The Story of Gridline
The 20+ year history of the evolution of the GRIDLINE football forecasting computer system. By Gridline's creator.

My First Bet
Playing for Real
The Birth of Gridline
The Evolution of Gridline
What's in a Name?
Gridline in the Private Sector
Tweaking the System (or how I learned to quit worrying and love the HFA)
Changing Environments
How Strong is That?
You Can Go Home Again
Can't Leave Well Enough Alone
The Internet Calls
What the Future Holds


I can't begin to tell the story of Gridline without revealing a little bit about myself, Gridline's creator. I am a baby boomer, born in the late 50's in New Orleans, where I have always resided. I graduated from the University of New Orleans with a Bachelor of Science degree in Mathematics. I've been employed as a Programmer/Analyst and then as a Systems Analyst for several small companies and I'm currently employed as something called a Network Systems Specialist with a well known megacorporation.

I married a beautiful librarian in the early 90's. No kids. One dog. One cat. My wife assists me with the website copy when she has the time. When she doesn't you'll see several typos and grammatical errors throughout the site.

My First Bet

I was introduced to gambling by my uncle when I was a prepubescent kid. I lost 5 bucks betting on the Raiders in what was then called the World Championship of Professional Football. I had 14 points but it wasn't enough as Green Bay walked from one end of the field to the other while Daryl Lamonica kept getting helped up off the ground by a jovial bunch of Packer defensive linemen. I was suprised when my uncle held me to the bet but I learned two things from that episode: 1) Don't bet everything you have on a single game and B) You lose you pay. I still wonder if he would have paid me had I won.

My dad played the parlay cards back then. He never bet more than a buck but he loved to collect that 6 dollars when he hit his 3 team card, which he did quite frequently. Eventually I started getting the cards for myself and my dad, and each week we would sit down and circle the numbers. As the seasons went by I upped the amount I would bet on them but Dad always stayed with that buck. He continued to win the 6 dollars on a regular basis, and my 10 team parlays never did come in.

Playing for Real

I was introduced to the bookmakers in the mid 70's when I found a club in eastern New Orleans that would let a slightly underaged, decently mannered kid sit at the bar with the hotshots. It contained a motley crew of characters, every one a gambler, and we all had action going on every contest imaginable. That place is now a donut shop but I think the lovely lady that was behind the bar is still there, serving donuts instead of doubles.

Eventually I turned into a punk kid: just out of high school, renovating houses, hanging with the ole farts and those wonderful barmaids. I got in some binds rolling dice and playing poker and pool, and I quickly learned that the safest and smartest bet was with the books. My first passion was football and I was always ready to lose a portion of my paycheck for the thrill of watching the games that I had money on. By the time I got my act together and enrolled at the University of New Orleans I found that my wallet was much lighter and the money I was losing was jeopardizing my college career. As far as gambling was concerned, I had to start winning or quit altogether.

The Birth of GRIDLINE

I made up my mind that I needed a system to keep from losing too much money. After all, a monkey can predict 50% of the games correctly, and he uses a system that gaurantees random selections on every game. That was a lot better than I was doing because I was still betting on the Raiders every week. I went the way a lot of bettors do when they feel the need for a system and I started comparing the points fonts scored and points allowed of the teams involved. Sure enough, it proved more effective than my own intuition and before long I found that I was hitting just as frequently as that monkey was. Fifty percent.

I soon realized that the point system was flawed. I was hitting one and missing the next, and each winning week had a corresponding losing week. I was doing a little better but I was still forking over the vig. It became obvious that the point system did not apply when a good running team came across a poor run defense, for example. It seemed to me that deeper analysis was needed.

I was currently registered in several math courses at the university and I was exposed to the algebra of matrices and the properties of determinants. I can't go into detail about the proofs, but I managed to come up with several equations that reflected how much a team relied on its rushing game, how much it relied on its passing game and the point production that could be expected when faced with a particular opponent. When the series of equations reduced nicely to a single, compact and symmetrical form, I knew I was on to something. To this day that same equation is at the heart of our system. Like the first Super Bowl, the system had not been named yet but sometime around 1980 GRIDLINE was born.

