Last Week In Review
As expected, the impact of resting starters caused a major blip in the prediction tool's results during Week 17. Since the tool's algorithms
do not anticipate such a strategy, Mr. Computer saw the Redskins favored by 9 over the Cowboys and thought Christmas wasn't over quite yet! Hopefully the
odd looking predictions made that fact obvious. Handicapping
Week 17 games will always have its challenges, and it's certainly not a time when historical analysis applies well across all game scenarios.
During Week 17, the predictions at default settings went 5-2 predicting winners at the 50%+ odds to win threshold in games that did not feature teams
that already had clinched playoff berths; they went 2-6 in games that did. Similarly, against the spread, default settings predictions went 2-1 at the 60%+ odds
to cover threshold in non-playoff-team games, but 1-5 in the others. All the wacky spreads resulted in a multitude of bold predictions to cover by Mr. Computer,
who literally could not believe his eyes...
Those outliers aside, Week 17 capped a phenomenal first "live" year for our beta stage predictions tool. Since its launch before Week 3 through Week 16 (208 games), predictions
at default settings went 144-64 (69.2%) picking straight up winners. Over the same period of time, this result outperformed all of our benchmarks: Vegas spreads, the Yahoo!
Pick'em user community consensus, the ESPN experts, and the Yahoo! Sports experts, to name a few. We have already heard from several people who won their pick'em pools
this year using the NFL GameZone predictions, and we couldn't be happier about that.
The tool's results were just as impressive ATS, with default settings predictions going 50-31 (61.7%) against closing Vegas spreads at the 60%+ odds to cover
threshold through Week 16. (Even if you hold the obvious outliers of Week 17 against it, the tool was still 59% ATS at default settings for the season.) The fact that
the tool achieved this ATS record over 80+ picks is very encouraging. Although the result set is still not large enough to reach any long term conclusions regarding performance, the combination
of the high number of 60%+ games and outstanding prediction performance meant that any bettors using the NFL GameZone analysis -- even just at default settings -- would have reaped handsome profits.
Even by wagering a unit that was only 1-2% of one's total bankroll -- a rule that smart bettors stick by -- blindly betting our 60%+ odds to cover games would have generated returns on total bankroll in the 15-30%
range in only 14 weeks, with measured risk. This equates to a prorated annual return of 55-110%. That's a lot better than the stock market or real estate, especially this year...
All of this was achieved with a tool that we consider perhaps 70% "baked," and that currently has no knowledge of injuries, weather, or bye weeks. We have many improvements still to make, and we
can't wait for next year. In addition, we are confident that our individual users are beginning to learn how to customize the tool's settings for specific game situations, enabling them to outperform even
the tool's own default settings analysis. And nothing says we can't build a service like this for other sports too.
Have any ideas on improvements we can make to the NFL GameZone, or
where we should be spending our next efforts? Please
let us know.
In closing, if you enjoyed using the NFL GameZone this year, and especially if you profited from it, a donation of any size would be greatly appreciated. While other handicappers charge
hundreds if not thousands of dollars for their advice and picks, we've kept the GameZone free all season. Both in terms of time and money, it costs us to keep our web site up, running, and
updated. We toyed with the idea of charging for the NFL Playoffs to help cover some of our costs, but decided against it.
NFL Playoffs Wild Card Games Preview
The Nerds are back this week -- apologies for the end of regular season disappearance. The prediction tool's analysis of
the 2007 NFL Wild Card matchups show mostly contrarian expectations; as of this writing, underdogs were favored to win two of the Wild Card games and cover
in three of them. As this is playoff season, it behooves us to examine each game in more depth, especially to try to identify some
of the less measurable factors that may significantly impact outcomes. Some examples:
- Week 16/17 performance skew. Teams that had made the playoffs already and were resting starters largely underperformed at the end of the season, and
these results are reflected in their season stats. While you can never be 100% certain how much of this underperformance is specifically attributable
to resting starters, because the prediction tool is driven by data, the power rankings of several leading teams were dragged down, and vice versa.
- Weather. The ultimate analysis destroyer, especially in the winter months. It takes a ton of historical data to be able to design a meaningful
data-driven system to forecast the precise effect of weather on a game outcome. Not only are there many factors involved (temperature, wind speed, precipitation, field conditions, etc.),
the chance that weather conditions at game time will be the same as forecast is by no means a certainty.
- Injuries. Especially in football, where so much position performance is not independently determined (e.g. the effect of an offensive line on a running back's stats), forecasting the
expected effect of a given injury (i.e., its value in relative points) is again an amazingly complex problem from a data standpoint.
- Bye weeks. There are a lot of people who quote "trends" about how teams with a bye week in the playoffs win a high percentage of their opening games against teams
that played the week before. While the prediction tool unfortunately can't help provide a better answer -- days of rest is a factor we will add next year -- just be
careful about the "correlation vs. causation" trap, and also the "single lens analysis" trap. For example, teams who get playoff byes are usually better teams than
their Wild Card brethren. So are they winning those games more because they are better teams, or more because of the bye week? It's not easy to parse out the relative impacts.
