Wraith

Two improvements immediately jump to mind: 1) Make the threshold a fixed percentage of drive length instead of a fixed number. Something like 12.5% (1/8th - 10 yards of an 80 yard drive or 5 yards of a 40 yard drive). Thus, a very short but very crucial yardage on a short drive could be included. 2) Normalize your results by number of drives or possession time. I would use drives. This would eliminate the possibility of the quarterback being punished because the defense cannot get off the field.

I can't believe I'm defending Holcomb's Arm here, but he has not committed the egregious "sins" that some have pinned on him. In fact, some of the claims are outright false (the claim above that the t-distribution was incorrectly applied, for example). Using statistical judgement to place a threshold is incredibly common in the real world and is usually necessary. The fact that tweaking the threshold in this test gives different results merely shows that the test is not sensitive enough to pick up a shift in Losman's production, or that there is no meaningful shift in Losman's production. It certainly does not mean that HA is deliberatly misleading with his statistics. Just as HA alluded to above, judgement calls are a part of statistics, so much so that there have been many tools devised to aid in that judgement (the whole topic of confidence intervals, for example, which is what HA was referring to with Alpha Level). An interesting way of examining HA's measurement system would be to apply the same standard to another QB, either one who's output is unquestionably higher than Losman's (Brady or P.Manning) or one who's output has meaningfully shifted year to year (D.Brees or E.Manning, perhaps. If HA's measurement system were truly capable, it would be able to detect the difference between Losman and Tom Brady or between Drew Brees "good" years and "bad" years. It is important to verify the location of the threshold is adequate by moving the thresold within reason and seeing whether the conclusions still hold. That's truly how you measure the adequacy of a measurement system. If can detect the difference between samples known to be different. This is coming from someone who does this sort of thing for a living.

Amherst borders the City of Buffalo on the north side....

I'm almost certain this study was done on a crimes per capita scale already. The term rate means a ratio of two things. That is almost certainly how Greece ranked so high. It is Rochester's largest population center.

Not to mention he's had two rediculously costly false start penalties. He's the freakin' fullback, how the heck does he false start?

Here's an idea, if you don't like the topic, don't read the post, and don't post in it. This is a discussion board. In a discussion, comparisons are made. It's part of normal discourse. If you've heard it too many times already, stop listening.

You make it seem like what people on this board say actually matters.... Losman will be fine regardless of what is said on this board after the game.

Which chart? The statistics in the tables by distance and down are current and provided via the splits on ESPN.com. The control chart is also up-to-date (takes about 30 seconds in JMP) and is my own creation. The last two games do not negatively impact his numbers at all (Detroit was right at the mean and NE was above it). CLARIFICATION: This particular control chart was compiled by me, they are a very well established statistical tool in general. :-P

The breakdown by distance and down are for 2005 as whole versus 2006 as a whole. In general, the distinction between the two stints is important, I agree. However, in the case of completion percentage, it's not. A control chart of his completion percentage from Stint A to Stint B to 2006 shows no statistical differences from Stint A to Stint B and a shift in the mean at the beginning of 2006. The comparisons of the aggregate are appropriate.

I have to agree. I don't think using the t distribution here is inappropriate. I'm not sure what having the NFL standard would do for you. The test of mean using the t distribution (I call it the T Test) is dependant only on the sample size n, the variability seen within that sample of size n, and the specified confidence interval. The only problem with the t test, as I mentioned in my previous post is that less sensitive to subtle but definite shifts in the mean over a small sample size.

Your definition of noise is fine, but the general explanation might be better to use here: Noise is variation that can be explained by the natural, random behavior of the system (I.E. cannot be explained by a special cause). A t-test is the general method for determining differences in mean to a specified confidence. However, the T-Test is not nearly as sensitive at picking up process shifts when samples sizes are small (the process being quarter backing a football team in this example). Something as simple as control chart with run tests would pick up the change quicker.

It is funny hearing an argument from a guy named Holcomb's arm AGAINST completion percentage. However, you are right, completion percentage can be artificially inflated. In the case of J.P. Losman, it hasn't. Pass Thrown in Air Table, 2006 Distance______Att_______Comp______Comp%_______% of Total Plays Behind LOS____46_______28_________60.9%__________23.4% 1-10 Yds______95_______70_________73.7%__________48.2% 11-20 Yds_____31_______17_________54.8%__________15.7% 21-30 Yds_____14_______6__________42.9%__________7.1% 31-40 Yds_____6________1__________16.7%__________3.0% 41+Yds_______5________0__________0.0%___________2.5% Pass Thrown in Air Table, 2005 Distance______Att_______Comp______Comp%_______% of Total Plays Behind LOS____46_______25_________54.0%__________20.4% 1-10 Yds______112______66_________58.9%__________49.5% 11-20 Yds_____42_______15_________35.7%__________18.5% 21-30 Yds_____10_______3__________30.0%__________4.4% 31-40 Yds_____9________2__________22.2%__________4.0% 41+Yds_______7________2__________28.6%__________3.0% Completion PCT by Down, 2006 Down___Comp%____Att 1______70.1%______77 2______57.1%______63 3______56.4%______55 4______50.0%______2 Completion PCT by Down, 2005 Down___Comp%____Att 1______41.0%______83 2______63.2%______68 3______46.6%______73 4______50.0%______4 It's been argued in the past that Losman is inflating his completion % by throwing shorter passes(<10 yards), which clearly is not the case (only 2.0% more over last year). He's just MUCH better at it. In addition, the most significant improvement is coming from his improvement in the 11-20 yard pass range. It has also been argued that Losman is inflating his completion % by throwing short on third and long (ala Mr. Holcomb). If that were the case, his completion percentage would have to be significantly better on 3rd down than the aggregate (to make up for the fewer attempts). From the table, that is also clearly not the case. The real change is that Losman is much better this year at the short and intermediate passes and is really damn good at the first down pass. This is the progress that I see. I still see a QB that loses the ball at odd moments and who should learn when to throw the ball away. But anyone who argues against progress is simply ignoring facts.

