East Coast Offense
By Christopher Liss
RotoWire Managing Editor
Probability and Prediction
I'm still in a handicapping slump against the spread, and even my once untouchable Fantasy Football Live Guru Picks have gone south, leaving me with just a 10-point lead over Brandon Funston. (This after I mercilessly trash-talked Funston and Brad Evans for 10 weeks). So what's my problem? Have I just hit a streak of bad luck, or am I missing something that I'd better become aware of quickly while there are still five weeks left?
Let's take a look at one of the games I got wrong. In Week 11, I took the Bears +3.5 in Green Bay. The Packers won 37-3. This wasn't the type of loss I could blame on a fumbled snap, or a botched chip-shot field goal. But what about the idea that the game played in the 99th percentile of possible outcomes favoring the Packers that day, and that if it played in the 50th percentile, the Bears might have lost by three and covered the spread? Looked at that way, my pick wasn't wrong - the game just played out in an extreme and uncommon way. (Or as I've written in the past: the game got itself wrong.
Maybe the way I'm picking is just fine, and I'm getting a combination of bad bounces (close/unlucky losses) and then extreme outcomes that happen from time to time. (That about covers every way one can lose against the spread - either narrowly or via blowout).
I could make a similar argument when I get a Survivor pick wrong. Let's say I picked a 14-point favorite to win outright and gave them a 92 percent chance to win. (That's very high for an NFL game). If I'm wrong, and they lose, I can say that the eight percent came in that day. After all, if you were on an airplane, and the pilot said: "Ladies and gentlemen, there are some problems with the plane's engine, but don't worry, there is only an eight-percent chance of a wreck", admit it, you'd probably crap your pants. (Also because the pilot is violating his training and likely the law by saying that and has quite obviously gone insane, so you'd suspect it was more like 75 percent, but let's assume he was telling the truth). The point is when we make a prediction about an NFL game, we know there's a realistic chance that we'll get it wrong, even in the easiest case when we're talking about a double-digit favorite in Survivor.
When handicapping games against the spread, each one is supposedly 50/50, but there's always that 8-10 percent chance that we go 4-12 or 5-11 in a given week. So in that case, how do you separate bad luck from bad handicapping?
I think the key here is the notion of probability. What we're saying when we put a percentage on a Bears-Packers prediction is if these two teams were to play a million times in the exact same conditions, this is the percentage of times Chicago would cover the spread. But that's ridiculous because those conditions will only obtain once, and you have to make a prediction for the outcome this time around. In other words, you're not just handicapping the average outcome of a million simulations, but predicting whether the actual outcome will be an outlier in the Packers' favor, and also to what extent. Probability's not going to help you much because you're predicting the result of a singular event. Even if you come up with some basis for assigning probability (games between those teams in the past, recent performance by both teams, games between divisional rivals with similar records generally, etc.), you're just estimating what the approximate average game between the two would be, and that's as of last week. It says nothing about how much this week's game will deviate from that average.
So, bottom line, am I doing a bad job, or just getting unlucky? Of course, I have no idea, but I prefer to think that I'm doing a bad job. Because if you retreat into the probability shell, then you can never take credit when things go well.
Norv Turner is Broomcorn's Uncle
When Turner opted for a game-tying field-goal attempt on 4th and 2 with a minute and half left Sunday night, he made the wrong call. The Chargers should instead have gone for the first down.
Even so, Turner did what a majority of coaches would likely have done because the chances of Nate Kaeding making a 47-yard field goal in good conditions is probably about 75 percent, while the chances of making the first down is probably closer to 60. So Turner's risk of losing immediately was significantly greater if he went for it. But what about the reward? When Kaeding made the field goal, the Chargers merely tied the game and gave the ball back to the Peyton Manning with plenty of time on the clock. Had San Diego made the first down, they payoff would have been far greater because they'd be in field-goal range, they'd have a decent chance for a game-winning touchdown, and they'd leave no time for the Colts to have another possession.
In an effort to play it "safe", Turner chose a far more dangerous course - like a poker player, who fearing to risk his money in a pot, winds up anteing himself to death.
Things to Take Away From Week 12
Surviving Week 13
Last week the Steelers won easily, and so did the Bucs, despite a first-quarter hiccup.
This week, we'd take Dallas at home against the Seahawks if you have the Cowboys available. Otherwise, it's a tougher call as I think the Jets-Denver game will be close, and I don't like Tennessee on the road in a nationally televised game on a short week - there's more volatility in a game like that, and when you're backing the favorite, you don't want volatility. Let's go Cowboys first, Tennessee second and the Jets third. We give Dallas an 82 percent chance to win this game. We reserve the right to change our minds before the full article comes out Wednesday night.
Article first appeared 11/26/08