Moving Beyond Excel For Your Hockey Analysis

# Moving Beyond Excel For Your Hockey Analysis
## Meghan Hall
### #HANIC April 18, 2020
### <a href="https://twitter.com/MeghanMHall">MeghanMHall</a> <a href="https://github.com/meghall06">meghall06</a> <a href="https://meghan.rbind.io/">meghan.rbind.io</a>

---

# About me

Data manager in higher education

]

Amateur hockey analyst ([Hockey-Graphs!](https://hockey-graphs.com))

]

R enthusiast

]

---

# About you

Working with data manipulation, visualization, and analysis

]

Maybe you use Excel?

]

Curious about learning to code (or already on your way!)

]

---

# Rule #1

### No Excel shaming

---

# Goals for tonight

### Limitations of Excel 
- When your goal is **efficient** and **reproducible** and **shareable** analysis

### Advantages of R 
- (Or whatever language you like)

### Confidence to move on
- Find resources and develop a learning roadmap

---
class: inverse, center, middle

# Data Example

## From Moneypuck

---

name: moneypuck
class: inverse, middle, center
background-image: url(img/Moneypuck.png)
background-size: contain

---

name: moneypuck2
class: inverse, middle, center
background-image: url(img/Moneypuck2.png)
background-size: contain

---

# Two questions we have from this data

### Longest average shifts?

- Forwards only
- Strength state: 5v4
- 100+ shifts

### Most shifts per game?

- Forwards vs. defensemen
- Strength state: 5v5
- 10+ games

---

# Easy enough to do in Excel!

### Longest average shifts?

- create a **shift_length** variable: `icetime / shifts`

- filter `position` to L, C, R

- filter `situation` to 5on4

- filter `shifts` to at least 100

- sort descending by `shift_length`

---

name: moneypuck4
class: inverse, middle, center
background-image: url(img/Moneypuck4.png)
background-size: contain

---

# Easy enough to do in Excel!

### Most shifts per game?

- create a **shifts_game** variable: `shifts / games_played`

- create a **position_grouped** variable to create D and F

- filter `situation` to 5on5

- filter `games_played` to at least 10

- find the average values for `shifts_game`

---

name: moneypuck3
class: inverse, middle, center
background-image: url(img/Moneypuck3.png)
background-size: contain

---

# But...

---

# ...what if...

&nbsp;someone asks you exactly what you did? &nbsp;😬

&nbsp;you get refreshed data? &nbsp;😬

&nbsp;you make a mistake, or there's something you want to make note of? &nbsp;😬

&nbsp;there's more data you want to join in? &nbsp;😬

&nbsp;you have to make a graph? &nbsp;😬

---

# Let's try this in R

---

# Longest average shifts?

```r
shift_length <- moneypuck %>%
 mutate(shift_length = icetime / shifts) %>%
 filter(position != "D" & situation == "5on4" & shifts >= 100) %>%
 arrange(desc(shift_length)) %>%
 select(name, shift_length)
```

```r
head(shift_length)
## # A tibble: 6 x 2
## name shift_length
## <chr> <dbl>
## 1 Alex Ovechkin 157.
## 2 Connor McDavid 138.
## 3 Leon Draisaitl 116.
## 4 Nathan MacKinnon 112.
## 5 Jack Eichel 108.
## 6 Auston Matthews 102.
```

---

# Most shifts per game?

```r
shifts_game <- moneypuck %>%
 mutate(shifts_game = shifts / games_played,
 position_grouped = ifelse(position == "D", "D", "F")) %>%
 filter(situation == "5on5" & games_played >= 10) %>%
 group_by(position_grouped) %>%
 summarize(avg_shifts_game = mean(shifts_game))
```

```r
head(shifts_game)
## # A tibble: 2 x 2
## position_grouped avg_shifts_game
## <chr> <dbl>
## 1 D 19.7
## 2 F 15.8
```

