THE BOOK--Playing The Percentages In Baseball

I guess they figure an average QB replacing him will complete better than 58% considering the offensive line and he’s throwing to Randy Moss.

Or that the running game will be asked to do more, and is capable of handling the extra workload. (or both)

Still, if Brady is only worth 1.2 wins over 16 games that’s the equivalent of a baseball player worth 12.  Like Barry Bonds 2001-2004.

"Still, if Brady is only worth 1.2 wins over 16 games that’s the equivalent of a baseball player worth 12.  Like Barry Bonds 2001-2004.”

I forgot about that. Silly me, thinking 1 win in a football season is worth as much as 1 win in a baseball season.

"Still, if Brady is only worth 1.2 wins over 16 games that’s the equivalent of a baseball player worth 12.  Like Barry Bonds 2001-2004.”

Right.  I figured he’s worth about 35% more, W%-wise, than Pujols (who I pegged at +9 wins).

1.2 definitely SEEMS low, but I’m not convinced that it’s incorrect.

The average completion percentage in the NFL last year was actually 61.2% the last two years, and 59.5% the year before.

You’d really want to use yards per attempt, though.  That average is roughly 6.4.  Replacement level is probably 5.0.  Brady’s career average is 7.2, so he’s 2.2 yards per attempt above replacement.  Say 35 attempts per game, that’s 77 yards per game.

You also have interceptions to account for; his career average is an interception percentage of 2.4%, and the league average is in the low 3s.  Call replacement level 5; he’s .91 interceptions per game better than replacement.  That’s basically -3 yards per game, so he’s now a total of 80 yards per game above replacement.

How many points is that?  I don’t know.

I suppose in football, the average team is some 7 or 8 WAR.  And I guess their payroll is, what, 100MM per team?  That’s 12-13MM$ per win.  So, a guy that is 1.2 WAR would make about 15MM$ a year.

That about how much top QBs make?

Just saw Hawk’s post.  It would seem that if we add up all the WAR for each position, we’re going to end up at more than 8 WAR for each team.

I’m not opposed to this necessarily, because the linearity of it make exceeding the boundaries we know possible.  But, it doesn’t make me feel good about it.

If the top-end salary is about 15% of team payroll, then the top-end WAR should be 15% of team WAR.  If you have 8 WAR per team, that’s 1.2 WAR max for the best player.

So, while we may be mathematically correct with all our gyrations about replacement, in the real world, it seems that the betting guy may have a better handle on this.

I think calling anyone in NFL more than a 1.5 WAR player to be really getting into a philosophical issue.

"You’d really want to use yards per attempt, though.  That average is roughly 6.4.  Replacement level is probably 5.0.  Brady’s career average is 7.2, so he’s 2.2 yards per attempt above replacement.  Say 35 attempts per game, that’s 77 yards per game.”

Can you do it like this though?  If you threw a replacement level QB into the New England offense, you’d expect him to average more than 5.0 yards per attempt, wouldn’t you?

I don’t think Tom Brady is being replaced by a replacement level quarterback, at least not in the way we use replacement level in baseball - there is no large, fungible source of quarterbacks out there for the Patriots to turn to.

Instead, what we’re referring to is Tom Brady’s value above Cassell, or more generically the value of Brady over a #2 quarterback. Those aren’t free - realistically the Patriots have invested a lot more into Cassell than an MLB team invests into a replacement level player. (This is especially true at quarterback, where you’re expected to know the offensive scheme very well. A guy who can play shortstop in Scranton-Wilkes Barre can play shortstop in Fenway. A guy who knows how to play quarterback in general is not necessarily ready to run the Patriot’s particular offensive scheme.)

I guess this is where we find out how much of a quarterback’s perceived value is actually him, and how much is the rest of the team.  The offensive line (combined) makes, what, 3 or 4 times as much as Brady?  Wouldn’t it be reasonable to assume that (as a WAG) at least a third of Brady’s above-average performance is due to the line?

Just a thought.

Can you do it like this though?  If you threw a replacement level QB into the New England offense, you’d expect him to average more than 5.0 yards per attempt, wouldn’t you?

Good point.  The answer:  I have no idea.  How much of YPA is QB, how much is receivers, how much is the blocking?

Instead, what we’re referring to is Tom Brady’s value above Cassell, or more generically the value of Brady over a #2 quarterback. Those aren’t free - realistically the Patriots have invested a lot more into Cassell than an MLB team invests into a replacement level player.

I don’t know if that’s true.  Cassell hasn’t started a game since high school, so the demand for his services is pretty low relative to the demand for some other #2 QBs.  I’m not sure how that all ends up comparing to a AAA player.

What’s with post #2?

FWIW, someone pointed out a slight error in my math.  That combined with the line shifting a bit, has us down to 0.96 wins.

I looked at all seasons where a QB had at least a 100 passer rating on football ref, and compared the record for those QB with the record of their backups in the games they missed.

The starters combined to go 414-108, a .793 percentage or 12.7 wins per year.  The backups? .627 (42-25), 10 wins per year.

If that represents the value of a great QB, then they are twice as valuable, relative to the schedule, as an MLB alltime great season.  But only half as valuable as an alltime great NBA season - a Jordan/Chamberlain/O’Neal at his best can be worth 25 or so wins in 82 games.

