Day of Reckoning

In honor of it being Super Bowl Sunday, I thought I'd take a look back at the few articles concerning NFL predictions that were posted here at LomHenn.com this past season.

How well did everyone do? Let's find out!

First off, I went after Michael Lombardi for his three stats that will define the Super Bowl teams. Lombardi's first prediction:

I can promise you with a 99.0 percent degree of certainty that the final four teams in the playoffs will be ranked in the top eight of point differential in the first half.

This year's final four playoff teams were Philadelphia, Arizona, Pittsburgh, and Baltimore. I responded to Lombardi's prediction at the time by showing that this prediction was wrong for the 2007 season. Also, why would first half scoring differential be more important than overall scoring differential? I then predicted that the Super Bowl participants would be in the top eight of overall scoring differential.

In 2008, Philadelphia (at #2, +82 in points) and Baltimore (#6, +58) were in the top eight in first half scoring differential. Pittsburgh was tied for 11th (+34), while Arizona was...28th (-53!!!). So Lombardi's first prediction--which he promised with "99.0 percent" certainty--was only 50% accurate.

For the overall scoring differential, Baltimore (tied for #1, +141), Philadelphia (#4, +127), and Pittsburgh (#5, +124) all ended up in the top eight. Arizona was 18th (+1). My "prediction"--that the two Super Bowl teams would be in the top eight--was 1 for 2, also 50% (if I had phrased my prediction like Lombardi had--saying that the final four teams would be in the top eight--I would have had that pick 75% accurate).

In any case, I think this supports my original point that first-half scoring differential is no more a meaningful predictor of playoff success than overall scoring differential. Arizona is a good case in point--the Cardinals were blown out in the first half of a couple of their games this season, which made their first-half scoring differential much worse than their overall scoring differential.

Let's see if Lombardi fared any better for his second can't miss Super Bowl prediction:

Throw the ball in the first half -- often

Lombardi's rationale here is that teams that establish the passing game in the first half are more likely to score in the first half, and therefore have a higher first-half scoring differential. Besides the obvious issue of Lombardi's second prediction being related to his first, my problem with his assertion is that it doesn't really tell you anything. Using 2007 statistics (since the predictions were made at the beginning of this season), 29 of 32 teams passed more than ran in the first half, so there's really no meat to his prediction.

Finding 2008 statistics regarding running/passing plays by half is like trying to find someone who still thinks Carrot Top is funny. After a couple hours of searching the Internet, the best I could compile were season totals for each team. Seven teams (out of 32) ran the ball more than they passed this season:

Atlanta (56% run/44% pass)
NY Giants (51%/49%)
Carolina (55%/45%)
Baltimore (58%/42%)
Minnesota (53%/47%)
Tennessee (53%/47%)
Oakland (52%/48%)

An interesting list. Discounting Oakland, the other six teams are all playoff teams, including Baltimore, who made it to the AFC Championship game running the ball 58% of the time. The remaining six playoff teams:

Indianapolis (39% run/61% pass)
Miami (48%/52%)
Whale's Vagina (47%/53%)
Philadelphia (41%/59%)
Pittsburgh (48%/52%)
Arizona (35%/65%)

So of the Super Bowl teams, Arizona was pretty pass-wacky, while Pittsburgh was more balanced. However, the league average was 46% run/54% pass, so most of the playoff teams (9 out of 12) ran the ball more often than the league average. Again, it's hard to analyze this directly without the first half statistics, but based on game statistics, I don't think Lombardi's prediction was valid.

What about Lombardi's third "stat that will define winning teams"?

Seven yards per passing attempt

The two teams that find their way down to Tampa for the Super Bowl will have above a 7.25 average per attempt passing for the season.

As I noted in the original article, 11 out of 32 quarterbacks were able to accomplish that goal in 2007. This season, only 10 QBs were able to do it. As far as the two teams in Tampa for the Super Bowl, Arizona averaged 7.42 yards per attempt--Pittsburgh averaged 6.5 (if we go by quarterbacks, Arizona's Kurt Warner averaged 7.66 YPA and Ben Roethlisberger averaged 7.04 YPA. Since I don't know if Lombardi was referring to individual or team stats on this, I'll include both).

