Loading in the tidyverse and data
# Loading tidyverse
library(tidyverse)
#Loading in Data
nhl_draft <- read_csv("nhldraft.csv")In this Data Dive we are looking at the probabilities behind different groups of variables. The groups that I decided to look at in this report each have summary statistics of other variables to go with them. Thus this should be able to give us a view of the data based on which groups tend to appear more or less when randomly picking a row.
In the code below, I grouped the data by nationality and included columns that summarize: The number of values per country, the means for age, games, played, goals, assists, points, plus_minus (how well they do on the field for the team), and their penalty minutes. This will give us an idea on why the means for all these values are high or low based on their given counts. I also calculated the amount of rows each nationality takes up in the data and divided it by the total number of rows in the data to get the probability of each group being picked randomly. Using the probability variable I was also able to create a “is_anomaly” variable that takes the lowest anomaly and classifies it as an “Anomaly”.
NOTE: For the calculations of the means, all of the “NA”s were removed and the means were rounded to 2 decimals.
To start, let’s look at player’s and their stats by Nationality.
nationality_summary <- nhl_draft |>
group_by(nationality) |>
summarise(count = n(),
avg_age = round(mean(age, na.rm = TRUE), 2),
avg_games_played = round(mean(games_played, na.rm = TRUE), 2),
avg_goals = round(mean(goals, na.rm = TRUE), 2),
avg_assists = round(mean(assists, na.rm = TRUE), 2),
avg_points = round(mean(points, na.rm = TRUE), 2),
avg_plus_minus = round(mean(plus_minus, na.rm = TRUE), 2),
avg_penalty_minutes = round(mean(penalties_minutes, na.rm = TRUE), 2)
) |>
mutate(prob = count/NROW(nhl_draft)) |>
relocate(prob, .after = count) |>
mutate(is_anomaly = case_when(
prob == min(prob) ~ "Anomaly"
)) |>
relocate(is_anomaly, .after = prob)