The Evolution of GRIDLINE

Each week I sat down with a notebook and a pencil and I dutifully copied the rushing, passing and point statistics from the Times-Picayune sports section, and broke it down into game averages and plugged the factors into my equation. After about ten hours of work I had projected icons scores for every game, on a neutral field that is, and then I added the oddsmakers 3 - 3 1/2 pt home field advantage before I placed the bets. The system wasn't hitting much better than 50% at that time, but I did notice that when my projected score was much different from the oddsmaker's that I was hitting at a higher rate. I decided that as long as that was the case I would stick with the system.

In the early 80's UNO opened its exquisite Computer Center on campus. It housed a monster IBM mainframe and it had a fantastic, efficient staff. I registered for Comp Sci courses with the idea of automating my system and eliminating the hours involved in those grueling weekly sessions. I coded the system in an experimental language called Pascal at a dumb terminal in the CC and each week I would create the input on punch cards and submit them to the CC staff to be run in batch mode overnite. The next day I would pick up one of those 11X16 computer paper printouts with the weekend's projected scores on it. I still had to create the punch cards each week, but I had eliminated the manual calculations and reduced the time it took to produce the scores from 10 hours to about 4 hours. If the manual version was GRIDLINE 1.0, you could say this batch system was GRIDLINE version 2.0. I utilized that system for one season until my parents gave me a fantastic piece of technology as a birthday present.

Adding machines were common equipment in the early 80's but some cutting edge technological companies had begun to produce something called the hand-held calculator. My model was a Radio Shack Scientific Programmable Calculator. The unit featured logarythms and exponential functions, summation calculations, integration capability and host of other mathematical functions. It was the 'programmable' feature that made my life easier. After reviewing the booklet I soon had my algorythm built into the calculator, and I had programmed it to prompt me for the statistical input. After punching in the numbers the LCD would display 10 black boxes for 5 seconds then it would post the score. GRIDLINE v3.0 only required about 2 hours of data entry per week and I wasn't losing any more money. In fact I had my best year ever, at least as far as pro football was concerned.

I went with the calculater for two good seasons and in my last year at UNO the accounting department secured a number of standalone units called personal computers. They ran on something called the DOS operating system and each came equiped with a fantastic program called the Lotus Spreadsheet Application. I was only allowed to work off of a 5 1/2" floppy but it was plenty enough to code the spreadsheet system and I could format the printouts in box score fashion. I carried those printouts with me throughout the week and soon my friends and some of those ole farts at the club began asking me for copies.

What's in a Name?

I was more than willing to distribute the printouts but I grew weary of being asked for "da computer sheet wit da scores". There was a neat local, sports newsletter that was being put out called "Gridweek". I thought the name was cool, the term gridiron was just coming into fashion but this was the first time I had seen the 'grid-' prefix used as 'of or relating to football'. I stole the grid part, and since my system was based on beating the point spread, or the latest line, I tacked on 'line'. Maybe it still sounds a little goofy, but technically GRIDLINE means "Related to football and the point spread". Nowadays, grid lines abound on several automated drafting free 3d models and graphics computer applications but that simply wasn't the case back in 1983. And the relatively recent practice of using the compounded form 'gridline' to refer to those invisible lines came way after the original GRIDLINE. I had it first.

Gridline in the Private Sector

I couldn't delay my graduation from college any longer so after I got my mathematics degree I ended up taking a job as a programmer with a small, locally owned software firm. I was part of a young and talented team of analysts that took on huge software development projects for huge oil corporations. I was separated from UNO's accounting building so I had to code the next version of Gridline in Basic on the company's DEC Alpha machine. A friend of mine with the firm coded the input routines for me, and each week we would sit down and knock out the data entry in about an hour. Over the years it had become much easier to produce the GRIDLINE picks - I guess the Basic version would be GRIDLINE 5.0 - but I had not made any major changes to the original algorythm or the body of data that fueled it. To the best of my knowledge, I was winning money in the pros but I really couldn't be sure what my hit percentage was and I know I was still making some irresponsible betting decisions. I was still in my twenties, after all.

