Orton's Arm Posted August 29, 2006 Author Share Posted August 29, 2006 Its humerous that you account O-line to wins, but when a team does good, you discount their O-line because it doesn't match your theory as not being the key for a good winning percentage. This is just a pet peeve of mine, so bear with me. But you should have written, "when a team does well," not "when a team does good." Nothing personal, it's just that that particular thing bugs me. I admit I sowed a little confusion about the role high draft picks play in determining a team's success. The first regression I ran sort of showed that simply using high draft choices on offensive linemen won't do much good. The second regression showed that when you actually succeed in filling starting OL positions with early draft choices, it helps a lot. At least on average. On average, men are taller than women. But you could find plenty of examples where an individual woman was taller than a specific man. Likewise, you could find examples where a team that built its offensive line through free agency or lower round draft choices had a better OL and a more successful team than a team which found its OL starters via high round draft picks. But if your intention is to build an offensive line through free agency or lower round draft picks, the odds are stacked against you. Teams typically don't let their best offensive linemen hit free agency. Orlando Pace is a Ram for life. Or when a good offensive lineman does hit free agency; it's often late in his career. Chris Villarrial is a good example of this, as is our own Ruben Brown. As for lower round picks, there was a thread about the topic a while ago. Apparently the last 15 or so day 2 offensive linemen the Bills drafted contributed very little to the team. (The initial post excluded anyone drafted more recently than 2004.) Link to comment Share on other sites More sharing options...
Ramius Posted August 29, 2006 Share Posted August 29, 2006 I admit I sowed a little confusion about the role high draft picks play in determining a team's success. The first regression I ran sort of showed that simply using high draft choices on offensive linemen won't do much good. The second regression showed that when you actually succeed in filling starting OL positions with early draft choices, it helps a lot. At least on average. 755673[/snapback] You arguement is still assinine. So spending high draft choices on the OL doesnt correlate to winning (notice i said correlate, not cause, a concept you have a lot of trouble grasping), but spending high draft choices on players that eventually turn out to be good does correlate to winning. simplifying: drafting players and OL that are good improves your team. No sh-- sherlock! Congrats, you've proved nothing! you've just spewed more assinine nonsense. kinda like every other post of yours. Link to comment Share on other sites More sharing options...
2020 Our Year For Sure Posted August 29, 2006 Share Posted August 29, 2006 Yeah, this would probably work with most positions. The higher the percentage of your first-day picks that go on to fill starting spots, the better your team will be. It's obvious and statistics aren't needed to grasp that. Because when your picks aren't filling starting spots, then they're busts, and thats wasting very good oppurtunities to get good players. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 29, 2006 Author Share Posted August 29, 2006 IMO this(bold) point virtually kills the usefulness of the statistics.My interpretation of it is..... When a team drafts a player in the the first round who pans out well, it reflects in the win/loss record. Well....at the sake of sounding rude....no shiit Sherlock Teams usually draft a position high because they are sorely lacking in that area. It is only logical that if the draftee succeeds, the upgrade for the team is noticeable. Add to that the fact that a higher percent of early picks(that succeed) will be 'very' good & the upgrade will be very noticeable to the team. I would bet that your criteria would pan out for pretty much all positions, not just the OL. 754890[/snapback] I know the outcome of the second regression may seem obvious in hindsight. But without the regression, could you really say for certain if it was better to fill your offensive line via first-day picks, second-day picks, or free agents? Teams have found successful offensive linemen through each of these three methods; so it's not immediately obvious that finding one's OL starters through the first day of the draft would correlate so strongly with winning. But the correlation is there, which is what the regression shows. Link to comment Share on other sites More sharing options...