Bottom line--sorry Hans, wrong guess! Would you like to go for double jeopardy, where the scores can really change?

I think it's pretty easy to say that Lombardi went 0-for-3 on his predictions.

From the same post, I also made some predictions. How did I do?

The Super Bowl participants will each be in the top 8 in the NFL in scoring differential.

See above. 3-out-of-4 for the teams in the championship games; 1-of-2 for the Super Bowl. For the final four teams, better than Lombardi; for the Super Bowl itself, the same as Lombardi.

Each Super Bowl team will be in the top 14 in offensive yards per game.

Arizona=4th (365.8 yards per game). Pittsburgh=22nd (311.9 ypg). Another 50% success, though Pittsburgh was pretty far down on the list.

Each Super Bowl team will have jerseys that don't read "Kansas City Chiefs."

Nailed that one!!

The takeaway here is that I did just as well as an NFL "expert" in making picks based on arbitrary criteria. My point is not to say how great a job I did in making picks; rather, it's to show how ridiculous it is to make these kind of picks at all. Year after year pundits make their picks, and year after year most of the picks are flat out wrong. And no one goes back and tells them how wrong they are.

Next up, an October prediction Zinglebert made concerning the Colts and Titans and the AFC South race (in an article critiquing the one and only Bob Kravitz):

I’m going to put on my Nostradumbass hat on and predict that the law of averages catches up with the Titans and they will come into Lucas Oil Stadium in Week 17 with a 10-5 record and either the division title or a playoff berth on the line.

Tennessee was 13-2 coming in, and the game was meaningless because both the Colts and Titans had clinched all they could clinch. If the Colts had won one more game early in the season, perhaps the division title would have been decided in the regular season finale. Instead, it was a chance to see Jim Sorgi in action!

To be fair, I had my own bad Colts prediction:

I'll make a prediction now: if the Colts win 10 games, they are definitely in the playoffs.

The Colts went 12-4 and easily made the playoffs. However, New England went 11-5 and stayed home. Obviously, 10-6 wouldn't have been good enough to get in. So another bad pick by me.

Finally, I'll wrap up with this little nugget: some of us here at LomHenn.com participate in our own prediction contest at the outset of each NFL season (we do this despite knowing how lousy our picks will be). We pick each of the division winners and wild cards, then take a shot at the playoffs. Zinglebert took the time to compare our picks to the "experts" at SI.com and ESPN.com to see how we did. Here is the result (correct picks are in white, incorrect picks in red):


Click on the picture to see it more clearly.

The column on the right is the total number of division winner/wild card picks that were correct. Although I'm proud of the fact that I got 6 out of the 12 possible picks correct, what I find more interesting is that of the 25 "experts" making picks, none had more than 5 picks correct. And absolutely NO ONE (including myself) had the Super Bowl matchup right, or even close (though I have to point out that both Dr. Z of SI.com and Mike Sando of ESPN.com had Philadelphia going to the Super Bowl, so at least those two writers had Philly advancing in the playoffs--unlike the rest of us. Those picks are denoted in grey in the picture).

So does my success this season make me an expert? Far from it. My point is that come next year, you have as much of a chance of picking the teams who will play in the Super Bowl as the people you read online or see on TV do. And unless you actually put your picks online, no one will hold you accountable if you pick poorly--just like no one holds accountable the "experts" who get all of their picks wrong year after year. You'll be in good company.

Since I've had soooo much success this season picking games, here's my prediction for this year's Super Bowl:

Arizona 33, Pittsburgh 31 (3 OT). The game will be decided on a safety with less than 2:00 to go in the third overtime.

Yeah, that could happen!