I used the Basic Gridline for a couple of seasons, then around 1987 three significant software breakthroughs made it feasible to transform the Gridline program into the Gridline system. The relational database was a theoretical concept that was invented in the seventies but it didn't take off until it was combined with the B-tree database server in the mid eighties. This combination of software technologies made it possible to access physical records in two disk reads, rather than the several reads required using sequential or binary searches. The third breakthrough was the advent of the "C" programming language. Back then you didn't buy a relational database, you coded it, and you didn't want to try coding a relational database management system in Basic. We designed our own RDBMS and we coded it in "C" to remain competitive as a software company. I designed the GRIDLINE 6.0 system utilizing the company's RDBMS to remain competitive as a bettor.

GRIDLINE 6.0 was a menu driven beauty. It ran on a state of the art 286 AT compatible multi-user Unix based system. It had data entry screens for defining the league and the teams, and a screen that served the duel purpose of accepting data from game box scores and displaying the GRIDLINE forecast. I duplicated the forecast printout from the Basic version and the first major addition I made to the system output was the Forecast History report. Its the same format that you see in the "Forecast Season to Date" selection on the site. For the first time I knew exactly what the system's hit percentage was. Over 54% for all bets, and 60% when GRIDLINE's point spread differed by 6 points or more from the oddsmaker's line. I marked those in the printouts with little BB's, GRIDLINE's Best Bets.

It was easy now to extract information from the GRIDLINE database so I decided to use this to my advantage. The first thing I wanted to know was how the point spread impacted the betting outcomes. I was still connected to the old farts at the clubs, and they all had theories about "home dogs" and "smart money" and "the team that wins straight up at home loses to the spread in the next game on the road." I had a full season of box scores and point spreads so I developed the "Team Records vs Spread Report" (yes, the same one you see on the site). That report not only showed me what teams were beating the point spread, but in 1988 I learned that the oddsmakers knew what they were doing. I found that favorites were hitting just as much as underdogs, road teams just as much as home teams, and road favorites just as much as home underdogs. There's no easy ride.

Tweaking the System (or how I learned to quit worrying and love the HFA)

I had a good season in '88 but the record vs spread report showed me that I was taking too many underdogs. If favorites were winning as much as dogs I should be betting favorites half of the time. I reviewed that data and found that as the point spread went up my projections tended to level off. I couldn't cover high point spreads. I wasn't about to alter the algorythm but I did the next best thing. I adjusted the team averages about the league norm before I plugged the data into the formula. Nowadays when you hear me talking about the "zoom" factor this is what I am referring to. The GRIDLINE system zooms in 25 percent around the league norm before we calculate the scores. Take a team that rushes for 140 yards/game, if the league average is 100 we'll adjust, or zoom in on the number, by upping it to 150 yards. It won't affect the projected score if teams are evenly matched but if one is deficient it will up our theoretical spread and we can lay more points.

There was still lots of room for improvement, and I was still using the oddsmaker's 3 - 3 1/2 point home field advantage but I felt that value was kind of arbitrary. I had 2 seasons of real data to work with so I went about determining a better HFA. I coded programs that adjusted all of the projections with different HFA's and compared the resulting hit percentages. I let them run in the office over the weekend, replaying each season with experimental HFA's and popping file systems off a RAM disk to keep from filling the company's hard drive. Then I came in early Monday and ripped the paper from the dot matrix network printer. It was a simple printout that had a single number in the top left corner, 2.5.

The zoom factor and experimental HFA didn't have a dramatic effect on my hit percentage, but it did allow me to lower the Best Bets spread differential from 6 points to 4 and that created several more bets that I could make with a reasonable expectation that they would hit at a 60% rate. To this day we still use the 4 point spread differential to identify our Best Bets.

