UK Premiership Fantasy Football game API/Data decoder interface & Analytics platform.

• (this doc is formatted as Github Markup lang)
• NOTE: The dataset structures & API of the fantasy football game at https://www.premierleague.com is proprietary and closed. It is not publicly documented and not open to public consumption. No subscription/consumption license is available or offered by https://www.premierleague.com. Subsequently, NO license to those services is provided, inherited or inferred via the use of this code.
• Updated for 2019/2020 season
  • As usual...during the 2019 off-season, the Premiership data Architects & Developers changed many components in the internal JSON data structures of the underlying data model. This is normal (since their API & data-model is not public). This breaks API extract operations & code.
  • Current codebase is now aligned to 2019/2020 season.

A companion to EPL Fantasy Football League game, played by 6,000,000+ users worldwide via web app/mobile app at the Barclay's EPL website: http://www.premierleague.com/en-gb.html.

v2.0 of the project is focused on the following priorities...

1. Live mem JSON ETL (Extract, Translate, Load), rather than save data to flat files & post processing datsets.
2. Native JSON data feed via API I/O. No scraping raw data from JavaScript pages @ http://www.premierleague.com/en-gb.html
3. Understanding all available REST API's used by EPL website app (i.e. a private API not publicly exposed/documented by EPL).
4. Prototyping & validating the API data I/O interface & technology, toolset & overall private game JSON data structures & accessors.
5. Processing API extracted JSON datasets in memory (live) for speed. (prototype v1.0 used flat file processing).
6. Push data into MongoDB database where/when needed.
7. Augment game decision making logic with key intelligence factors not present/provided/computed/offered in the native EPL game logic.
8. Augmenting the game dataset with additional data/info where key data/fields/structures are not present/provided in native EPPL game & datasets.
9. Enable users with ability to apply State-of-Art Datascience logic to game decision making. e.g. analysis, statistics & probably A/B Hypothesis testing against EPL game data (via Datasciecne tools such as: Numpy, Pandas, Matplotlib, Scikit-Learn, Scipy & the Berkley Datascience module). Goal is to provide this capability via a Web/GUI experience, not force EPL game players to use shell & write python code.
10. Make core game data sets available, especially full historical data in non-aggregate form
11. Augment EPL game data with additional correlated game data from other non-EPL sources. e.g. FIFA, gambling website sites (e.g. odds), football status sites, Football news/rumor sites etc. Enable this additional data to be leveraged by users in advanced inferential Datascience based analytical game decisions.

• Not all of these capabilities are available in v2.0

epldecoder.py [-h] [-d] [-l LEAGUE] [-p PLAYER] [-q QUERY] [-r] [-v] [-x] -u USERNAME -c PASSWORD [-g GAMEWEEK]

optional arguments:

• -c PASSWORD, --password PASSWORD password for accessing EPL website
• -d, --dbload save JSON data into mongodb
• -g GAMEWEEK, --gameweek GAMEWEEK game week to analyze
• -h, --help show this help message and exit
• -l LEAGUE, --league LEAGUE league entry id - (a league that you want to evaluate)
• -p PLAYER, --player PLAYER team player id - (a valid user ID registered in the EPL websit)
• -q QUERY, --query QUERY squad player id - (a valid selectable squad player...eg. Harry Kane's player code)
• -r, --recleague recursive league details - (scan every league you are registered in)
• -u USERNAME, --username USERNAME a valid username for accessing EPL website
• -v, --verbose verbose error logging
• -x, --xray dump out some raw JSON data dicts & structures

How to find some of the critical EPL game id's that are needed for cmd line options...

For example...

• Username & Password are required
• Player id is required
• league id is optional

Player_ID - required

You must know your Player id (or a player you want work with) from the real EPL website game: http://www.premierleague.com

To find this number...

1. Log on the Fantasy football website game
2. Click on the 'Points' menu option
3. Look up at the browser URL bar. You will see a URL like this...https://fantasy.premierleague.com/a/team/7766554/event/34
4. Your player ID is the number after the word 'team'. e.g. the number '7766554'

To lookup the player ID of some other game opponent/user, you need to do this...

1. Click on the 'Leagues' menu option
2. Select a league that has an opponents team in it that you are interested in
3. Look up at the browser URL bar. Yo will see a URL like this...https://fantasy.premierleague.com/a/team/33221100/event/34
4. The players ID is the number after the word 'team'. e.g the number '33221100'

League_ID - optional

To find this number...

1. Click on the "Leagues" menu option
2. Select a league that you are interested in
3. Look up at the browser URL bar. You will see a URL like this...https://fantasy.premierleague.com/a/leagues/standings/255552/classic
4. The league ID is the number after the word 'standings'. e.g. the number '255552'

The goal is to augment the game decision making process with additional analytical data sets that are sources from other locations. These will be created & stored within a MongoDB and cross-referenced with the basic game data-set. Today, this is not easy to do via the current game website: http://www.premierleague.com/en-gb.html. Doing this makes deeper team & player analysis more possible (but more complex). and therefore, better team/player decisions are possible. Many of these data sets come from non EPL websites. e.g. FIFA, gambling sites, football stats sites etc.

Also, the MongoDB data source will be used to drive richer analytical charts that are not available on http://www.premierleague.com/en-gb.html, never were available and are very much needed by the 5,000,000+ users of the Barclays EPL website & App.

These charts can graph info based on the standard dataset from the EPL website as well as graph additional advance charts.

That EPL website used to provided a set of charts to users but this service was removed and discontinued. Fantasy Football players no longer have any charts/graphing service to support their team decisions

A ML/AI engine will also be built and offered to individual players. A player will be able to teach the ML/AI (via a web app/mobile app) to emulate his/her individual analytical team decision making style. Thereby automating the mundane process of choosing a squad by churning through pages of weekly data & relying too much on fuzzy personal logic, gut-feel, intuition, rumors and voodoo on what squad to pick & set-up. This will also allow EPL Fantasy Football players to run 1000's of possible permutations against potential squads (e.g., who to captain, who to vice-captain, when to play a wild card, what to play special options etc...all before the game day cut-off time. The ML/AI engine will operate under a set of personalize-able & tunable rules that "you" can define and control and do things like ...

• Scan for injury rumors, news and info then compute player not-starting probabilities
• Scan for hot & cold trends
• Extrapolate scoring runs and trends and predict possible 'Black Swan' events
• Compute best captain probabilities
• Compute best & worst games to play/avoid.
• Compute highest scoring probability of each game and player scoring probabilities
• Propose possible teams, transfers and team configs

...etc, etc

Data Science

As of 2019/2020 season, the code now implements & leverages Data Science code & functions provided by NUMPY & PANDAS.

This was always the long term strategy.

These python Data Science modules work well in a LINUX env & are easy to install, but can be tricky/problematic to get installed & running in a local Windows OS env. My Windows Dev environment uses Anaconda Python because it's the only way I can get a working Matplotlib for Windows Python module that works).

More to come. ~Orville