The Ted Baxter System

Ah, yes, I remember it well. It was a cold Saturday, the 11th of January, 1975. I had been living in Western Pennsylvania for almost a year.

On that night, all of Western Pennsylvania was excited: our Pittsburgh Steelers had at last reached the Super Bowl! The next day in New Orleans, we were going to meet the Minnesota Vikings for the National Football League championship.

When I tuned in that week's episode of The Mary Tyler Moore Show, I discovered that football fever had infected even this CBS comedy.

Since gambling is of questionable morality, this is the only way that a proper sitcom episode can end. However, considering that the fictional WJM-TV was in Minneapolis, I was puzzled that the writers didn't have Lou place his bet on the hometown Vikings. Maybe the episode had to be filmed before the NFC team had been decided.

Over the closing credits, Mary Tyler Moore voiced an apology to Pittsburgh fans for depicting our team as losers. In real life the next day, the Steelers actually defeated the Vikings to win the first of their four Super Bowl championships. [Update: make that six.]

This episode was reportedly based on the real-life experience of its co-writer, Stan Daniels. That suggests the obvious question: Does the Ted Baxter System work in real life? And if it does work, is the magic number — we'll call it BN, the Baxter Number — really 11 points? Or might some other point spread give better results?

The following season, I tried to find out (without actually gambling, of course). For a few weeks, I kept a record of the Las Vegas odds, "The Latest Line," which I generally obtained from the Wednesday newspaper. Then I compared these point spreads to the actual outcomes of the games.

The preliminary results were favorable, but the experiment became tedious and I soon abandoned it.

This year, armed with a modern computer and a spreadsheet program to make the calculations easier, I decided to repeat the experiment, continuing it for a full season this time. Now that the final regular-season game of 2002 has been played, I can share the results with you.

The short answer is yes, the Ted Baxter System does work! Using a Baxter Number of 11 points would have been successful in 2002. However, a better choice would have been either 2½ points or 13½ points, depending on your strategy. I'll explain the details below.

But first, I should point out that over the 28 years since Ted Baxter invented his system, the National Football League has changed. They've added six expansion teams and a salary cap. No longer are the best players concentrated in a few powerhouse franchises; instead, they're scattered around the league. This competitive balance is called "parity." The NFL likes parity because it means that almost every team has a chance to win.

However, many fans don't like it. Here's what a columnist wrote in the Pittsburgh Post-Gazette on November 30, 2002:

My data for the 2002 season bears this out. Let me show you, using charts and graphs. Paul Shaffer, do we have some theme music for this segment?

Suppose that there are two bettors: Franz, who always bets on the Favorite, and Ulf, who always bets on the Underdog. Also suppose that there are four games, A, B, C, and D, and that the spread on each game is three points.

The horizontal scale represents the point spread: half a point, one point, and so on. Our four games A, B, C, and D are represented by blue diamonds at the horizontal position "3," for three points.

The vertical scale represents the actual results of the games: by how many points did the favored team win?

In game B, the favorite won by three points, exactly the spread. This is a "push." No one won or lost; Franz and Ulf each got their money back. The red line, called the "push line," connects all the places on the chart where the spread and the result are the same.

In game C, the favorite won, but by only one point. Since this was less than the spread, Franz lost his bet, and Ulf won.

In game D, the underdog upset the favorite by four points. On the chart, this is shown as an result of -4, meaning that the favorite's "winning" margin was a negative four points. Again Franz lost his bet, and Ulf won.

You can see that Ulf, who always bets on the underdog, will win every game that's below the push line. Franz, who always bets on the favorite, will win only those games that are above the push line. Since there's more area below the line than above, it would seem that Ulf has the advantage — if the games are scattered evenly across the chart.

But we wouldn't expect them to be scattered all over the chart, now, would we? The betting public ought to know enough about football to establish point spreads that are somewhere close to the actual outcomes of the games.

The blue diamonds should cluster around the push line, some of them a few points above it and others a few points below.

Therefore, we would expect a chart for a full season to look something like this, where the result of every game is within a touchdown of the point spread.

However, the actual chart for the 2002 season looks like this. There's only a slight correlation between the spreads and the results.

September started with a string of upsets: in 18 of the first 31 games played, the underdog won outright.