nationality_summary |>
print(n = 40)## # A tibble: 47 × 11
## nationality count prob is_anomaly avg_age avg_games_played avg_goals
## <chr> <int> <dbl> <chr> <dbl> <dbl> <dbl>
## 1 AT 19 0.00155 <NA> 19.5 322. 88.6
## 2 AU 1 0.0000816 Anomaly 18 24 2
## 3 BE 2 0.000163 <NA> 18 2 0
## 4 BN 1 0.0000816 Anomaly 19 951 55
## 5 BR 2 0.000163 <NA> 18 546. 18
## 6 BS 2 0.000163 <NA> 18 31 0
## 7 BY 32 0.00261 <NA> 18.6 223. 30.1
## 8 CA 6498 0.530 <NA> 18.6 319. 53.7
## 9 CH 73 0.00596 <NA> 18.9 203. 30
## 10 CN 1 0.0000816 Anomaly 18 NaN NaN
## 11 CZ 479 0.0391 <NA> 19.1 314. 52.8
## 12 DE 81 0.00661 <NA> 18.4 319. 54.7
## 13 DK 27 0.00220 <NA> 18.4 335. 53.9
## 14 EE 2 0.000163 <NA> 19 491 63
## 15 FI 497 0.0406 <NA> 19.1 233. 37.2
## 16 FR 9 0.000735 <NA> 20 188. 50.3
## 17 GB 22 0.00180 <NA> 18.6 296. 61.4
## 18 HT 1 0.0000816 Anomaly 19 89 21
## 19 HU 3 0.000245 <NA> 18 NaN NaN
## 20 IT 2 0.000163 <NA> 18 276. 10
## 21 JM 1 0.0000816 Anomaly 20 NaN NaN
## 22 JP 3 0.000245 <NA> 19.5 18.5 0.5
## 23 KR 2 0.000163 <NA> 18 478. 53.5
## 24 KZ 24 0.00196 <NA> 18.5 244. 27.6
## 25 LT 3 0.000245 <NA> 18.3 723 85
## 26 LV 39 0.00318 <NA> 19.9 214. 23.0
## 27 ME 1 0.0000816 Anomaly 18 NaN NaN
## 28 NG 2 0.000163 <NA> 18.5 38 2
## 29 NL 1 0.0000816 Anomaly 18 202 46
## 30 NO 23 0.00188 <NA> 18.6 88.6 6.8
## 31 PL 8 0.000653 <NA> 18.8 463 82.8
## 32 PY 1 0.0000816 Anomaly 20 834 222
## 33 RU 724 0.0591 <NA> 19.1 266. 53.3
## 34 SE 800 0.0653 <NA> 18.8 312. 50.0
## 35 SI 9 0.000735 <NA> 18.1 628 184.
## 36 SK 166 0.0136 <NA> 18.7 289. 58.2
## 37 SU 2 0.000163 <NA> 27.5 281 18
## 38 TH 1 0.0000816 Anomaly 18 NaN NaN
## 39 TW 1 0.0000816 Anomaly 20 994 51
## 40 TZ 1 0.0000816 Anomaly 18 52 6
## # ℹ 7 more rows
## # ℹ 4 more variables: avg_assists <dbl>, avg_points <dbl>,
## # avg_plus_minus <dbl>, avg_penalty_minutes <dbl>When looking at the data we can see that the counts for each Nationality ranges from 1 to 6498 counts. There are also people who are listed with no country. For the probabilities of each country we need to look at why the draft included international players.
Back in 1979, the rules of the draft were changed to allow players who had previously played professional hockey to be drafted. During that time there were 2 leagues: The NHL and the World Hockey Association. The World Hockey Association, however, dissolved because the NHL decided to merge 6 WHA teams (the Edmonton Oilers, New England Whalers, Quebec Nordiques, Cincinnati Stingers, Houston Aeros, and Winnipeg Jets) into the league and the merger was approved. A lot of international players had the opportunity to play during that time but most of the teams and players came from Canada. Thus this is why Canada has so many players. As the years went on, hockey became a more popular sport and the reputation of the NHL became well known across the world. Now the NHL has a variety of players from various countries but there’s only a small amount of those players due to location and climate that hockey plays in. This is why Canada and the United States are the 2 major country groups due to the popularity in those countries during the fall and winter.
The small count of some countries means that only 1 player played in the league from that country and thus their mean is skewed to only that sample population. Canada has the largest sample size of any country and thus is able to have the highest probability of any group and the most accurate means for draft picks. Thus, players from Canada and the United States are what most of the draft pool for a given draft will look like.
One interesting thing to note about this data is that a former country like the Soviet Union exists in this data due to the 1972 International Summit Series between team Canada and the Soviets. WHA players were not permitted
Next lets look a grouping by NHL team
team_summary <- nhl_draft |>
group_by(team) |>
summarise(count = n(),
avg_age = round(mean(age, na.rm = TRUE), 2),
avg_games_played = round(mean(games_played, na.rm = TRUE), 2),
avg_goals = round(mean(goals, na.rm = TRUE), 2),
avg_assists = round(mean(assists, na.rm = TRUE), 2),
avg_points = round(mean(points, na.rm = TRUE), 2),
avg_plus_minus = round(mean(plus_minus, na.rm = TRUE), 2),
avg_penalty_minutes = round(mean(penalties_minutes, na.rm = TRUE), 2)
) |>
mutate(prob = count/NROW(nhl_draft)) |>
relocate(prob, .after = count) |>
mutate(is_anomaly = case_when(
prob == min(prob) ~ "Anomaly"
)) |>
relocate(is_anomaly, .after = prob)
