-=-**-=-¢ 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=+=¢¢