart
metrics_name_plot <- c(
  "npG/90", "xG/90", "Ast/90", "xAG/90", "SCA/90", "GCA/90", "tklW/90", "Recov/90", "Drib/90", "Aerial%", "Sh/90", "SoT/90"
)

# Colors

colors_radar <- c("#EF0107", "#00AFBB", "#f7d62d", "#8DBF8D", "purple4")[1:(nrow(df_final_plot) - 2)]


# Create Radar chart

op <- par(mar = c(1, 2, 2, 2))
create_radarchart(
  data = df_final_plot,
  color = colors_radar,
  vlabels = metrics_name_plot
)

legend("bottomleft",
       legend = rownames(df_final_plot[-c(1,2), ]),
       horiz = FALSE,
       bty = 'n', pch = 20,
       col = colors_radar,
       text.col = "black", cex = 0.7, pt.cex = 2)

title(
  main = "Arsenal Striker Search\nComparing Realistic Targets to the worlds best (2024/25)",
  cex.main = 1.1, col.main = "#5D6D7E" 
)

Conclusions and Recommendations

Based on our analysis, I have offered further context to the search for Arsenals next forward, and made recommendations based not just on data, but other mitigating football circumstances.

The Benchmark - Arsenal would love to be able to sign Ousmane Dembele, or his French compatriot Kylian Mbappe. However these players are currently unattainable and would cost astronomical money.

**Realistic Alternatives* - Julian Alvarez and Mateo Retegui represent excellent players, who would fit Arsenals style, intensity and system excellently. While both players would cost more then 60m Euros, they represent excellent market opportunities that are more realistic. Out of these 2, I would recommend Arsenal to sign Julian Alvarez. While his metrics are slightly lower then Reteguis, he is slightly younger and based on his previous successful stint in England (with Man City), he is sure to hit the ground running and could be an instant impacter for Arsenals recruitment team. While his metrics are impressive and well rounded, there is also lots of upside to improve the defensive aspect of his game, as well as his creativity, which I am confident under Arteta, and playing with other world class players like Saka, Odegaard etc, this would be something Alvarez can surely live up to .

Other Options Leroy Sane represents an excellent tactical fit for Arsenal, as well as being Premier League proven and has also previusly worked with Arteta, however at 29, he does not represent a huge amount of growth potential. Michael Olise and Bradley Barcola would be other excellent young options to take a look at. Both would fit Arsenals style very well, have massive upside potential however it would take a lot of money to prize either player from their clubs.

Limitations

Age of Data I only had data from the 24/25 season available to me for this study. Therefore, players who moved in the summer transfer window (Isak, Ekitike) are no longer available on the market. It also does not take into account current form in the 2025/26 season.
League Difficulty Goals in the Premier League may be more valuable then goals in a slower league, such as Serie A. This is not taken into account in the analysis, hence my preference for Alvarez or Retegui, who has already proven his ability in the Premier League.
Team Quality Our analysis does not account for players who may have inflated statistics due to playing with better attacking teams/players.
Injury Status While I tried to account for injury prone players, and included a min number of minutes played, this is not fully captured in my analysis.

Future Possibilities - To enhance my work in the future, these are areas that I can improve upon my work.

Transfer Value This can be implemented to take a deep dive into the players representing the best value in the market
Trajectory Is the player trending upwords or have they reached their peak?
Big Games Do some of these players have inflated stats in smaller games? Who are the best big game performers?
Injury History Assess Physical Robustness across our talent pool.

Final Recommendation

In my opinion, Arsenal should already be doing their due dilligence on if a deal for Julian Alvarez is possible this summer. His playing style, age, robustness and proven premier league ability make him a perfect match for Arsenal and by signing him, Arsenal could truly take a step towards becoming the best team in Europe.

sessionInfo()

## R version 4.5.2 (2025-10-31 ucrt)
## Platform: x86_64-w64-mingw32/x64
## Running under: Windows 11 x64 (build 26200)
## 
## Matrix products: default
##   LAPACK version 3.12.1
## 
## locale:
## [1] LC_COLLATE=English_Canada.utf8  LC_CTYPE=English_Canada.utf8   
## [3] LC_MONETARY=English_Canada.utf8 LC_NUMERIC=C                   
## [5] LC_TIME=English_Canada.utf8    
## 
## time zone: America/Vancouver
## tzcode source: internal
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] scales_1.4.0    DT_0.34.0       knitr_1.51      lsa_0.73.4     
##  [5] SnowballC_0.7.1 fmsb_0.7.6      lubridate_1.9.4 forcats_1.0.1  
##  [9] stringr_1.6.0   dplyr_1.1.4     purrr_1.2.1     readr_2.1.6    
## [13] tidyr_1.3.2     tibble_3.3.1    ggplot2_4.0.1   tidyverse_2.0.0
## 
## loaded via a namespace (and not attached):
##  [1] gtable_0.3.6       jsonlite_2.0.0     compiler_4.5.2     tidyselect_1.2.1  
##  [5] jquerylib_0.1.4    yaml_2.3.12        fastmap_1.2.0      R6_2.6.1          
##  [9] generics_0.1.4     htmlwidgets_1.6.4  bslib_0.9.0        pillar_1.11.1     
## [13] RColorBrewer_1.1-3 tzdb_0.5.0         rlang_1.1.7        stringi_1.8.7     
## [17] cachem_1.1.0       xfun_0.55          sass_0.4.10        S7_0.2.1          
## [21] otel_0.2.0         timechange_0.3.0   cli_3.6.5          withr_3.0.2       
## [25] magrittr_2.0.4     digest_0.6.39      grid_4.5.2         rstudioapi_0.18.0 
## [29] hms_1.1.4          lifecycle_1.0.5    vctrs_0.7.0        evaluate_1.0.5    
## [33] glue_1.8.0         farver_2.1.2       rmarkdown_2.30     tools_4.5.2       
## [37] pkgconfig_2.0.3    htmltools_0.5.9

Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.

M9 - Individual Activity

Adam Walsh

2026-01-31