
     Notes for the 1992 NFL data provided for use with Front Page 
     Sports: Football.
     
     
     Front Page Sports: Football in a very short time has already 
     become the mark by which all other computer football simulations 
     will be judged.  It is an amazing program and the creators 
     deserve all the praise we can heap on them for pulling off so 
     much that other developers claimed to be impossible with today's 
     technology.  That said, like other all other programs FPSF is 
     not perfect.  One item that is missing in FPSF is actual NFL 
     data though much of what is included is based on that same data.  
     There are those who argue that FPSF was not designed as an NFL 
     simulator but as a pure football simulator (designers and buyers 
     included) but I don't think that's representing 100% of the 
     situation.  Any simulation you create must be based on some 
     model and where football is concerned that model is obviously 
     the NFL.  The basis for this argument is found in FPSF as all 
     the NFL cities, among much other data, are included in the 
     shipping version.
     
     I'm not here to argue what factors went into the design, I'm 
     here to provide what I, and many others like me thought FPSF was 
     missing, simulated NFL data.  Basically I have taken the time 
     (and quite a bit of it) to go in and rearrange the data so that 
     much of the NFL is recreated.  Certain limitations have arisen 
     due to factors beyond by control.  For instance, it seems that 
     though most of the included data for the teams in FPSF is based 
     on actual NFL players, there are still 8-10 players on every 
     team that are purely fictional.  Almost all of those are deep 
     bench players, but they're there.  I've made every attempt to 
     reduce their role in this NFL data and when they do play simply 
     try and think of them as the rookies that come and go each and 
     every year.

     Also, through all of my work I've uncovered a few interesting 
     things about the way FPSF works.  For instance, FPSF does not 
     work on a Starter, Backup basis by position.  Instead it appears 
     that it actually evaluates each player at each position and then 
     makes a determination as to who should start.  This 
     determination is based on the 8 rating values Dynamix has chosen 
     for player replication.  The problem is that the first 6 of 
     these ratings are biased towards younger players.  They're 
     ratings like speed and agility which almost all younger players 
     have an edge over older players.  The final two ratings 
     basically represent all those years of experience a true veteran 
     acquires.  A human coach can look at Dan Marino, slow as he may 
     be on his feet, but realize that he's a Hall-of-Famer for his 
     experience and that he's a "winner".  FPSF takes all 8 ratings 
     and gives them equal weighting.  The end result is that rookie 
     QB's (and other positions as well) almost always start over 
     veterans.  In one season I just finished testing Dan Marino 
     never once started a single game for Miami.  Obviously an 
     oversight.

     To address this I've taken two steps the second of which I'll 
     explain a little further on down the line.  The obvious thing 
     was to bring the rookie back down to earth by trimming his stats 
     a bit.  Basically I reduced their last two ratings so that they 
     truly became what they are, students of the game still learning 
     the ropes.  This fixes most of the rookie/veteran mishaps, at 
     least at the QB position where it was most notable.  One funny 
     note about this.  Matt Bahr, the kicker for the NY Giants 
     recently was the team hero for the Giants in a test season I 
     just ran.  He carried them far into the playoffs.  Did he kick 
     a lot of field goals?  Nothing more than average, but he did 
     lead the team as their starting QB for the entire season.  He's 
     listed as the emergency QB for the Giants on their team roster.  
     Not only that but he had GREAT full season stats.  I've 
     addressed this but here's another one for Dynamix to dig into 
     further.

     The final hurdle appeared after all the work I did getting as 
     much of the data as close to right as I could get it.  I started 
     a season and after having fixed the veteran/rookie problem I 
     noticed a new problem, and I noticed again in Miami.  I thought 
     I had defeated the computer AI by making it realize that Dan 
     Marino was the right choice, at least I thought I had.  It 
     fought back.  The artificial intelligence looked over the 
     situation and said "Dan Marino? He stinks, we can't win with 
     this bonehead at QB" and so the computer manager traded a player 
     to Philadephia and obtained Jim McMahon to play OVER Dan Marino.

     Obviously this had to be addressed.  You can't achieve any kind 
     of accurate simulation if players are being traded all over the 
     place throwing off the balance that was sought in the first 
     place.  FPSF could make things much easier for us in this regard 
     if we could just click a button removing trading as an option 
     for computer owners but until that happens the only option is to 
     make each team human owned.  The drawback to this is that each 
     week you'll need to go into the Schedule portion of the league 
     and move the highlight from PLAY to SIMULATE for every game that 
     week.  Not a big deal but more than should be necessary. By the 
     way, this is also the second way I'm hoping to force the 
     Dolphins to use Dan Marino.  Maybe this guy should stick with 
     selling gloves on TV as this might be some sort of omen.  Oh, 
     yeah, human owned teams need passwords to make changes to data.  
     The password for every team is just an ENTER key.

     All-in-all the results of my testing are near what I was looking 
     for.  Real players are doing 95% of the playing and giving 
     rather realistic results though somewhat different each time 
     through.  I'd like to get as much input back about this data 
     with the hopes of continuing to make it as real as it can be.

     Hope you enjoy it as much as I am.

     Rich Heimlich
     President 
     Top Star Computer Services, Inc.
     CIS: 70670,2517