Prediction: Two Teams Will Play In the Super Bowl

If you ever pay attention to most preseason Super Bowl predictions, you'd find that they are generally wrong. Yes, a person can usually figure out which teams will be competitive, but pundits who predict Super Bowl participants in September are just guessing.

Michael Lombardi of SI.com has taken a different approach. Instead of picking the teams, he'll tell us what stats will be important factors for the Super Bowl participants.

Frankly Football: Three stats that will define the Super Bowl teams

My guess: wins, losses, and jerseys that don't read "Kansas City Chiefs."

There is nothing better than the start of "real football," but with the beginning of the NFL season comes the usual predictions of which teams will get to the Super Bowl. And with all due respect to my colleagues here at SI.com, 99.9 percent of the time those predictions are wrong.

(I included this part to show that Lombardi's article starts out reasonably. Then, it all goes horribly wrong...)

I am sure there are some MIT students who can craft the right formula for the Super Bowl participants...

Yeah: wins, losses, and jerseys that don't read "Kansas City Chiefs."

...but for me it's way too hard. Yet I do know several key statistics that are essential to the makeup of a Super Bowl-caliber team:

Hit me.

1. Have a successful first-half point differential

So...have a halftime lead? Last year, ALL teams were 182-53 when leading at the half, a 77.4 wining percentage. Since 1998, NFL teams leading at halftime have a 74.7 winning pct.

Lombardi is saying that teams that outscore their opponents in the first half during the season will be successful. In general, I agree with this idea, but what if a team leads 44-0 at halftime of one game, but is down 7-6 at halftime of the other 15 games? That team's +29 differential probably won't be that meaningful later.

I can promise you with a 99.0 percent degree of certainty that the final four teams in the playoffs will be ranked in the top eight of point differential in the first half.

99% degree of certainty? That means this criterion will only be wrong 1 out of 100 times, theoretically.

For example, last year the Patriots lead the league in first-half point differential with 196. That means the Patriots, on average, went into each halftime with a lead of roughly 12 points.

The Patriots were one of the final four playoff teams last year. Check.

Rounding out the Top 10 in this category were the San Diego Chargers at 104,

Also a final four team. Two for two.

the Colts at 102, Tampa Bay at 76, Washington at 62, Pittsburgh at 59,

All playoff teams, but only the Colts got past the Wild Card round...and lost in the Divisional.

Green Bay at 56,

Check.

Seattle at 55, Jacksonville at 49 and the Cowboys at 23, all playoff teams last season. (The Giants had a plus-two point differential last season.)

You're "99% degree of certainty" doesn't even hold up for last year? The Giants won the Super Bowl, but were only +2 in first half scoring differential last year. But, despite this, you're saying that it's 99% certain that the final four teams will be in the top 8??? Talk about arbitrary. The bottom line is that good teams typically lead at halftime and they typically lead at the end of the game, too. First half point differential is simply a by-product--you could have just as easily used "teams who lead more often than trail at halftime."

Now, what is so important about halftime leads? Well, it forces the opponent to play a near-perfect second half. It also requires the defensive play-caller to not make one mistake, lest he limit his ability to be creative in attacking the passer. Calling a defensive game is a challenge in itself, but calling it from behind is very taxing. Every third down is crucial as your team needs to get the ball back to close the gap. So if the defensive coach makes one mistake -- like calling a blitz and giving up a big play -- then the 10-point deficit might turn into a 17-point deficit, essentially putting the game out of reach. Forcing a defense to play a cautious and conservative game is what most offenses thrive on to be successful.

Does it really? It is possible that teams who are trailing will take more chances on defense--for example, trying to force a turnover--than play conservatively?

And aren't third downs also crucial in close games, or in games your team is leading?

On average there are slightly more than 13 third-down situations in every game. Teams that convert above 45 percent of their third downs are considered excellent. Every week coaches put tremendous time and energy into breaking down the third-down tendencies of their opponent's defense and learning how their foes plan their exotic blitzes on those downs.