- Playing style matchups. The prediction tool's approach values actual results (points, wins, etc.) over more subjective evaluation of how a given team's offensive and
defensive schemes may play against their opponent's. Given the variation in play calling and schemes that occurs from game to game, this level of analysis tends to be too
subjective and data-poor for our tastes, although we intend to look into it more in the future.
- Individual player / player group matchups. Another highly subjective and risky approach due to the multitude of dependencies involved in football. On paper, Team A's awesome receivers look like they
should dominate Team B's secondary -- but they won't if Team B's defensive line dominates Team A's pass protection. Or if Randy Moss and Wes Welker wake up with the Ebola Virus. This type of break-'em-down analysis may sound great on
sports radio, and it's definitely entertaining. But given the complexities and interdependencies, the tool favors a more abstract approach.
- Emotion, morale, revenge, and other "soft" factors. Totally unpredictable to all but Tuesday Morning Quarterbacks. However, if you are chained to an electric chair and forced to bet a game you consider a total toss up,
your gut feel in this area may be a slightly better guide than just flipping a coin. Maybe.
With all this said, remember that there are only 11 games in the playoffs, so anything can happen. The default settings predictions could have a great run or
they could underperform just flipping a coin. But the long term value of sticking by a system like the GameZone tool is what matters most.
Have a great playoffs and don't forget about our content in other sports, specifically college hoops, which should be ramping up in the weeks to come.
NFL Playoff Wild Card Games Predictions Summary
Below is our Team Rankings GameZone predictions summary for all
NFL Week 18 games. Although we update this page
periodically, predictions can change multiple times per day based on
new information such as shifting NFL betting lines. Click on
"View Predictions" to see the most recent lines
and predictions and to customize factor settings.
 Washington Redskins at  Seattle Seahawks
|
View Predictions |
|
Based on a point spread of Seattle -3.5, analysis at default settings
shows Washington as the favorite to win the game, with win odds of 60.5% and an expected
margin of 2.8 points. Similar games analysis indicates Washington as the strong favorite
to cover the Vegas point spread with 76.7% cover odds.
Notes from the Nerds: What a fantastic game to try to break down. After a disppointing end to their regular season, the Seahawks return to the
friendly confines of Qwest field (#2 home field advantage power rating) to take on the red-hot Redskins, who in addition to stomping Dallas
at home last week, won their two previous games on the road against good teams. The Redskins appear generally healthy and have outperformed expectations under backup
QB Todd Collins in their last four games. While far from a proven talent, Collins has been consistent in the few games he has played. For the Seahawks, both starting QB Matt Hasselback and RB Shaun Alexander are tending to
minor injuries; Alexander seems to be the greater injury risk, but he hasn't been a huge factor this year. The forecast is for rain, and if the field gets wet enough,
it may give a slight advantage to the Redskins' slightly better running game (117 rushing ypg vs. 102 for Seattle); the Skins also yield fewer yards per game to the run.
To account for the playoff situation, we adjusted the importance of
the "week of game" factor to the maximum level, and also did some sensitivity analysis on the power rating, distance traveled, and penalties factors, given how Seattle's fans
and stadium tend to manufacture an inordinately high number of false start penalties on opponents. The short answer is, while certain scenarios pulled Seattle up to as much as
very very very slight favorites to win the game, none favored the Seahawks to cover a -3.5 spread in this situation. While the Hawks are indeed 7-1 at home this season, they also did not
have a strong home schedule -- the best opponent they beat at Qwest was Tampa Bay, currently ranked #15 in our overall power ratings. The Skins have a significantly higher power rating. All told, if Todd Collins can maintain his consistency, we don't see much in terms of fudge factors that would erode our support of the data-driven predictions for this game.
|
 Jacksonville Jaguars at  Pittsburgh Steelers
|
View Predictions |
|
Based on a point spread of Jacksonville -2.5, analysis at default settings
shows Jacksonville as the very slight favorite to win the game, with win odds of 55.5% and an expected
margin of 1.8 points. Similar games analysis indicates Jacksonville as the toss up favorite
to cover the Vegas point spread with 54.2% cover odds.
Notes from the Nerds: These two teams met in Pittsburgh less than a month ago, and the Steelers were favored by three; although the Jags pulled off a seven point upset, the
lines currently only favor them by 2.5. That may appear to scream opportunity, but there is all sorts of crazy information out there on this game. One note that stuck out: the Steelers have never been beaten twice in a season
by the same opponent in their 75 year history. All streaks must come to an end, but that does sound a bit...well...imposing. The Steelers also have much more playoff experience (and success) under their belts than
Jacksonville does. That being said, the Steelers are pretty banged up (although it appears safety Polamalu will
play) and will be missing top rusher Willie Parker; replacement Najeh Davenport has definitely yet to prove himself.
