KRC Posted August 28, 2006 Share Posted August 28, 2006 Good one, Kenneth. I believe that may have something to do w/ why the Texans went after a defensive player w/ their first pick instead of a running back. After all, they're getting production out of the RB position in Domanick Davis. What would Reggie Bush add to that? A few more TDs? How does that help if you have a defense ranked last statistically in just about every category? You're still going to lose unless you address that. Even though it wasn't the sexy pick, I applaud Houston for going that route. 753852[/snapback] I would agree with your assessment. They are getting decent numbers out of Davis. There was no reason to get Bush when they had other more pressing needs. That is why you cannot focus on just one thing when evaluating winning percentages (thus making this statistical analysis useless). As Bill_from_NYC mentioned, you need balance. If you have a great OL, you can still lose. None of this was addressed in the analysis. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 28, 2006 Author Share Posted August 28, 2006 Let me address the objections that were made to my analysis Valid Objections - Too short a timespan was covered. Agreed. A good offensive lineman lasts a long time. - Offensive linemen drafted in 2005 shouldn't have been included. Agreed. I watched Orlando Pace get schooled as a rookie. - R^2 cannot be negative. True. It was very late when I posted that. I meant to write that the correlation was very slightly less than zero. Borderline Objections - Only first and second round offensive linemen were looked at. It's possible that the difference between a good and bad offensive line is taking linemen (especially interior linemen) in the third round. My analysis didn't address the effect of any round after the second. Invalid Objections - The success or failure of each offensive lineman taken was ignored. The point of my analysis was to see if the simple act of taking more offensive linemen in the early rounds would increase a team's winning percentage. - Other factors were ignored, such as the quality of the defense, etc. The beauty of statistics is that you can isolate one particular factor and test for significance; while ignoring everything else. For instance, let's say you wanted to test to see if ninth graders were, on average, taller than eighth graders. Many factors go into height, including diet, heredity, gender, etc. But assuming you use good random sampling techniques and large enough sample sizes, all that other junk will cancel itself out, and the height difference between the two grade levels will shine through. Link to comment Share on other sites More sharing options...
stuckincincy Posted August 28, 2006 Share Posted August 28, 2006 Let me address etc... 754121[/snapback] Teams know a lot more about their OL players than you and I. The good coaches know when they have so-so talent. They go into the clubhouse and vomit. Then they come out, and work on plays, formations, blocking schemes, quick-counts or whatever they can think of to maximize what they have. Even with crappy tools, useful work can be accomplished. It's nice to have a high pick OL now and then. It's also nice to draft a couple in the lower rounds per year and season them. Doesn't always work out, but a gem surfaces occasionally. Link to comment Share on other sites More sharing options...
tennesseeboy Posted August 28, 2006 Share Posted August 28, 2006 I think you get your offensive line through free agency (Steve Hutchinson comes to mind) through the draft and through trades. Our offensive line sucks. Getting good offensive linemen should be a priority. Or maybe Losman and McGahee can play just fine without an offensive line. Link to comment Share on other sites More sharing options...
KRC Posted August 28, 2006 Share Posted August 28, 2006 - Other factors were ignored, such as the quality of the defense, etc. The beauty of statistics is that you can isolate one particular factor and test for significance; while ignoring everything else. For instance, let's say you wanted to test to see if ninth graders were, on average, taller than eighth graders. Many factors go into height, including diet, heredity, gender, etc. But assuming you use good random sampling techniques and large enough sample sizes, all that other junk will cancel itself out, and the height difference between the two grade levels will shine through. 754121[/snapback] Let me see if I get this straight...Anything that ruins your hypothesis can be ignored for the simple fact that it ruins your hypothesis? You make assumptions like "well, it probably wouldn't make an impact. I have nothing to support this claim, but since it goes against what I am proving, I am just going to ignore it and make up what I want to be the truth." That explains why I could never be a statistician. You can just make crap up and just ignore things that do not support your theories. Link to comment Share on other sites More sharing options...
Chilly Posted August 28, 2006 Share Posted August 28, 2006 Let me see if I get this straight...Anything that ruins your hypothesis can be ignored for the simple fact that it ruins your hypothesis? You make assumptions like "well, it probably wouldn't make an impact. I have nothing to support this claim, but since it goes against what I am proving, I am just going to ignore it and make up what I want to be the truth." That explains why I could never be a statistician. You can just make crap up and just ignore things that do not support your theories. 754191[/snapback] Not really. Taking his example, in a large population, diet, heredity, and gender factors tend to come in at the same rates for both eight and ninth graders. The percentage that have diet influence their height will be almost the same for 8th grade as 9th grade, if not the same. If you have a large population, and take a valid random sample, you can ignore the other factors because they will be the same for both 8th graders and 9th graders, thus having the same effect on both (in effect canceling each other out). This is also why confidence levels and what not do exist (random samples are hardly perfect) Link to comment Share on other sites More sharing options...