But when a team falls behind and is concerned about not allowing the big play on third down, this reduces their blitzing and makes the game much easier on the quarterback. This is to not imply that if you get a lead, your team will never see any blitzes, but it does reduce the amount of blitzes and makes their predictability much easier to decode.

Any data to back this up? Again, I disagree wholeheartedly. I think that teams that are behind tend to take more chances on defense--that is, blitz more--than teams in the lead. Case in point: teams in the lead using a prevent defense late in the game. In trying to protect the lead and not give up a big play, the defense backs off into coverage (and allows a disciplined offense to march down the field).

Of course, I don't have any hard data for that assertion, either. But I don't claim to be making any predictions.

2. Throw the ball in the first half -- often

Last year only three teams (Minnesota, Oakland and Jacksonville) ran more than they passed in the first half. Most of the playoff teams averaged a 44 percent run to 56 percent pass ratio before intermission. Seattle, Green Bay, Indy, N.E., Dallas, Pittsburgh and the Giants came out trying to establish the pass, therefore most of them had a positive halftime point differential. San Diego and LaDainian Tomlinson had a 48 percent run and 52 percent pass ratio.

This point might as well be "use 11 players on offense -- often." Lombardi admits that 29 of the 32 teams in the league passed more often than ran in the first half. So how does this differentiate the playoff teams? He uses the ambiguous "most playoff teams" had a 44/56 run/pass split to try to make his point. But then he shows that San Diego had a 48/52 split, and he already mentioned Jacksonville running more often than passing. Since he doesn't include how he came up with his averages, there is no way to break down the numbers. However, since 2 of his 12 playoff teams don't come close to following his assertion (including a team that made the AFC Championship game), I'm calling bullshit here.

3. Seven yards per passing attempt

The two teams that find their way down to Tampa for the Super Bowl will have above a 7.25 average per attempt passing for the season.

11 of 32 starting quarterbacks did this last year. Only one of the QBs in the Super Bowl did it.

Last year, Eli Manning played on a team that led the NFL in dropped passes, so his stats in this area were a little off, but of the 25 top-rated passers in yards per attempt, 17 had above 7.00. Completions are nice, they look good on the stat sheet and it's nice to have a high percentage, but making yards matters most.

But Eli was still 28th in yards per attempt. 28th! Even with the dropped passes, I don't think Lombardi's point is a good one. Let's see about the 4 Conference Championship participants:

NYG: Eli Manning, 28th
GB: Brett Favre, 5th
NE: Tom Brady, 1st
SD: Philip Rivers, 18th

Half of those four quarterbacks were in the bottom half of the league in yards per attempt. How about the quarterbacks for the other playoff teams?

IND: Peyton Manning, 3rd
JAC: David Garrard, 7th
PIT: Ben Roethlisberger, 4th
TEN: Vince Young, 22nd
DAL: Tony Romo, 2nd
SEA: Matt Hasselbeck, 14th
TB: Jeff Garcia, 10th
WAS: Jason Campbell, 24th

7 of the 12 playoff teams were in the top 12 in yards per attempt. While I agree with Lombardi that yards per attempt is a good stat for evaluating QB performance, it's easy to see that the stat is not a reliable predictor of who will make the Super Bowl.

The bottom line is that predictions made at this point of the season are mostly garbage. How can anyone predict injuries? How many people predicted Tom Tom would go down in week one? Do you think that might affect some people's predictions for the season?

I fully expect San Diego and Seattle to achieve high rankings in all three areas and I believe they have the key elements of teams that reach the Super Bowl.

The biggest problem I have with this article is that these stats are too general to make any accurate predictions from. I can define three arbitrary statistics and give general stats in order to make a Super Bowl prediction, too. All this will prove is that if you're generic enough, no one can tell you that you are wrong...even when you are.

Here are my arbitrary stat predictions:

1. The Super Bowl participants will each be in the top 8 in the NFL in scoring differential.

2. Each Super Bowl team will be in the top 14 in offensive yards per game.

3. Each Super Bowl team will have jerseys that don't read "Kansas City Chiefs."