Weather-wise, given both teams' strong rushing offenses and defenses, a classic Steel City winter game in the muck,
should it arrive (40% chance of rain last we checked), will likely not give any relative advantage to the Steelers, as the December game against the Jags demonstrated. Finally, both teams rested starters in Week 17, and their late season
performance excluding those games doesn't reveal too much in terms of relative momentum. Besides the Pittsburgh win, Jacksonville beat up on a couple of poor opponents, while Pittsburgh underperformed expectations at New England. Slight momentum
advantage to the Jags.
The most important component of our analysis, the GameZone situational tool, indicates that the Vegas spread looks pretty darn efficient at -2.5. Sensitivity analysis on factor settings largely suggest an even closer outcome. In addition, even factoring in the prior loss
at home to the Jags, Pittsburgh's home power rating is about equal to Jacksonville's away power rating, and one game result is not a sound basis for a future prediction. Man, this is a tough one.
Personally, we would lay off it. But the tool has proven effective (~58% ATS) even at the 55% threshold at default settings, so if forced, we'd have to buck history and go with Jags to win and cover, with significantly more confidence in the former than the latter.
|
 New York Giants at  Tampa Bay Buccaneers
|
View Predictions |
|
Based on a point spread of Tampa Bay -3, analysis at default settings
shows New York as the very slight favorite to win the game, with win odds of 52.9% and an expected
margin of 0.6 points. Similar games analysis indicates New York as the favorite
to cover the Vegas point spread with 61.9% cover odds.
Notes from the Nerds: The tool's results are highly sensitive to power ratings in this game, and Tampa Bay's underperformance
at the end of the season, coupled with the Giants' impressive showing against the Pats in Week 17, affected those numbers. As as result, especially
considering that the Giants took a fair amount of health damage in their last regular season game, we'd probably consider the Bucs as slight
favorites to win this game. (52.9% win odds, given margin of error in the tool's calculations, cannot be mathematically interpreted as a plus for
the Giants.) Spread-wise, however, most angles show the Giants barely covering +3; again, decrease the importance of power ratings even one notch
and you will see the impact it has. As a result, we would consider the calculations a weak endorsement of the Giants to cover at full strength. If you
think their injury status is significant, may be wise to lay off. Weather does not seem like it will be a factor.
|
 Tennessee Titans at  San Diego Chargers
|
View Predictions |
|
Based on a point spread of San Diego -10, analysis at default settings
shows San Diego as the heavy favorite to win the game, with win odds of 76.6% and an expected
margin of 9.2 points. Similar games analysis indicates Tennessee as the slight favorite
to cover the Vegas point spread with 57.1% cover odds.
Notes from the Nerds: Since their early season slump ended, the Chargers have been pretty nasty at home; even though their win over the Colts was downright ugly,
the lines still expected them to lose that game by 3.5 points. With Vince Young now looking like he will start, the Titans should at least preserve continuity at the QB spot,
although they also are batting some injuries to other contributors such as Albert Haynesworth and LenDale White. The injury issue is always tough to gauge, especially in the playoffs
where supposedly hurting players seem to magically heal come game time. With a 30% chance of rain last we checked, we don't see much that would change the prediction of the
Chargers as strong favorites to win, while the Tennessee injury situation and San Diego home advantage would lead us to downgrade the prediction of Tennessee as slight
favorites to cover. We're certainly not comfortable with all the injury uncertainties.
|
|
New England Patriots
San Diego Chargers at 
Green Bay Packers
Seattle Seahawks at 
Indianapolis Colts
Dallas Cowboys
New England Patriots at 
New York Giants
San Francisco 49ers at 
Cleveland Browns
Cincinnati Bengals at 
Miami Dolphins
Buffalo Bills at 
Philadelphia Eagles
Dallas Cowboys at 
Washington Redskins
Houston Texans
Detroit Lions at 
Carolina Panthers at 
New Orleans Saints at 
Chicago Bears
Atlanta Falcons
Pittsburgh Steelers at 
Baltimore Ravens
Oakland Raiders
St. Louis Rams at 
Arizona Cardinals
Minnesota Vikings at 
Denver Broncos
Kansas City Chiefs at 
New York Jets
St. Louis Rams
Carolina Panthers
Cleveland Browns at 
Cincinnati Bengals
Detroit Lions
Miami Dolphins at 
Philadelphia Eagles at 
New Orleans Saints
Buffalo Bills
Green Bay Packers at 
Oakland Raiders at 
Jacksonville Jaguars
Minnesota Vikings
Houston Texans at 
Atlanta Falcons at 
New York Jets at 
Tennessee Titans
Baltimore Ravens at 
Tampa Bay Buccaneers at 
San Francisco 49ers
Denver Broncos at 
Arizona Cardinals at 
Kansas City Chiefs
Indianapolis Colts at 
Chicago Bears at