Chilly Posted August 29, 2006 Share Posted August 29, 2006 I know the outcome of the second regression may seem obvious in hindsight. But without the regression, could you really say for certain if it was better to fill your offensive line via first-day picks, second-day picks, or free agents? Teams have found successful offensive linemen through each of these three methods; so it's not immediately obvious that finding one's OL starters through the first day of the draft would correlate so strongly with winning. But the correlation is there, which is what the regression shows. 755748[/snapback] Doesn't the correlation only show that improving your offensive line through the first two rounds helps a team? It doesn't include free agents or lower round draft picks! Link to comment Share on other sites More sharing options...
syhuang Posted August 29, 2006 Share Posted August 29, 2006 No. According to the regression I ran, each time you fill a starting OL spot with a first day pick, it's worth one extra win. 755663[/snapback] You need to analyze more data before you can conclude one first day pick in the OL means one win. As Ramius mentioned, Correlation does NOT mean causation. Here is another example: Based on the stats since superbowl I, the 4th overall draft pick generates more pro bowl appearances than 1st overall pick. Does this mean the quality of 4th overall pick is better than 1st pick? Although it's against the common sense, People only look at this data may still say 4th pick has better quality than 1st pick. But if we look more carefully, the most pro bowl appearances come from undrafted free agents, more than 1st and 4th picks combined. Does this mean teams should forfeit their draft picks and just sign undrafted rookies? I believe any fan with football knowledge doesn't agree this. We can not just base on insufficient data to conlude causation. In this case, we should analyze more data, for example, total number of undrafted rookies, the successful rate of each pick, and so on. In your case, you should also analyze more data like other people suggested, you miss too many factors to make valid conclusion. Link to comment Share on other sites More sharing options...
Ramius Posted August 30, 2006 Share Posted August 30, 2006 You need to analyze more data before you can conclude one first day pick in the OL means one win. As Ramius mentioned, Correlation does NOT mean causation. Here is another example: Based on the stats since superbowl I, the 4th overall draft pick generates more pro bowl appearances than 1st overall pick. Does this mean the quality of 4th overall pick is better than 1st pick? Although it's against the common sense, People only look at this data may still say 4th pick has better quality than 1st pick. But if we look more carefully, the most pro bowl appearances come from undrafted free agents, more than 1st and 4th picks combined. Does this mean teams should forfeit their draft picks and just sign undrafted rookies? I believe any fan with football knowledge doesn't agree this. We can not just base on insufficient data to conlude causation. In this case, we should analyze more data, for example, total number of undrafted rookies, the successful rates of each pick, and so on. In your case, you should also analyze more data like other people suggested, you miss too many factors to make valid conclusion. 755769[/snapback] Thank you! using Holcombs arm lahjik, i should stop eating ice cream, because ice cream consumption is highly correlated to drowning deaths. Never mind that the only reason they are correlated is because both activities are high in the summer. They are correlated, so one must cause the other! Do not ice cream or you will drown!!! Link to comment Share on other sites More sharing options...
OCinBuffalo Posted August 30, 2006 Share Posted August 30, 2006 I know the outcome of the second regression may seem obvious in hindsight. But without the regression, could you really say for certain if it was better to fill your offensive line via first-day picks, second-day picks, or free agents? Teams have found successful offensive linemen through each of these three methods; so it's not immediately obvious that finding one's OL starters through the first day of the draft would correlate so strongly with winning. But the correlation is there, which is what the regression shows. 755748[/snapback] Allright - enough! I have tried to stay out of this because I didn't want to show my nerd side BUT- I can't handle this anymore. While I appreciate your regression attempt - it's mostly right - you really have got to understand the concept of a control. Are u a pschology guy or a pure science guy? (Correlation vs. Causation). I think the reason you are driving some folks nuts here is that your correlations are not enough since they do not show a clear pattern based on consistent(important), raw data - they are just correlations. By messaging the raw data enough I am sure that I could draw a correlation between the appearance of the new Sabres logo and the number of Offesive lineman drafted. You aren't controlling for variables properly here - or at least attempting to hold something constant - or tossing them out altogether. And, I have yet to hear about a consistent model that defines a "good" offensive lineman. So far, what we have is good draft picks = good team = wins = good draft picks = good team. This is defined as circular reasoning. Is it any wonder we have identified a correlation in the midst of a circle? I see what you are trying to do, and like I said, I appreciate it. But you have to hold at least one costant - some standard to which the rest of the data can be compared. Try defining "good" for an O lineman by itself - not related to anything else,(don't forget to tell us your definition ), hold that standard constant and throw out any O lineman who doesn't qualify. I.E. - I would use something like "played for the team that drafted him for >= 5 years" as part of my definition since we are trying to find the relationship between drafting O line and team wins(other parts might be games started, sacks allowed, etc.). This would remove the free agent, and one year wonder, variables that seem to be tripping you up, and, it would give us a real look at the long term value of drafted O lineman that stay with the team. The nice part about a clear definition like this is that then you can test the opposite definition - O line that played <5 years - while holding the other parts of your definition constant and see what you end up with. Anyway, once you have cast the data against your definiton now you have your raw list of "good" O lineman. Then tell us what you find as it relates to the draft round selected, number of wins for their team, playoff appearances, or some of these together, or anything else that suits your fancy. I think if you can get this all done effectively, you might be on to something. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 Allright - enough! I have tried to stay out of this because I didn't want to show my nerd side BUT- I can't handle this anymore. While I appreciate your regression attempt - it's mostly right - you really have got to understand the concept of a control. Are u a pschology guy or a pure science guy? (Correlation vs. Causation). I think the reason you are driving some folks nuts here is that your correlations are not enough since they do not show a clear pattern based on consistent(important), raw data - they are just correlations. By messaging the raw data enough I am sure that I could draw a correlation between the appearance of the new Sabres logo and the number of Offesive lineman drafted. You aren't controlling for variables properly here - or at least attempting to hold something constant - or tossing them out altogether. And, I have yet to hear about a consistent model that defines a "good" offensive lineman. So far, what we have is good draft picks = good team = wins = good draft picks = good team. This is defined as circular reasoning. Is it any wonder we have identified a correlation in the midst of a circle? It seems we're not on the same page here. My regression was intended to answer a very specific question: "Do teams which build their starting offensive lines through first day picks tend to win more games than teams that don't?" The answer is a strong yes. You suggest various tools that I should be using: control groups, a definition of a good offensive lineman, means of controlling for other variables, etc. While these tools aren't required to answer the question I asked, they may be needed to answer a different question you have in mind. What specific question is it you would like to see answered? Link to comment Share on other sites More sharing options...
OCinBuffalo Posted August 30, 2006 Share Posted August 30, 2006 It seems we're not on the same page here. My regression was intended to answer a very specific question: "Do teams which build their starting offensive lines through first day picks tend to win more games than teams that don't?" The answer is a strong yes. You suggest various tools that I should be using: control groups, a definition of a good offensive lineman, means of controlling for other variables, etc. While these tools aren't required to answer the question I asked, they may be needed to answer a different question you have in mind. What specific question is it you would like to see answered? 755817[/snapback] You are still confused I see. Let's breakdown the Question you posed: Do Teams ( groups not individuals so we will need to group o lineman later on) which build their starting offensive line ( so now we will have to look at individual and starting offensive lineman as well) through fist day draft picks(1st and 2nd round individuals drafted) tend to win more games than teams that don't( now compare one group of individuals to another group - back to groups again) This should be enough to make my point. You have to identify what you are looking at; indivuduals or teams first - not both at the same time, or worse, relating one to the other prior to drawing a conclusion. Again, define your data as it relates to individuals first, then move on to teams. You can't say a 2nd round pick that plays for 3 years and moves on due to FA, while his team makes the playoffs the next year = good thing they drafted an o lineman high. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 You need to analyze more data before you can conclude one first day pick in the OL means one win. This is a potentially confusing issue, so I'll try to make things clear. For each OL starter you obtain via a first day draft choice, your expected number of wins increases by one. Individual results will vary. Let me put this another way: suppose you were going to meet someone for the first time, and were asked to guess that person's height in advance. If you're told that person is a seventh grader, the expected value of that person's height is going to be lower than if you're told that the person is an eighth grader. Likewise, if you were asked to blindly guess a team's winning percentage, your guess should be higher if you know that team obtained its starting offensive line via the first day of the draft. While I'm at it, I may as well address that whole correlation/causation issue that's been raised. Ramius gave a good example of ice cream being positively correlated with drowning deaths. In that case, both the ice cream consumption and drowning deaths were driven by some third factor (hot weather). You raised a different type of correlation/causation issue: the 4th overall pick tending to get more Pro Bowl appearances than the 1st overall pick. This, I believe, is caused by random chance; and by the excellent play of Mike Williams. Or at least by random chance. Is the correlation I've observed driven by some third factor? Those who suggest it might be would do well to articulate what they feel this third factor might be. As for the random chance possibility, that can be tested with a nice F test. My knowledge of stats is a little rusty, but I'm pretty sure we're looking at 30 degrees of freedom for X and for Y. I obtained an F value of 3.72; which means my findings are significant at alpha levels of both 0.05 and 0.01. In English, this means I'm more than 99% sure the results I obtained aren't due to random chance. Link to comment Share on other sites More sharing options...