Changing Environment

By 1989 4gl's became the rage and the GRIDLINE 7.0 system was coded in a development environment that allowed it to be ported to different hardware and operating systems. It was a virtual carbon copy of the "C" version and as a Systems Analyst I used it to evaluate CPU and disk access speeds for several different platforms. The company ran a Novell network now with Windows running on top of DOS, but we still had our conventional computer systems in house and we were connected to a fortune 500 company's IBM mainframe. There were probably about a dozen different instances of GRIDLINE running on a half-dozen platforms by the time that little company disbanded. I know most of our in-house disks were wiped time and again but I bet that fortune 500 company still has a fully functional version of GRIDLINE collecting dust somewhere in the clusters of an old IBM mainframe.

In the early 90's I went to work for a software design and civil engineering firm that did contracts for the public sector. The 4gl version of GRIDLINE proved just as portable across workplaces as it did across platforms. For a brief period it ran on the local government's IBM AS/400 computer but for the most part I reined it in and kept just a couple of copies in DOS and UNIX within the office.

I didn't own a personal computer until 1992, when I purchased an ACER 486 that ran Windows on top of DOS. The home system allowed me to spend more time creating reports and analyzing past performance. In quick fashion I was able to code the Power Ratings, the Team by Team performance report, Series History and the Teams Schedule Strength report.

How Strong is That?

The Schedule Strength report was a minor breakthrough because while I had seen schedule strength listings before, they all reflected the teams won-loss record with the previous opponents records. This is fundamentally incorrect, since a listing sorted by the teams won/loss record would tend to show weaker schedules for more successful teams, and harder schedules for the weaker teams. GRIDLINE's schedule strength report subtracts the team's record from the previous opponents, and the listing reduces to the opponents record against the rest of the league. As a bonus, I created a data element called the performance factor which is simply the team's current record times the previous opponents record, and I used it to project the team's final record by multiplying the record of the future opponents by that same perfomance factor.

I was still looking for ways to improve the system's hit percentage. In 1994 I felt that 65% among the Best Bets was an attainable goal and I still feel that way today. It was going to be hard to pump up the performance now that GRIDLINE was winning, but the Schedule Strength factor seemed like a logical candidate to incorporate into the system. In the same manner that I implemented the Zoom factor, I began adjusting each teams vital statistics about the norm relative to their schedule strength. It didn't make a great or immediate impact on my projections but it did keep me from getting suprised by teams that looked good on paper due to a weak schedule.

You Can Go Home Again

My hit percentage was now up over 60% for the Best Bets but it had not reached the 65% goal - the point where I would consider the system a success. For over 15 years I had tweaked and prodded the system but I was fresh out of new ideas. After serious soul searching I decided to revisit the home field advantage, and in 1995 I made the last adjustment to the system as we know it today.

The 2 1/2 point HFA that I was using is a good number if you are using a power ratings based system, but ours is not a power ratings system, its a mathematical model. In reality, the home field advantage is different for each team. In other words, it might mean more to the Titans than it does to the Bengals. Historically, I had better results adjusting data BEFORE I ran it through the algorythm, so I opted to take the same approach to determining a more realistic and mathematically sound HFA. Rather than simply appending points to the computer projection after the calculations had run, I decided to logically enhance the performance of the home team and dampen the performance of the visitor. I used the same drill I had used before when I computed the HFA constant, only this time I was looking for a percentile constant that would affect home teams and visitors with respect to their previous performance. It took some doing but I finally came up with a percentage that closely mirrored the actual HFA of the NFL. Unfortunately, I can't share this value because it is Top Secret ... it doesn't even echo when I type it into the GRIDLINE system.

That did it! In 1996 I posted my best NFL season ever. The GRIDLINE 9.0 system yielded over 60 computer generated Best Bets and it hit 67% of them! It was also the year that a single week's forecast went 100%, 13 out of 13 ATS, the first time I had ever seen anyone project a full weekend's slate perfectly. I didn't take every team that weekend but it was my biggest scoring week ever, still is, and it would have been even better had I taken the time to fill out one of those silly 10 team parlay cards.

Can't Leave Well-Enough Alone

For over 16 years I had been tweaking the system. Despite the success I had in 96 I was still determined to make adjustments. Revising the system had become an unhealthy obsession.

