-=-**-=-›     SPREADSHEETS AND PREDICTIONS!›     By THOMAS R. SMITH, member of›     THE OL' HACKERS AUG, N.Y. › Though this is TOMS first article it's› a WINNER. Hope we get  more  from him!›› HEEDING THE "CALL"››    Alan Sharkis' call in the  May-June› Newsletter for us to "SPREAD THE WORD"› was quite a motivator to share a phase› of  8-bit  use that  has  been fun and› exciting----spreadsheet  handicapping:› turning   sports    statistics    into› predictions! ›› APPLICATION OF PRINCIPLES››    Though my examples will be based on› football,   many   of  the  principles› described  can  be  applied  to  other› sports. ›› A LITTLE KNOWLEDGE GOES A LONG WAY››      With     adequate     programming› expertise,  you  could write a program› to   make   predictions,  conveniently› prompt for inputs, and print an output› in  the  desired  format.   The   most› important  part  of  any  such program› however, would be the formula used  to› turn   the   statistical   data   into› forecasts.  Finding this formula takes› experimentation,    for    which     a› spreadsheet  is  ideal.  A spreadsheet› can  carry  on  numerous  simultaneous› experiments,   as  well  as  organize,› update, store and retrieve the numeric› data  you'll  be  dealing with in this› application.  A spreadhseet will  tell› you basically what you want your fancy› program to do.›› TRY, TRY, AND TRY AGAIN››    I   was   in   the  midst  of  such› experimentation when an issue of Antic› came  out  in  October  of 1987 with a› COVER   photo  of  a  football  player› running  with  a  130XE,  promoting  a› program to "Pick the  pros  -beat  the› spread!"   It  turned  out  that  this› program's formula to predict the score› of a football game  was  basically  to› split  the  difference  between   each› team's  offensive  points  scored with› their opponent's points given up.   My› spreadsheet  predictions  at  the time› were doing  about  10%  better,  so  I› continued  with  them,  though  always› testing new ideas.›› THIS IS ONE WAY TO DO IT››     For   a   hypothetical   game   of› Mustangs  vs  Zebras,   a   simplified› spreadsheet  version  of  that   Antic› program would look like this: ››         A    B    C    D    E    F›   1  Team   PF   PA  SPD  PSC  MGN›   2   MUS   30   14   -2   24    5›   3   ZEB   21   17   +2   17   -5›    › where   PF = points for›         PA = points against›        SPD = spread        › all from newspaper inputs, and ›› PSC = predicted score, (B2+C3)/2›                       for cell E2›                        (B3+C2)/2›                       for cell E3› MGN = margin, D2+(E2-E3) for cell F2›               D3+(E3-E2) for cell F3›› This  MGN  calculation  indicates that› the Mustangs are  predicted  to  "beat› the  spread" by 5, quite a comfortable› "margin".›› WHY DOES IT WORK?››     How  was  I  able  to  get  better› results   than   the   Antic  Football› Predictor?  The  logic  of  the  Antic› formula  is  actually quite sound -the› key is having that formula operate  on› better  data, -data that a spreadsheet› is ideally suited to provide.›› USING BETTER DATA››    Part of that  better  data  key  is› gathering  it over the right period of› time.  Teams change week by  week,  so› the  team  on  the  field the upcoming› weekend is not the same team you  have› statistics  for  6  or  7 weeks past. › Data you have  for  a  team  beyond  a› certain  point  does  not   accurately› represent    the    current     team. › Countering  this,  teams  have bad and› good days, luck, and officiating  that› results  in  single  game  scores that› aren't   representative.    It   takes› several weeks worth of scores to get a› realistic measure.›    Another part of the better data  is› taking  into  account  the strength of› opponents.  Winning by  14  points  is› not  as  significant  against  a  weak› losing  opponent  as  a strong winning› one.  Over the proper period of  time,› most teams will have accumulated their› statistics   against   an   adequately› similar  range of opponent strengths. ››› IS THERS A BETTER WAY?››    I don't know of  a  better  way  to› determine  this period of time than to› experiment,  and  to  my  delight  and› surprise, a certain  number  of  weeks› worth  of  data  has  always  produced› better results that greater and lesser› numbers of weeks. I'd spoil  your  fun› by telling you how many!›› MATH TO THE RESCUE››    There  are  creative   mathematical› ways  to  factor in opponent strength,› formulas easily experimented with  and› tested  by  spreadsheet.  For example,› you could create a new scoring  system› of your own in which the winner scores› the opponents's  wins  going  in  plus› one,  and  the loser scores nothing or› maybe the winner's wins minus one. ›› A NARROW VIEW››    Each  of  the  numerous  statistics› published in the Sports section of the› newspaper  is  a reasonable measure of› performance, albeit a narrowly focused› aspect.  Each statistic, by comparison› with   the   same   statistic   of  an› opponent,   makes   a  prediction.   A› spreadsheet makes it  relatively  easy› to  keep  track of numerous statistics› and   their   predictions,   and  with› follow-up, keep track of how effective› a  predictor each statistic was.  With› this effectiveness information  it  is› possible  to  refine  each statistic's› prediction, and  give  it  appropriate› "weight".›› TWO SPREADSHEETS USED››    The  two  spreadshheets  that  I've› used  are  the  GT Albert Spreadsheet,› that came with Indus drives, and  Mike› Silva's  SynCalc,  marketed by Synapse› and   Broderbund.    SynCalc   is    a› powerful,  user  friendly gem that was› also produced for the  Commodore  64. › This  C64 version was on one side of a› disk, the Atari version on the  other.› If you find SynCalc, BUY IT! ›› OTHER SPREADSHEETS››     (NOTE:    several    other   Atari› commercial programs were produced this› way,  Mercenary,  for  example, so you› just might score some  terrific  Atari› software  on the back of a C64 disk at› a garage sale or swap meet somewhere.)› Other spreadsheets that probably would› be OK, though I  haven't  tried  them,› are  Calc  Magic  and  VisiCalc.  Calc› Magic received a  rave  review  in  an› L.A.  area  (Antelope  Valley  -AVACE)› user group newsletter a few years ago,› and   was   available   through    C&T› ComputerActive, P.O. Box 893, Clinton,› OK 73601.) Another great  way  to  use› the  lil  ol'  ATARI, and maybe make a› BUNDLE!›              =+=end=+=››