Fool’s Errand

This page should provide you with all the background you need to understand how we make our weekly picks here at Thoughtful Fool Orbiting Headquarters.  Broadly speaking we look for situations where the public perception of a team’s quality relative to their opponents does not jive with measured reality.  We can easily determine public perception simply by looking at the point spreads and money lines. 

Defense-adjusted Value Over Average

In the world of football, statistical analysis of the type that has been going on in baseball for decades has only been around in the past few years. It is centered around a website called Football Outsiders (FO), and even though they expend a Herculean effort on this each year, the movement is still in its infancy.

The guys at FO have, among other things, developed a statistical measure called DVOA (Defense-adjusted Value Over Average). You can read about it in detail here, but what it involves is looking at every play of every game and evaluating how well it succeeds relative to the league average for the given situation. Weightings are applied and a few other adjustments are made, and eventually the result is the team's DVOA, a single number which can be thought of as a measure of their performance relative to other teams.

DVOA is cold. It does not care who your favorite team. It does not show preference if Team A has a receiver who is a complete idiot or Team B has a QB who is a paragon. It is not influenced by what John Madden said just before half-time or what Peter King wrote in his last column. It is an unemotional evaluation of what actually happened during the games.

Later in the season, usually about Week 9, there is enough of a sample set that the DVOA can be weighted such that the more recent games in the season have a stronger affect. This measure is called adjusted DVOA and I use that instead of standard DVOA, once it is available.

The Point Spread

Contrast that to odds set by the sports books. These are heavily dependent on public opinion.

A good place for a sports book to be regarding the spread is where there are an equal amount of bets on either side. To bet on the point spread in Vegas, you typically have to bet 110 to win 100. So I place $110 as my wager, if I win I get my $110 back, plus my winnings of $100. If I lose, I lose the entire $110.

In a hypothetical situation where there is exactly one bet on either side of the spread, the sports book comes out up $10. It has taken 2 bets, both for $110, so has taken in $220. The winner of the two bets gets a payout of $210 (the original $110 wager + the $100 winnings), the sports book pockets the remaining $10.   Extrapolating from this, we see that the sports book can assure itself revenue if it keeps an equal amount of money on either side of the spread. They do this by adjusting the spread over time, although they are generally sharp enough in setting the spread that adjustments are usually very small.

The important point is that the spread is only indirectly related to the quality of the teams.  It is directly related to the public perception of the quality of the teams.  For my analysis, I typically grab the spread from one of the online services on late Tuesday or early Wednesday.

DVOA vs. the Spread

A glib way of looking at this is that DVOA measures the reality of team performance and the point spread measure the perception of team performance. It is my pet delusion that in games where we find a significant disconnect between the DVOAs of the two teams and point spread, we have a gambling opportunity. Each week I provide summary spreadsheet comparing this data.  Note: I include raw DVOA numbers and also Home Adjusted DVOA, where in the home team gets a 16.5 point DVOA boost.

Category 1: We look for games in which the team with the superior DVOA is the underdog. For example, let's say Team A has a DVOA of 19.8 and Team B has a DVOA of -15.6, but the point spread is Team B, -3.5 over Team A. In these games we are likely to pick the underdog as we believe the objectively better team is the one getting points instead of giving them.

Category 2: We look for games where there is a substantial DVOA difference between the two teams, but the point spread is small. For example, let's say Team A has a DVOA of 49.8 and Team B has a DVOA of 8.6, but the point spread is Team A, -2.5 over Team B. DVOA suggests that Team A is markedly superior to Team B, yet they are giving Team B less that three points. In this case we are likely to pick the favorite, believing that the DVOA measure suggests that they can win by more than 3 points.

Category 3: We look for games where the DVOA differential is small, but the point spread is inordinately large. For example, let's say Team A has a DVOA of 29.8 and Team B has a DVOA of 26.5, but the point spread is Team A,-9.5 over Team B. DVOA suggests that Team A is not all that much better than Team B, yet they are giving them a significant number of points. In this case, we are likely to pick the underdog since DVOA suggests that, while they may not win, they are a good shot from covering the spread (losing by less than 9.5 points, in this case).

