New to MyFootballNow?  Click here for a tour!

Saturday, March 21, 2015

Random MFN Trivia

Did you know...?


  • Player and coach names are generated from US Census data, driven by probabilities of each name actually appearing in the United States population.
  • When testing the random name generator, the 5th name generated was "James Bond."
  • Weather data is not randomly generated, but is using actual historical weather conditions for the location. So if you notice all the teams in the same region experiencing similar inclement weather, it's not a coincidence.  Each season is mapped to a specific year in history and weather data for that location is looked up via foreacst.io. Each league will use a different year-to-year mapping.
  • League championship game locations are chosen randomly based on the average weather conditions for games played in each city in the league during the months of December and January. A league championship venue will not be repeated for 7 seasons.
  • While many football text simulations generate the overall result of each play by using probabilities based on the players on the field, in MyFootballNow each player in the simulation is autonomous, meaning that they make independent decisions based on information that only they are aware of.
  • A secret "development" simulation environment runs through a full season every two days. This is used to test the long-term effects of changes to the game engine.  Because of the complexity of the simulations, simming a whole season nonstop takes about 13 hours.
  • Each season of each league contains about 1GB of data, including all of the game playback information and the player historical attributes.
  • Early in development, an idea was floated out where the players would not be rated numerically, but the users would be required to actually study game film to make personnel decisions. This clearly would have been too time consuming to play.
  • Another idea in the early development was to simulate the referees on the field in addition to the players.  The plan was to only call penalties if the referees saw it, and also to add the possibility for referees to impact the play. Referees were going to be assigned attributes like players and would be more or less likely to call various penalties based on those attributes. There would have been bad refs and good refs, and they would have moved up and down the chain like players and coaches do.
  • If those scrapped ideas were not enough, one idea was to simulate every second of the real time game and allow watching the players run to and from the sidelines, into huddles, etc. This was dropped when it became apparent that the time it would take and the playback data it would require would be prohibitive.
  • The game engine alone took a full calendar year to bring to a point where it would be acceptable for testing.  Early testing with human testers saw games with scores in the three digits, which obviously meant that there was much to be done.

2 comments:

  1. I am not sure where to post but have a question about this league. The one who is going to win my division just scored 91 points has a rookie wide receiver playing tight end and has over 2000 yards receiving his QB is up to around 9000 passing am I missing something? I like to play but why with that?
    Thanks.

    ReplyDelete
  2. You are correct in that there are still cases where certain combinations of gameplanning and player strengths will cause extreme outliers as you describe. As the game develops, those outliers are getting more and more narrow, but because of the nature of the simulation and the complexity of what we are simulating, there will likely always be a measure of that.

    Feel free to post in the community forums your concerns or evaluations - there are lots of people playing who will give excellent feedback, and the community discussions drive what the focus for the next updates are.

    ReplyDelete