---

# ...what if...

&nbsp;someone asks you exactly what you did? &nbsp;😄

---

```r
# Always comment your code as if others will read it
# "Others" includes yourself in 3 months when you've 
# forgotten everything

# This is the "Skaters" file from MoneyPuck as of 2020-04-11

moneypuck <- read_csv("MoneyPuck.csv")

# Created a new variable to calculate average shift length
# Filters: forwards only, 5on4, at least 100 shifts 
# Can easily include justification for why you did all this
# Because you WILL forget

shift_length <- moneypuck %>%
 mutate(shift_length = icetime / shifts) %>%
 filter(position != "D" & situation == "5on4" & shifts >= 100) %>%
 arrange(desc(shift_length)) %>%
 select(name, shift_length)
 
```

---

# ...what if...

&nbsp;someone asks you exactly what you did? &nbsp;😄

&nbsp;you get refreshed data? &nbsp;😄

&nbsp;you make a mistake, or there's something you want to make note of? &nbsp;😄

---

```r
# You can create functions for common tasks
# Let's say you hate the Leafs and always want to 
# remove them for your data
# That would be a pain to do every time by hand, so:

no_leafs_no <- function(original_data, team_variable) {
 
 original_data %>%
 filter(team_variable != "TOR")
 
}

moneypuck_no_leafs <- no_leafs_no(moneypuck, moneypuck$team)

# Or more realistically, you always want to remove a certain game
# Or you have tasks that you do frequently
# If you're copying and pasting code, it's time for a function
 
```

---

```r
# You can also run tests to check for mistakes
# Should have 31 teams and 883 players

moneypuck %>%
 summarize(teams = n_distinct(team))
## # A tibble: 1 x 1
## teams
## <int>
## 1 31

moneypuck %>%
 summarize(players = n_distinct(name))
## # A tibble: 1 x 1
## players
## <int>
## 1 883
```

---

# ...what if...

&nbsp;someone asks you exactly what you did? &nbsp;😄

&nbsp;you get refreshed data? &nbsp;😄

&nbsp;you make a mistake, or there's something you want to make note of? &nbsp;😄

&nbsp;there's more data you want to join in? &nbsp;😄

---

```r
# Read in your tracking data that has your proprietary statistic:
# Gutsy Recoveries In Transition

GRIT <- read_csv("your_tracking_data.csv")

# Join that into the moneypuck data by both player and season

moneypuck_w_tracking <- moneypuck %>%
 left_join(GRIT, by = c("name", "season", "team"))
 
```

---

# ...what if...

&nbsp;someone asks you exactly what you did? &nbsp;😄

&nbsp;you get refreshed data? &nbsp;😄

&nbsp;you make a mistake, or there's something you want to make note of? &nbsp;😄

&nbsp;there's more data you want to join in? &nbsp;😄

&nbsp;you have to make a graph? &nbsp;😄

---

# Longest average shifts?

```r

shift_length %>%
  top_n(10) %>%
  ggplot(aes(y = shift_length, x = reorder(name, shift_length))) +
  geom_col() +
  coord_flip() +
  labs(
    y = "Average Shift Length",
    x = NULL,
    title = "Average Shift Length Among Forwards",
    subtitle = "2019-20 NHL Season, 5v4 Only",
    caption = "Data from Moneypuck"
  ) +
  geom_text(aes(label = round(shift_length, 1)), colour = "white", 
            size = 3, position = position_stack(vjust = 0.07)) +
  meg_theme()
```

---

# Longest average shifts?

]

---

# Most shifts per game?

```r

shifts_game %>%
  ggplot(aes(y = avg_shifts_game, x = position_grouped)) +
  geom_col() +
  labs(
    y = "Average Shifts per Game",
    x = NULL,
    title = "Average Shifts per Game By Position",
    subtitle = "2019-20 NHL Season, 5v5 Only",
    caption = "Data from Moneypuck"
  ) +
  geom_text(aes(label = round(avg_shifts_game, 1)),
            size = 4, position = position_stack(vjust = 1.05)) +
  meg_theme()
```

---

# Most shifts per game?

]

---

# Other things you can do with R

## Github

- Integrates easily so you can collaborate, save, share. Version control!