KRC Posted August 28, 2006 Share Posted August 28, 2006 Not really. Taking his example, in a large population, diet, heredity, and gender factors tend to come in at the same rates for both eight and ninth graders. The percentage that have diet influence their height will be almost the same for 8th grade as 9th grade, if not the same. If you have a large population, and take a valid random sample, you can ignore the other factors because they will be the same for both 8th graders and 9th graders, thus having the same effect on both (in effect canceling each other out). This is also why confidence levels and what not do exist (random samples are hardly perfect) 754228[/snapback] So, to apply this to the topic at hand, the defenses cancel each other out, the QB's cancel each other out, injuries cancel each other out, etc? Sorry. Not buying it. Link to comment Share on other sites More sharing options...
Chilly Posted August 28, 2006 Share Posted August 28, 2006 So, to apply this to the topic at hand, the defenses cancel each other out, the QB's cancel each other out, injuries cancel each other out, etc? Sorry. Not buying it. 754303[/snapback] I wasn't saying that its applicable to a football situation, but in many situations it is. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 28, 2006 Author Share Posted August 28, 2006 Let me see if I get this straight...Anything that ruins your hypothesis can be ignored for the simple fact that it ruins your hypothesis? If you read my sig, you'll see I feel TD neglected the offensive line. I ran the analysis thinking there would be a correlation between the number of offensive linemen taken and the number of wins. Well, there wasn't; at least not the way I analyzed the data. I would gleefully have posted my analysis had it turned out the way I thought it would. It didn't, but I figured I'd put it out there anyway in hopes people would shoot it down. But I want it definitively shot down, in a way which clearly and quantitatively shows the relationship between taking offensive linemen early and winning football games. Link to comment Share on other sites More sharing options...
Dibs Posted August 29, 2006 Share Posted August 29, 2006 So, to apply this to the topic at hand, the defenses cancel each other out, the QB's cancel each other out, injuries cancel each other out, etc? Sorry. Not buying it. 754303[/snapback] I think the concept has merrit. If you take a large enough sample over say 15 years of all 32 teams, there is an average percent that each individual team makes(or wins) the superbowl. If you then figure out the percentage of....in this case high OL draftees taken by each team & then correlate the actual teams that make(win) the superbowl & you find that teams that drafted high with OLmen won the SB much more often than ones that didn't, you could fairly safely conclude that drafting above the average with OLmen would increase your chances of winning....it's always statistcally possible to be fluke.....highly unlikely though if the percentage difference is pronounced. Link to comment Share on other sites More sharing options...
finknottle Posted August 29, 2006 Share Posted August 29, 2006 My take, unburdened by actual research: Drafting OL high (1st round, not second) hurts you as a franchise because you are locked in to a higher salary curve than (other than left tackle) what you would expect for the position (see Mike Williams); and for a player who is unlikely to contribute for a year or two at that! So the real recipe is to draft OL in rounds 2-4, and simply be better at evaluating the talent. It would be nice to see if the facts back this up or not. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 29, 2006 Author Share Posted August 29, 2006 My take, unburdened by actual research: Drafting OL high (1st round, not second) hurts you as a franchise because you are locked in to a higher salary curve than (other than left tackle) what you would expect for the position (see Mike Williams); and for a player who is unlikely to contribute for a year or two at that! So the real recipe is to draft OL in rounds 2-4, and simply be better at evaluating the talent. It would be nice to see if the facts back this up or not. 754784[/snapback] I happen to have the data nicely typed in with which to answer a similar question to the one you asked. According to the regression I just ran, each time you succeed in finding an offensive line starter through the first round, you will win nearly two more games each year. On the other hand, each time you succeed in filling a starting OL position with a 2nd or 3rd round pick, you'll increase the expected number of wins by only half a game each year. Probably the first effect is so strong because of people like Orlando Pace, Jonathan Ogden, and other very successful first round OL picks. Link to comment Share on other sites More sharing options...
Dibs Posted August 29, 2006 Share Posted August 29, 2006 I happen to have the data nicely typed in with which to answer a similar question to the one you asked. According to the regression I just ran, each time you succeed in finding an offensive line starter through the first round, you will win nearly two more games each year. On the other hand, each time you succeed in filling a starting OL position with a 2nd or 3rd round pick, you'll increase the expected number of wins by only half a game each year. Probably the first effect is so strong because of people like Orlando Pace, Jonathan Ogden, and other very successful first round OL picks. 754794[/snapback] That is intriguing. A few Qs on it.... How many years was the data gathered for it? Can rounds 4-7 be added? It would be interesting to see if rounds 2-3 differ much from 4-7. I'd be very interested to see a similar analysis done on QBs, LTs(not the entire OL), DEs & RBs....QB in particular.....hell, all positions would be good to have a look at. This could be an interesting way of statistically determining the impact(importance) of each position on a team. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 29, 2006 Author Share Posted August 29, 2006 That is intriguing. A few Qs on it.... How many years was the data gathered for it? The data were gathered according to the methodology I described in the part II thread. That is, I looked at each team's starting offensive linemen, and counted how many had been obtained via that team's own first day draft picks. As for the other questions you asked, you are a most greedy person! That said, I'm tempted to at least do the QBs, because it would be a little easier than some of that other stuff, and because QB is such an important/controversial position. Link to comment Share on other sites More sharing options...
Recommended Posts