team_summary |>
print(n = 40)## # A tibble: 44 × 11
## team count prob is_anomaly avg_age avg_games_played avg_goals avg_assists
## <chr> <int> <dbl> <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Anah… 224 1.83e-2 <NA> 18.8 290. 42.2 73.8
## 2 Ariz… 74 6.04e-3 <NA> 18.3 120. 22.0 30.8
## 3 Atla… 78 6.37e-3 <NA> 19.7 332. 63.8 109.
## 4 Atla… 107 8.73e-3 <NA> 18.5 276. 43.2 62.8
## 5 Bost… 474 3.87e-2 <NA> 18.8 316. 54.8 95.3
## 6 Buff… 496 4.05e-2 <NA> 18.6 359. 56.9 93.4
## 7 Calg… 397 3.24e-2 <NA> 18.7 300. 52.7 84.5
## 8 Cali… 51 4.16e-3 <NA> 19.8 217. 50.1 73.3
## 9 Caro… 202 1.65e-2 <NA> 18.4 234. 36.6 55.3
## 10 Chic… 545 4.45e-2 <NA> 18.5 301. 51.2 86.1
## 11 Clev… 9 7.35e-4 <NA> 19.7 160. 24.8 34.4
## 12 Colo… 221 1.80e-2 <NA> 18.4 255. 40.2 70.0
## 13 Colo… 45 3.67e-3 <NA> 19.2 218. 34.1 76.1
## 14 Colu… 183 1.49e-2 <NA> 18.5 238. 33.6 55.4
## 15 Dall… 234 1.91e-2 <NA> 18.3 290. 45.0 69.4
## 16 Detr… 534 4.36e-2 <NA> 18.6 358. 67.0 106.
## 17 Edmo… 405 3.31e-2 <NA> 18.7 307. 56.0 92.7
## 18 Flor… 252 2.06e-2 <NA> 18.4 281. 37.5 66.1
## 19 Hart… 178 1.45e-2 <NA> 18.8 400. 65.8 109.
## 20 Kans… 26 2.12e-3 <NA> 19.6 285. 72.8 92.6
## 21 Los … 475 3.88e-2 <NA> 18.9 280. 46.9 80.8
## 22 Minn… 261 2.13e-2 <NA> 19.4 271. 40.9 73.5
## 23 Minn… 165 1.35e-2 <NA> 18.6 270. 38.8 67.7
## 24 Mont… 627 5.12e-2 <NA> 18.8 360. 57.4 102.
## 25 Nash… 205 1.67e-2 <NA> 18.6 308. 43.5 79.4
## 26 New … 386 3.15e-2 <NA> 18.5 306. 46.2 75.7
## 27 New … 496 4.05e-2 <NA> 18.7 347. 56.2 96.0
## 28 New … 551 4.50e-2 <NA> 18.8 334. 55.8 96.3
## 29 Oakl… 20 1.63e-3 <NA> 19.8 261 30.3 80.7
## 30 Otta… 250 2.04e-2 <NA> 18.6 313. 54.6 87.1
## 31 Phil… 509 4.16e-2 <NA> 18.8 283. 50.9 85.2
## 32 Phoe… 140 1.14e-2 <NA> 18.4 247. 34.8 64.4
## 33 Pitt… 463 3.78e-2 <NA> 18.7 285. 51.8 85.0
## 34 Queb… 182 1.49e-2 <NA> 19 362. 64.5 107.
## 35 San … 259 2.11e-2 <NA> 18.6 334. 51.5 82.8
## 36 Seat… 18 1.47e-3 <NA> 18.2 10 3 6
## 37 St. … 498 4.07e-2 <NA> 18.9 299. 46.4 79.3
## 38 Tamp… 269 2.20e-2 <NA> 18.6 236. 39.4 67.2
## 39 Toro… 509 4.16e-2 <NA> 18.7 310. 51.7 83.3
## 40 Vanc… 452 3.69e-2 <NA> 18.7 307. 56.0 87.0
## # ℹ 4 more rows
## # ℹ 3 more variables: avg_points <dbl>, avg_plus_minus <dbl>,
## # avg_penalty_minutes <dbl>We can see the average age of players for the draft is 18 which makes sense since players either can come straight from high school or play one year in college and enter the draft. Also different countries treat 18 year olds different when it comes to be considered as an adult. Teams also find this age useful because it allows for players to adapt and grow at a time that will allow them to perform better in their prime years.
Also what is interesting is that we can see all of the teams that were added or removed during the history of the NHL up to the present. Teams like Cleveland Barons and California Golden Seals are teams that played during the 1960s and 70 while the Seattle Kraken and Vegas Golden Knights are expansion teams that were founded within the last decade.
The counts for the numbers support the fact that these teams have not had long histories and thus when picked out of random will have more skewed means than teams who have been in the league for a long time.
