mockdraft

Tommy Hegerich

Project

Create a Fantasy Football Mock Draft simulator. The program will (if it hasn't been done that day) retrieve the adp data from a site (currently the only site is fantasypros). It will then fill out the SQL database with the players.

The mock draft is done through commandline (maybe javafx) with a CPU advancement level chosen beforehand (stupid, random, decent). The results of the user's team may be later stored in a json file for future retrieval.

Maybe include something to determine the best pick available. Use adp and standard deviations to run simulations for the user pick. This works by running separate simulations for the next X players (maybe the players on the user's queue?) and returning the one with the best average adp. Simulations work recursively (this would be exponential so whatever is done calculation wise must be trivial).

SQL

The sql is only local now. Must setup a local sql database (local instance 3306. I set up the user to be root and password as root but you can change that for your local system if you want).

The sql is pretty slow and useless. Might as well just store everything in data structures or json as of now (with just the commandline option). I remember sql fitting nicely with javafx (so it would be really easy to have a table of available players in the gui). But if gui isn't a feature then just add the json option that will primarily be used.

Purpose

To gain experience in java web scraping, SQL, JSON, and javafx. Also make a cool algorithm that we could use come fantasy football season. Maybe this could actually be used for quick mock drafts against the cpu.

Libraries Used

org.json:json:20190722
org.jsoup:jsoup:1.13.1
mysql:mysql-connector-java:8.0.16
org.springframework:spring-test:4.2.4.RELEASE
org.springframework:spring-jdbc:5.2.6.RELEASE
org.apache.commons:commons-math3':3.0
org.glassfish.jersey.containers:jersey-container-servlet:2.25.1
org.glassfish.jersey.media:jersey-media-json-jackson:2.25.1
javax.activation:activation:1.1
jakarta.xml.bind:jakarta.xml.bind-api:2.3.2
org.glassfish.jaxb:jaxb-runtime:2.3.2

Outlook

Data

-get ADP/STDV data from FFCalculator
    -place in draft data table and JSON file
-get Projection data from FantasyPros
    -update draft table with data
-ADP/STDV JSON file -> 2D Array of Player-Pick-Probs
    -in Suggestor Class
-do all of these 3 times (one for each scoring type)
    -store these on server
    -these tables read-only, except for scraping

-when draft starts, copy entire table to local arraylist from server
-each player starts with team # -1
    -updated once drafted
-use arraylist stream to present players

Draft

-get score type, league size, user pick
-start draft
-Controller gets players from DataGetter
-Controller deletes players with DataStorer
-DataGetter and DataStorer determine which table to use by the score type
    -constructor changes the url of the connection
-each user pick gets suggestions
    -CLUI -> Controller -> Suggestor -> Simulator
    -Suggestor uses DataGetter to get all the players
    -Suggestor calls a bunch of threads to get vals for each player
        -this is done in the Simulator class (concurrency)

Sims

-each sim is a thread, run concurrently
    -in the Simulator class, simulate() method
-run a bfs for 5-6 levels (basically rounds)
-each player is a node
    -outgoing edges to the next N players (or something like that, 
     we could figure out which players to go to based on stdev or 
     something)
        -maybe create an actual graph before draft if it is too 
         complicated
        -maybe every available player is a neighboring node, but only 
         actually consider the players who you would consider at the next 
         pick (their prob of being at the next pick is within 2 
         arbitrary bounds. for example kickers are all available in 
         the 2nd round but you wouldnt take them)
-each node has its own value
    -recursively go thru a bfs, using each value to get the original 
     node's value (by adding, multiplying, not sure yet)
    -once value is found, store it
        -stored in Hashmap, passed into function
            -or Hashmap is owned by the Simulator class, which is 
             passed by reference through its constructor, so all the 
             threads share it
        -because of concurrency, need locks for this