For the season as a whole, Ulf's underdogs won 55% of their games against the spread (depicted by the diamonds below the red line). They won 37% outright (the diamonds below the black line).

This chart is a picture of parity. You might as well flip a coin.

A January 2006 update:

Jay Kornegay, the executive director of the Las Vegas Hilton sports book, notes that "the novice bettor, the casual fan, almost always bets on the better team." As a result, NFL favorites get more support than they deserve, the point spread gets pushed up, and in the long run the favorites don't cover the spread. "Ulf" would be a winner. In 16 seasons from 1989 to 2004, only twice did the favorites cover more times than not, and in those two years the figures were barely above 50 percent.

However, in the 2005 pro football season, "the favorites covered the point spread an incredible 63 percent of the time," according to a column by Jim Armstrong of the Denver Post. Kornegay remarked, "I've been in this business 18 years, and it's the most I've ever seen."

Meanwhile, "freakfish" on a sports journalism message board offers the corresponding wisdom for the NCAA basketball tournament: "Take the points in the first round, the favorites in the second, and spend the winnings."

So much for the charts, which we've used to visualize the fact that NFL games are highly unpredictable. Now let's get to our graphs, which we'll use to determine one or more optimal Baxter Numbers.

Suppose that Ted chooses a BN — on the show it was 11 points — and for every game where the point spread is BN or more, he will "take the points" and wager $10 on the underdog. (No money changes hands until after the game is played. If Ted wins, he gets $10, but if he loses, he owes the bookmaker $11. The extra dollar is how the bookie makes his living.) However, if the point spread is less than BN, Ted will not bet on either team.

Using the spreadsheet, I can analyze my 2002 data for all the different possible values of BN and see which would have worked best.

First, I count the winning and losing bets for various Baxter Numbers.

As BN increases from left to right, bets are placed on fewer games, so the numbers of wins and losses both decline. But notice that the blue line is almost always above the red.

The exception is at BN=8, where the underdogs won 18 and lost 19 against the spread. Ted would have won $10x18 = $180 and lost $11x19 = $209. His "profit" (actually a net loss) would have been -$29.

At BN=8½, he would have had a winning record, barely, as the underdogs were 16-15. But because wins are worth only $10 and losses cost $11, he would still have ended up in the hole by -$5.

For example, Ulf's technique — always bet the underdog — corresponds to a Baxter Number of ½, which would have resulted in a record of 131-106 and a net profit of $144.

An even better Baxter Number was 2½, where the underdogs had a record against the spread of 119-91. Ted would have won $10x119 = $1190 and lost $11x91 = $1001, for a net profit of $189.

This graph shows that BN=2½ would have yielded the biggest profit.

Surprisingly, however, that BN did not correspond to the largest winning percentage.

All of the small Baxter Numbers, from ½ through 7, had winning percentages around 56% (give or take 1½%). With a small BN, you place a lot of bets and make your profit on volume.

But if you're looking for a bigger winning percentage, you need a BN greater than 11. You won't bet very often, but when you do, you'll win at least two-thirds of the time.

For example, at BN=11½, Ted would have made only six bets all season, winning four of them or 67%. Of those six games, five were in the first five weeks. After October 6, the wagering public apparently recognized the truth of parity; there were few heavily favored teams, and Ted would have found only one more game to bet on for the rest of the year.

Taking it to the extreme, 13½ was the best BN percentagewise this year, with a 100% success rate. Ted would have wagered on only three games, but by backing the underdog, he would have won all three bets:

Date

Favorite

Spread

Underdog

Underdog Result

9/29

Philadelphia

20

Houston

Lost by only 18

10/6

Indianapolis

13½

Cincinnati

Lost by only 7

12/8

Pittsburgh

13½

Houston

Won

However, this strategy wouldn't be much fun. For weeks on end, Ted would have had no stake, no interest, in what happened on Sunday. A real gambler like Lou Grant wants to have "action" every week — preferably every game. He can't stand not being involved.

Additionally, a real gambler wants to feel that he's outwitting the other bettors by being smarter at predicting the final score.

To use the Baxter system, you don't need to know anything more about football than Ted did himself. It's a mechanical way of exploiting the statistical probabilities, something like putting your money in a mutual fund.

The Baxter system can work for you, but only because your competitors (the other bettors) have the hubris to believe that they know something.

They think that they can predict the results of an increasingly random and unpredictable sport.