## RMarkdown

- Useful for creating and sharing all kinds of content

- Like slides &nbsp;😇

---

```r
shift_length %>%
  top_n(10) %>%
  mutate(shift_length = round(shift_length, 2)) %>%
  knitr::kable(format = "html")
```

<table>
 <thead>
 <tr>
 <th style="text-align:left;"> name </th>
 <th style="text-align:right;"> shift_length </th>
 </tr>
 </thead>
<tbody>
 <tr>
 <td style="text-align:left;"> Alex Ovechkin </td>
 <td style="text-align:right;"> 156.64 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Connor McDavid </td>
 <td style="text-align:right;"> 138.04 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Leon Draisaitl </td>
 <td style="text-align:right;"> 116.45 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Nathan MacKinnon </td>
 <td style="text-align:right;"> 111.85 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Jack Eichel </td>
 <td style="text-align:right;"> 108.48 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Auston Matthews </td>
 <td style="text-align:right;"> 102.48 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Nicklas Backstrom </td>
 <td style="text-align:right;"> 102.39 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Patrick Kane </td>
 <td style="text-align:right;"> 101.51 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Ryan Nugent-Hopkins </td>
 <td style="text-align:right;"> 99.27 </td>
 </tr>
 <tr>
 <td style="text-align:left;"> Brad Marchand </td>
 <td style="text-align:right;"> 98.24 </td>
 </tr>
</tbody>
</table>

---

# Other things you can do with R

## Github

- Integrates easily within RStudio so you can collaborate, save, share

## RMarkdown

- Useful for creating and sharing all kinds of content

- Like slides &nbsp;😇

## Make a website

- Using `Shiny` (like Evolving-Hockey!)

- Using `blogdown` (like my website!)

---

# Okay but how?

---

# Moving past Excel conceptually

### It is comforting to *see* all of your data

- but that's not really necessary!

- much more efficient to look at the pieces you need

### Excel quickly becomes unwieldy with too much data

- if you're working with NHL play-by-play data? 1M+ rows

### Easier to keep track of data frames in an R file

- rather than a lot of workbooks and sheets

### Separate your data from your analysis!

---

# Treat it like learning a new language

### Excel

Create a new variable that creates **position_grouped**

]

### R

```r
mutate(position_grouped = 
         ifelse(position == "D",
                "D", "F"))
```

]

---

# Treat it like learning a new language

### Excel

Sort descending by `shifts_game`

]

### R

```r
arrange(desc(shifts_game))
```
]

---

# Treat it like learning a new language

### Excel

Make a pivot table by position and find average shifts

]

### R

```r
data %>%
  group_by(position_grouped) %>%
  summarize(avg_shifts_game = 
              mean(shifts_game))
```
]

---

# Treat it like learning a new language

### Excel

Filter to forwards only

]

### R

```r
filter(position != "D")
```
]

---

# Tips for learning

&nbsp;&nbsp;&nbsp;Practice a little bit frequently!

&nbsp;&nbsp;&nbsp;Pick **one** resource and go all the way through

&nbsp;&nbsp;&nbsp;Get really good at Googling

&nbsp;&nbsp;&nbsp;Use a project or a question as motivation

---

# My favorite resources

### R for Data Science &nbsp; [](https://r4ds.had.co.nz/)

- Focuses on the tidyverse
- Can skip from chapter to chapter to focus on what you need
- Really helpful examples you can follow along with

### R for Excel Users workshop &nbsp; [](https://rstudio-conf-2020.github.io/r-for-excel/)

### My tutorials — gotta plug my own stuff &nbsp;🤷

- [Intro to R on Hockey-Graphs](https://hockey-graphs.com/2019/12/11/an-introduction-to-r-with-hockey-data/) (with lots of exercises)
- [Intro to R with Swirl](https://meghan.rbind.io/post/hockey-data-with-swirl/) (interactive learning in the console)
- Various ad hoc tutorials (exploratory data analysis, tidymodels, etc.)

---

# Where to find me

&nbsp;[MeghanMHall](https://twitter.com/MeghanMHall)

&nbsp;[meghall06](https://github.com/meghall06)

&nbsp;[meghan.rbind.io](https://meghan.rbind.io)

---

# Go forth and code &nbsp;🙏

Slides created via the R package [**xaringan**](https://github.com/yihui/xaringan).