Brian Burke generally knows what he’s doing.  Keep in mind this is based on actual results, not necessarily true-talent level.

http://www.advancednflstats.com/2007/12/week-15-qb-ratings-adjusted-for-pass.html

Last year Brady was about 2.5 wins above average, with Peyton Manning at about 1.3.  (The previous year Manning was more in the 2.5 to 3.0 range I think.) The 30th best QB was about -1.5 wins per 16 games, while the next five take it down to -2.0 wins.  If you arbitrarily regress Brady to +2 wins and put replacement-level at -1.5 wins, that’s a 3.5 win swing.  Of course, Cassell could be better than replacement level (or worse—he hasn’t really played much since high school).

Rally/13: Ideally, you would take the top QB with the passer rating in year X, and look at the team performance with him in year X+1.  This will take care of the regression issue.

Tango/15: That’s impossible to do, though. NFL teams change so much from year to year there’s no way you can compare a QB who is dependent on the 5 linemen, a tight end, and several receivers to help him out. Just look at Brady in 06 versus 07. He’s always been a very good QB, but he didn’t suddenly morph into the best ever in one season-- he got Randy Moss and Wes Welker to throw to whereas he had mediocre guys there before.

Using the last 5 seasons of data, and based on a linear regression using passing and running efficiencies, interception rates, fumble rates, and penalties, each additional yard of offensive passing efficiency means (about) +1.43 additional wins.

Every % point increase in interception rate means (about) -.54 wins.

The NFL average offensive passing efficiency (net, which means including sack yards) is 6.0 net yds per att. Brady’s in 2007 was 7.8 net yds per att. This would equate to +2.5 wins above average.

The NFL average int rate is 3.2%, and Brady’s was 1.5%. This would equate to another +0.9 wins above average, for a total of +3.4 wins above avg for his 2007, arguably the best year for a QB ever.

One question is, how far would we expect Brady to regress this year? The second question is, what is Cassel’s/other NE replacement QB’s expected ability?

Brady’s career net efficiency and int rate are 6.5 yd/att and 2.4%, for a career average of +1.1 wins abv avg. Just guessing, say he would have regressed 2/3 toward his career mean this year. That would give him a +1.9 expected wins abv avg for ‘08.

A .9-win drop from Vegas means that bettors are expecting Cassel/whoever is going have substantially above average numbers this year, enough for +1.0 win based on the passing game alone. This is effectively saying: assuming NE’s running, defense, and penalties are totally average, their passing game could provide a 9-7 record.

With NE’s receivers and line, I think that’s reasonable. Cassel’s only thrown 57 attempts in his career, most of them without Moss and Welker to throw to, so his 5.9 yd/att and 3.5% int rate could be expected to improve. But now I’m just getting into conjecture.

Using the last 5 seasons of data, and based on a linear regression using passing and running efficiencies, interception rates, fumble rates, and penalties, each additional yard of offensive passing efficiency means (about) +1.43 additional wins.

By additional yard, you mean YPA, right?

Anyone ever do this kind of analysis for Mike Vick?  His team had a good winning % when he was the QB but was terrible when he didn’t. 

His ypa was slightly above average, his sacks were bad but his running of course was awesome.  I’ve always thought of his running game as throwing himself screen passes the way Drew Brees dumps it off to Reggie Bush all the time to inflate his accuracy %.  Vick’s accuracy % was bad but was it really when you consider that he was essentially completing passes to himself 8 times a game.  When you look at his numbers, he threw about 25 times a game, ran 8 times and got sacked 2.5 or 3 times.  I forget the exact numbers.  So you can see that he what he lost in pass attempts, he made up for with running attempts

It’s interesting though if you look at the record with and without Vick over the last 7 seasons.

They were 38-28-1 when Vick started for a 57% winning %.  They have been 13-32 when he hasn’t started the last 7 seasons for a 29% winning %.  Seems like your theoretical replacement player right there. 

In the five seasons he was the starter from 02 through 06, they were 37-27-1 when he started for a 58% win % and 3-12 when he didn’t for a 20% win %. 

I dunno.  I’ll leave it to the math wizards on here to figure out how the numbers explain the difference between the Falcons with and without Vick.

Colin-yes.

Interesting about Vick. He was so unconventional, it’s hard to compare him to anyone. A little off topic, but you’re right. Vick was basically his own dump off pass. That’s why his comp% was so low. Whenever he did throw, he threw relatively deep passes. He consistently ranked at or near the top of yards per completion, mostly because he rarely threw the dink and dunk stuff.

Have any of you actually read the VW post?  This is one of the worst misinterpretations of betting odds I’ve ever seen.

If you haven’t loaded the Matchbook page for NFL over/unders, go do so now.  There is literally zero trading activity, on the Pats or any other team.  No one can possibly glean anything useful from sportsbook odds when there is no liquidity.  With no bets, whose opinions are reflected in the lines?  Nobody’s.

If some small-cap stock has a bid price of $1/share and an ask of $100/share, that doesn’t make the market price $50.50 simply because that number splits the difference.

If that example isn’t clear enough for you, imagine that McDonald’s decides to change the price of their double cheeseburger from $1 to $100.  Has its value gone up to $50.50 in response?  Pretty soon the double cheeseburger will be more valuable than Brady.