In 1997 I focused on series histories and I allowed past games to influence the algorythm's output. Bad mistake. Even where the history is overwhelmingly lopsided it doesn't mean the next game will follow suit. GRIDLINE 10.0 was a bust.

In 1998 I made an extremely radical departure from the way I determined each teams vital statistics. This was, and is, actually a very promising theory, and I can't reveal details because I am still testing it, but at the time it did not produce as good results as the standard averaging method. I tabled GRIDLINE 11.0, then I lost it entirely in a hard drive accident. In 1999 I returned exclusively to GRIDLINE 9.0 and immediately began chalking up high hit ratios again.

GRIDLINE 2000 the Internet Calls

Since the mid 80's I had been printing the system output and distributing it to interested parties. In early 90's I was seriously contemplating publishing a newsletter. I had hoped to sell advertising to casinos that were starting to pop up around the city and distribute the issues to bars and news stands. I even went so far as to produce a prototype. I scrapped the idea when I realized how much work was involved, and how much work I had already invested. I am a computer analyst, after all, not a magazine editor.

When the internet starting growing into the mass communication monster that we know today I knew it was the proper medium for GRIDLINE. I had to overcome my natural inertia, but it was just a matter of securing some server space and learning HTML. Securing server space was easy, you just open your wallet. Learning HTML turned out to be tougher than I expected.

A lot of software engineers have a snooty habit of referring to HTML as a 'glorified word processor', and I had seen webmasters on the evening news who said that it was easy, "You simply find a page you like, copy it and change the text and graphics to your liking." Let me say that I made the mistake of going in with this attitude and it cost me at least two weeks of frustration before I forced myself to do the right thing. Learn HTML. The web is a great place to learn how to code in HTML and there are several sites that provide invaluable tag definition lists and special features. For the actual tutoring process I would refer anyone who wants to learn HTML to Dave Kristula's website,, and select "HTML: An Interactive Tutorial for Beginners." He's a talented kid with a fantastic and humorous approach not only to web mastering but to life in general. Stop by his site and see what I mean.

What the Future Holds

Nobody really knows what the future holds. Heck, our algorythm only hits Sixty-something percent of its best bets, but we have some definite plans for the next version of the GRIDLINE software, and we have some tentative plans regarding the website.

Recall that GRIDLINE 10.0 and GRIDLINE 11.0 were busts and we are currently using a DOS, character based version of GRIDLINE 9.0. At present, I am working on GRIDLINE 12.0, what I hope will be the mother of all computer based projection systems for football games. The system is being developed in MS Access to take advantage of GUI features that should allow us to mix and match the games we use to determine the statistical input for the GRIDLINE formula. In other words, we may opt to run the games using stats culled from the two team's common opponents and/or games from the series history. Another feature of the future GRIDLINE software will be the ability to assemble statistics based on multiple data collection methods. The vast majority of game predicting computer systems, even our current version of GRIDLINE, are based on the use of per game averages to determine input for the algorythms. Those of us with mathematical or statistical backgrounds know there are several ways to determine a viable set of statistics. I am currently experimenting with with different collection methods, but it is a long and arduous process to determine viability. Expect GRIDLINE v12.0 to come into production sometime before 2005, while the millennium is still new.

As for the website, I really can't promise it will be around forever. One thing I've learned is that our traffic is based on our hit percentage, and if we find ourselves in the throes of an extended losing streak the site could turn into a ghost town on the information highway. Right now, we are simply focused on attracting more clients. The website is a relatively low cost operation but I would like to see it generate some revenue. If it does we can think about extending our expertise to college football and possibly other sports. One thing is for certain, we are committed to keeping access to this site and all of our official projections free of charge and we will continue to resist the temptaion to overload the site with irritating banners and hard sell tactics.

There you have it. An extensive history of the evolution of GRIDLINE along with a peek into the future and a promise of our commitment to excellence and free access. I'll update the Story of GRIDLINE every ten years or so, but for now its a wrap. Adios and good luck everyone.