s4-advanced-analytics.Rmd

---
title: "Session 4: advanced analytics"
author: "Ben Raymond, Adrien Ickowicz"
date: "<div id='logo_footer'></div>"
output:
  xaringan::moon_reader:
    lib_dir: libs
    seal: false
    self_contained: false
    css: ["extra/extra.css"]
    chakra: extra/remark-latest.min.js
    nature:
      highlightStyle: github
      highlightLines: false
      beforeInit: "extra/macros.js"
---

layout: true

<div class="my-footer">
  <div class="my-footer-box"><a href="https://openvolley.org/"><img style="display:inline;" src="extra/ovoutline-w.png"/>openvolley.org</a></div>
  <div class="my-footer-box"><a href="https://https://volleyball.ca/"><img src="extra/vc-w-wide.png"/></a></div>
  <div class="my-footer-box"><a href="https://untan.gl/"><img src="extra/su_title-w.png"/></a></div>
</div>

---

```{r xaringanExtra-clipboard, echo=FALSE}
xaringanExtra::use_clipboard()
xaringanExtra::use_panelset()
```

```{r setup, include=FALSE}
options(htmltools.dir.version = FALSE)
options(knitr.kable.NA="")
knitr::opts_chunk$set(echo = FALSE, warning = FALSE, message = FALSE, cache = FALSE, dpi = 120)
library(dplyr)
library(formattable)
library(ggplot2)
library(ggsci)
library(datavolley)
library(ovdata)
`%eq%` <- function(x,y) x==y & !is.na(x) & !is.na(y)

```

class: inverse, logo, center

<img src="extra/3logo2.png" style="width:65%; margin-bottom:50px;" />

## Session 4: Advanced analytics

### Adrien Ickowicz, Ben Raymond

##### with valuable contributions from many others...

---

## What are we going to talk about

- Statistical modelling
    - descriptive statistics vs inference
    - Is causality reachable ?
    - Inference via simulation vs Inference via statistical modelling
- Simulating matches
    - the <span class="pkg">volleysim</span> package
    - the <span class="pkg">bandit</span> package
- Statistical (inference-based) models
    - Use the example of the effect of player height on passing performance
    - the effect of serve speed on outcome? Serve error vs breakpoint ?


---

## Statistical modelling
#### descriptive statistics vs inference

From wiki:

A descriptive statistic (in the count noun sense) is a summary statistic that quantitatively describes or summarizes features from a collection of information, while descriptive statistics (in the mass noun sense) is the process of using and analysing those statistics.

Example of descriptive statistics:
* Number of serve errors per set
* Break-point rate per set

Example of inference:
* How do these two quantities relate to one another?

---

## Statistical modelling
#### Causality

<img src="extra/xkcd_causality.png" style="width:70%; margin-bottom:50px;" /> 

credit: https://xkcd.com/

---

## Statistical modelling
#### Causality

<img src="extra/xkcd_causality_v2.png" style="width:90%; margin-bottom:50px;" /> 

credit: https://xkcd.com/

---

## Statistical modelling
#### Causality and correlation

https://www.tylervigen.com/spurious-correlations

So what is the problem? 

- An increase in the number of people at the beach causes a higher number of ice cream sales
- An increase in the number of people at the beach causes a higher number of shark attacks
- Therefore, people need ice cream for comfort after a shark attack; or is it that sharks attack people after they ate ice cream because they taste better?

---

## Statistical modelling
#### Inference via simulation vs Inference via statistical modelling

When do we use either:
- When the world can be simplified without loss of (too much) information, we can simulate;
    - Example: Can we simplify a volleyball game if we only want to know who is going to win? Anyone has an opinion?
- When the world is too complex, and we have no clue or the process is convoluted, we may want to try something simpler at first;
    - Example: Are tall players better passers than short players?

In a nutshell, when you cannot narrow down the factors that influence the uncertainty to a handy few, forget about simulation.
    
    
---

## Simulating matches
#### the <span class="pkg">volleysim</span> package:

```{r list_function_vs, results='asis'}
library(volleysim)
list_fun_vs <- lsf.str("package:volleysim")
cat(paste('-', list_fun_vs), sep = '\n')
```


---

## Simulating matches
#### the <span class="pkg">volleysim</span> package:

A quick overview: https://untan.gl/r-package-simulating-volleyball.html

Our turn. Open `s4-example-volleysim`.

---

## Simulating matches

The bandit functions in the <span class="pkg">ovlytics</span> package

```{r list_function_ovl, results='asis'}
library(ovlytics)
list_fun_ovl <- lsf.str("package:ovlytics")
cat(paste('-', list_fun_ovl[c(14,11, 10, 8, 4)]), sep = '\n')
```

What is this? See https://untan.gl/setter-choice.html

Our turn. Open `s4-example-bandit`.

And if you dont want to code... https://apps.untan.gl/bandit/

---

## Do you know...

More analytics articles:

https://untan.gl/index.html

---

## Statistical models
#### Player height and passing performance

Question: Does height influence the passing quality?

Open `s4-player-height-passing`.

---

## Statistical models
#### Serve agression, speed and outcome

Question 1 : Should we ban serving errors ?

Question 2 : Is there an optimal serve speed for which the expected break-point is maximum ?

Open `s4-serve-speed-example`.

---


## Statistical models
#### Identifying a defensive system

A blog post: https://untan.gl/identifying-defensive-systems.html

An app: https://apps.untan.gl/volleydef/

<img src="extra/defensive_zones.png" style="width:90%; margin-bottom:50px;" /> 

---

## Statistical models
#### Identifying a reception system

Can we do the same with reception? 

An app also: https://apps.untan.gl/volleypass/

<img src="extra/reception_zones_P1.png" style="width:70%; margin-bottom:10px;" /> 

WMCH 2018 - Italy - P2. What is happening?