Finally let’s look at grouping based on Overall Draft Pick
overall_pick_summary <- nhl_draft |>
group_by(overall_pick) |>
summarise(count = n(),
avg_age = round(mean(age, na.rm = TRUE), 2),
avg_games_played = round(mean(games_played, na.rm = TRUE), 2),
avg_goals = round(mean(goals, na.rm = TRUE), 2),
avg_assists = round(mean(assists, na.rm = TRUE), 2),
avg_points = round(mean(points, na.rm = TRUE), 2),
avg_plus_minus = round(mean(plus_minus, na.rm = TRUE), 2),
avg_penalty_minutes = round(mean(penalties_minutes, na.rm = TRUE), 2)
) |>
mutate(prob = count/NROW(nhl_draft)) |>
relocate(prob, .after = count) |>
mutate(is_anomaly = case_when(
prob == min(prob) ~ "Anomaly"
)) |>
relocate(is_anomaly, .after = prob)
overall_pick_summary |>
print(n = 40)## # A tibble: 293 × 11
## overall_pick count prob is_anomaly avg_age avg_games_played avg_goals
## <dbl> <int> <dbl> <chr> <dbl> <dbl> <dbl>
## 1 1 60 0.00490 <NA> 18.3 809. 250.
## 2 2 60 0.00490 <NA> 18.3 789. 226.
## 3 3 60 0.00490 <NA> 18.4 708. 153.
## 4 4 60 0.00490 <NA> 18.4 632. 152.
## 5 5 60 0.00490 <NA> 18.4 621. 113.
## 6 6 60 0.00490 <NA> 18.3 550. 123.
## 7 7 60 0.00490 <NA> 18.4 597. 124.
## 8 8 60 0.00490 <NA> 18.3 485. 94.8
## 9 9 60 0.00490 <NA> 18.4 528. 99.2
## 10 10 60 0.00490 <NA> 18.4 450. 82.0
## 11 11 59 0.00482 <NA> 18.4 563. 109.
## 12 12 59 0.00482 <NA> 18.4 428. 78.0
## 13 13 59 0.00482 <NA> 18.4 486. 86.0
## 14 14 59 0.00482 <NA> 18.3 528. 84.5
## 15 15 59 0.00482 <NA> 18.4 397. 96.0
## 16 16 59 0.00482 <NA> 18.6 394. 72.3
## 17 17 59 0.00482 <NA> 18.5 471. 82.6
## 18 18 59 0.00482 <NA> 18.4 375. 53.5
## 19 19 58 0.00473 <NA> 18.4 362. 64.2
## 20 20 58 0.00473 <NA> 18.4 393. 64.9
## 21 21 58 0.00473 <NA> 18.4 385. 62.2
## 22 22 57 0.00465 <NA> 18.4 393. 84.1
## 23 23 57 0.00465 <NA> 18.3 377. 55.3
## 24 24 57 0.00465 <NA> 18.5 357. 62.6
## 25 25 54 0.00441 <NA> 18.5 336. 51.3
## 26 26 54 0.00441 <NA> 18.3 338. 65.6
## 27 27 54 0.00441 <NA> 18.5 375. 50.7
## 28 28 54 0.00441 <NA> 18.4 312. 54.3
## 29 29 54 0.00441 <NA> 18.4 268. 44.4
## 30 30 54 0.00441 <NA> 18.5 295. 43.0
## 31 31 54 0.00441 <NA> 18.5 160. 17.2
## 32 32 54 0.00441 <NA> 18.4 273. 40.4
## 33 33 54 0.00441 <NA> 18.4 389. 70.2
## 34 34 54 0.00441 <NA> 18.5 206. 21.9
## 35 35 54 0.00441 <NA> 18.5 277. 44.7
## 36 36 54 0.00441 <NA> 18.6 324. 48.4
## 37 37 54 0.00441 <NA> 18.4 234. 28.1
## 38 38 54 0.00441 <NA> 18.7 311. 32.8
## 39 39 54 0.00441 <NA> 18.5 270. 40.7
## 40 40 54 0.00441 <NA> 18.6 341. 53.6
## # ℹ 253 more rows
## # ℹ 4 more variables: avg_assists <dbl>, avg_points <dbl>,
## # avg_plus_minus <dbl>, avg_penalty_minutes <dbl>Even though overall pick doesn’t have a huge impact in terms of player performance, we can get a glimpse of how each pick has done over time and how their impact has been in the league.