This is the most reasonable statement you've ever made on these boards in regards to statistics. You are probably right. But have you done the work to prove it or are you just assuming?

Their system is interesting, I will give them that. They attempt something very important, putting the data in context. However, I've read their book and their methods, while a huge step in the right direction, are still amateurish. Some of their comments in regards to statistics make me cringe. In addition, by their own admission, these rankings are much more for the passing game as a whole rather than the individual quarterback. They are unable to decouple the quarterback from the offensive line and wide receivers. It's a difficult issue to avoid. Two weeks ago Losman was 14th in their rankings. Now he is 29th. Where he ends up at the end of the year will be interesting, and possibly somewhat valid, but not before then.

Tom Brady missed two wide open receivers deep in the first freakin' quarter. Losman missed one, big deal. Hell, Brady has completed 38 % of his passes past 10 yards this season. Losman 43%. So what? If you're going to use Brady as the standard to which Losman fails to reach, you should at least use aspects of the game that Brady is better at. There are many.

Of course. 99/162 61.1% 1131 yds 6.98 yds/comp 5 int Also, the Bills have only attempted to pass 10 times in 7 games with less than two yards to go.

But drives are not sustained solely on third down. In fact, third down performance is largely a function of first and second down performance. An interesting metric is percentage of first downs responsible for. Currently the Bills have 100 total first downs from plays (with 9 coming from penalties), with 39 coming from run plays and 61 through the air. That means that Losman's passing performance is accounting for 61.0% of the offenses first downs. That compares favorably with the league: J.Losman 61/100 61.0% Brady 71/118 60.1% Grossman 68/103 66.0% P.Manning 90/136 66.2% E.Manning 78/129 60.5% Palmer 70/100 70.0% McNabb 92/135 68.1% Rivers 70/118 59.3% The Bills problem, and Losman's problem, has not been sustaining drives as you assert. It's scoring points. The Bills, and Losman, have been fairly succesful in between the 20s. Where both Losman and the Bills in general have suffered is in the red zone. But the blame does not lay solely with Losman, as Willis McGahee is averaging 1.53 yds/carry (43 yards on 28 carries) in the red zone and has 1 TD. Hell, Losman has 77.7% of the Bills TDs on offense (7/9) and McGahee has 11.1% (1/9).

Where did I omit that information? I gave attempts, completions, yards, and TDs in my posts. You want me to calculate out every ratio for you? The statistic you cite was available from the information I presented. You also have an interesting leap of logic in your second paragraph. How exactly does an increase in completion percentage coupled with a level ouput in yardage mean he's NOT more effective at moving the offense down the field? Those two statistics coupled together mean his yards per completion has dropped off, that is correct. That does not necessarily mean he is completing fewer long plays. It may also mean he is completing more short passes. The statistic you presented is at worst, inconclusive, and at best, supportive of my argument rather than yours. You need to complete passes to sustain drives.

That's because the person who originally cited those statistics (incorrectly) omitted the statistics that would hurt his/her argument. The missing statistics: '05 134 Att 68 Comp 50.7% *(5 Games Cited Above) '06 197 Att 122 Comp 61.9 % (+23.47%) That's progress.

Your numbers are off. The totals at the bottom of the '05 set are for the full year, not just the five games he cites. The rest of the data was truncated. For the five games cited in '05 versus the seven in '06 the statistics you refer to are actually: '05 7 TD 907 YDS in 5 games (1.4 TD/Game, 181.4 YDs/Game) '06 6 TD 1212 YDS in 7 games (0.85 TD/Game, 187.4 YDs/Game, +3.3% YoY)

Were you expecting playoff's this year? What exactly has changed between the beginning of the season and now? Coming in, it was a young team hoping to improve as the year went along and expecting a roller coaster ride along the way. That's exactly what's happened. The thing worth watching, just as it has always been for this team, is improvement. I watch every week because I want to see individual plays, because that's where I get to see if players have improved. It's that simple.

The differences in the league from then to now do not matter. The actual statistics do not matter (because the league is different now than then). The only thing that matters is the change in the statistics over time. It can be seen from Bradshaws career numbers (partially displayed in this thread), that Bradshaw was a markedly different QB his first five years than he was for the rest of his career. That is the point people should be taking away from this thread. Don't let the specific numbers cloud your mind.

....and that's Losman's fault how? Eli Manning came at a pretty high price, too.

Tom Brady overthrew wide open receivers deep twice in the first quarter against the Bills. You can't say any QB should make that throw with their eyes closed, and you can't point to two throws in 7 games as a case for anything.

Ah, so he didn't misuse a word. He misspelled a word. Perhaps he's not the only one who occasionally misuses a word.