What constitutes "substantial" differences and "significant" point spreads is subjective. In fact, there are things that DVOA does not take into account. Injuries will only be taken into account as they affect the outcome of games and get folded into DVOA. A first string QB going down in the fourth quarter of a game will have virtually no effect on DVOA rating, but could potentially have a large affect on the next game. DVOA also doesn't take into account the warm-weather team in a cold-weather stadium effect, as described here.  The spreadsheet analyzes both raw DVOA and DVOA with a home team bonus.  (In measuring DVOA home teams do perform better on average, but it is highly variable from season to season and team to team so we are careful not to accept it blindly.)  But in general, we try to be very reticent in using judgment to override our methods since our judgment is quite probably no better than the public opinion we are trying to outsmart.

As a final rule, I do not attempt to use this system until Week 8 at the soonest. Any statistical system needs a solid dataset to be accurate and I have selected Week 8 as the point where the dataset has enough data to be valid, but is still early enough in the season that I can have some fun. After three years in the black, last year (2006) was the first time I came out behind on point spread picks.

The Money Line

As of 2005 I began making money line predictions. In some ways, money line bets are much simpler than betting the point spread. The bet is on whether a team wins or loses straight up. The adjustment for team strength is made in the payouts.  As with point spreads, I grab money lines from one of the online services on Tuesday night or Wednesday morning.

Example: Team A -142 v. Team B +132. 

The money line on Team A is -142. The minus means they are the favorite, so you must wager $1.42 to win $1.00. Let's round up to hundreds so I don't have to type decimal points. So that means I give the nice man behind the counter in the sports book $142. If Team A wins, he gives me my $142 back, plus an extra $100. For convenience sake let’s normalize all bets to $100, so I give the man $100 for a chance to win $70.42.

The Team B is +132. The plus means they are the underdog. (Don't ask me why plus is underdog and minus is favorite, it's one of life's mysteries.) That means you wager $1.00 to win $1.32. I give the nice man $100 on Team B and if Team B wins he gives me back my c-note plus and extra $132.

So the key to understanding the money lines is that the number next to the favorite will have a minus sign and that means that is how much you must risk to win $100, and the number next to the underdog will have a plus sign, that tells you how much you win if you bet $100.

DVOA vs. The Money Line

Like spread picks we’ll have a spreadsheet every week with the money line analysis. (Note: the "Return 100" column in the spreadsheet indicates the amount that you would win for a hundred dollar bet.)

Before we make picks we need some rules. These rules are from my gut, but my intention with money line picks is to allow no judgment in the process at all, no matter how strongly I feel about mitigating factors. The money line picks are something that I can simply program into Excel and have it spit out the picks.  After I have a couple of years under my belt the formula might be refined.

Rule A: Always pick underdogs (line is a "plus'') that have a superior DVOA. Similar to our Category 1 picks against the spread, but in this case we believe the better team has the superior payoff.

Rule B(1): Never pick a team that has a payoff less than $50. We'll just draw an arbitrary line and say betting $100 to win less than $50 is not worth the trouble.

Rule B(2): Always pick the favorite when the payoff plus the DVOA difference is 90 or greater. This is a linear way to assure that in situations where the teams are close in capability (a small DVOA difference) there is enough of a payoff to justify the risk of picking them. This rule implies that we will take any favorite where the payoff is $90 or higher. As the payoff drops from $90 we need a stronger DVOA difference (more confidence that the favorite is better). Thus, as the payoff drops to $80, we are going to need at least a 10 point DVOA difference to justify the bet. At $70 we need 20, at $60we need 30, at $50 we need 40. We'll call 90 the Favorite Benchmark and incorporate it into the spreadsheet so we can change it easily in the future if necessary.

Rule C: When the DVOA inferior team has a payoff greater than 30 times the DVOA difference, we will pick the underdog. Essentially we are saying that a payoff that high justifies the risk. 30 is the arbitrarily select factor we will use here, so we'll call it the Dog Factor in the spreadsheet in case we want to change it in the future. So, for a team that is 10 points inferior in DVOA, we need a money line of at least +300.

In three years of making money line picks they have never let me down over the course of a season and, in fact, have rarely let me down week to week.

Simple, Eh?

OK, maybe not. But it's actually simpler once you look at the spreadsheets and follow it through for week or two. And you don’t have to worry about anything because I’ll prepare the spreadsheets and summarize the picks each week.  For the record, I do not wager actual money each week on this.  Only sometimes.  I look on this more as an experiment; a way on which I might stumble on to some piece of information that gives me a bit of an edge. The fact is I'd watch football, and occasionally bet on it, anyway, so why not see if being thoughtful can help me be a little less foolish.