Let’s look at something about the draft picks that is something that is different than the other tables.
overall_pick_summary |>
arrange(desc(avg_points)) |>
print(n = 40)## # A tibble: 293 × 11
## overall_pick count prob is_anomaly avg_age avg_games_played avg_goals
## <dbl> <int> <dbl> <chr> <dbl> <dbl> <dbl>
## 1 1 60 0.00490 <NA> 18.3 809. 250.
## 2 2 60 0.00490 <NA> 18.3 789. 226.
## 3 3 60 0.00490 <NA> 18.4 708. 153.
## 4 4 60 0.00490 <NA> 18.4 632. 152.
## 5 6 60 0.00490 <NA> 18.3 550. 123.
## 6 7 60 0.00490 <NA> 18.4 597. 124.
## 7 5 60 0.00490 <NA> 18.4 621. 113.
## 8 264 10 0.000816 <NA> 18.5 491 114
## 9 11 59 0.00482 <NA> 18.4 563. 109.
## 10 9 60 0.00490 <NA> 18.4 528. 99.2
## 11 134 50 0.00408 <NA> 18.6 451. 87.7
## 12 15 59 0.00482 <NA> 18.4 397. 96.0
## 13 8 60 0.00490 <NA> 18.3 485. 94.8
## 14 14 59 0.00482 <NA> 18.3 528. 84.5
## 15 17 59 0.00482 <NA> 18.5 471. 82.6
## 16 13 59 0.00482 <NA> 18.4 486. 86.0
## 17 171 47 0.00384 <NA> 18.8 310. 91.4
## 18 22 57 0.00465 <NA> 18.4 393. 84.1
## 19 12 59 0.00482 <NA> 18.4 428. 78.0
## 20 210 46 0.00376 <NA> 18.7 280. 76
## 21 231 25 0.00204 <NA> 20.2 371 81.3
## 22 120 51 0.00416 <NA> 18.7 326. 84.5
## 23 214 33 0.00269 <NA> 19.8 461. 62
## 24 10 60 0.00490 <NA> 18.4 450. 82.0
## 25 288 5 0.000408 <NA> 18.8 462 104
## 26 183 46 0.00376 <NA> 18.6 448. 77.9
## 27 71 54 0.00441 <NA> 18.6 393. 69.6
## 28 257 11 0.000898 <NA> 20.6 357 82.3
## 29 156 48 0.00392 <NA> 18.7 436. 64.2
## 30 20 58 0.00473 <NA> 18.4 393. 64.9
## 31 16 59 0.00482 <NA> 18.6 394. 72.3
## 32 33 54 0.00441 <NA> 18.4 389. 70.2
## 33 250 19 0.00155 <NA> 19.7 428 34.2
## 34 21 58 0.00473 <NA> 18.4 385. 62.2
## 35 69 54 0.00441 <NA> 18.4 315. 70.4
## 36 166 48 0.00392 <NA> 18.6 370. 67.9
## 37 133 50 0.00408 <NA> 19.6 318. 59.3
## 38 19 58 0.00473 <NA> 18.4 362. 64.2
## 39 26 54 0.00441 <NA> 18.3 338. 65.6
## 40 121 51 0.00416 <NA> 18.6 302. 73.3
## # ℹ 253 more rows
## # ℹ 4 more variables: avg_assists <dbl>, avg_points <dbl>,
## # avg_plus_minus <dbl>, avg_penalty_minutes <dbl>Looking at the data above, if you sort the data by the average points there is draft pick 264 which pops up as 8th on the ranking! This means that those 10 players are skewing the mean of their points for the time that they played. Thus some overall picks we need to weight more than the ones with way less data.
This is also interesting data because we can see how players based on their draft number contributed to the value of their draft pick. This confirms that the 1-4 picks in drafts tend to actually live up to the expectation they hold as a number 1 pick.
After looking at these groups, there is a lot to unpack on why some groups are smaller than others. The probabilities of these different groups are impacted based on the year that they were drafted as well as the culture and decade they played in. After learning this information, there needs to be a deeper dive on each decade and how those players did.