Dibs Posted August 30, 2006 Share Posted August 30, 2006 It seems we're not on the same page here. My regression was intended to answer a very specific question: "Do teams which build their starting offensive lines through first day picks tend to win more games than teams that don't?" The answer is a strong yes.......755817[/snapback] I think a lot of people were assuming you were trying to prove more than you were here.... IMO the obvious answer(which I posted earlier) is "Der...of course it is a yes." And the answer will obviously be yes for the following questions. "Do teams which build their starting defensive lines through first day picks tend to win more games than teams that don't?" "Do teams which build their starting offensive skills players through first day picks tend to win more games than teams that don't?" "Do teams which build their starting secondaries through first day picks tend to win more games than teams that don't?" Actually, upon second thought, your question appears to be.... "Do teams which build their starting offensive lines through first day picks tend to win more games than before they had those first day picks as starters?" To quote myself from the earlier post.... "Teams usually draft a position high because they are sorely lacking in that area. It is only logical that if the draftee succeeds, the upgrade for the team is noticeable. Add to that the fact that a higher percent of early picks(that succeed) will be 'very' good & the upgrade will be very noticeable to the team." The only useful thing that you might be able to show from using this sort of data that I can see is whether 1st day picks have a higher upside on average than later picks. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 This should be enough to make my point. You have to identify what you are looking at; indivuduals or teams first - not both at the same time, or worse, relating one to the other prior to drawing a conclusion. Again, define your data as it relates to individuals first, then move on to teams. You can't say a 2nd round pick that plays for 3 years and moves on due to FA, while his team makes the playoffs the next year = good thing they drafted an o lineman high. Ultimately, I'm looking at the effect individual linemen have on their teams' success. If obtaining a starting lineman via a first day pick is helpful to a team's winning, then the one ought to be correlated with the other. And so we find. My definition of a successful offensive lineman is someone who helps his team win. I object to using winning percentages to evaluate specific players--I hate hearing that QB so-and-so is 41-10 as a starter. QB so-and-so's team was 41-10 while that particular QB happened to be on the field. Maybe QB so-and-so is Trent Dilfer; a very average player benefiting from the excellent performance of those around him. While winning percentage can't be used to evaluate individual players, it can be used to evaluate general trends or strategies. In general, eighth graders tend to be tall than seventh graders (with many individual exceptions to this rule). In general, teams which obtain their starting offensive lines via first day draft choices win more games than teams which use other means to build their lines (again with many exceptions). Link to comment Share on other sites More sharing options...
OCinBuffalo Posted August 30, 2006 Share Posted August 30, 2006 Is the correlation I've observed driven by some third factor? Those who suggest it might be would do well to articulate what they feel this third factor might be. As for the random chance possibility, that can be tested with a nice F test. My knowledge of stats is a little rusty, but I'm pretty sure we're looking at 30 degrees of freedom for X and for Y. I obtained an F value of 3.72; which means my findings are significant at alpha levels of both 0.05 and 0.01. In English, this means I'm more than 99% sure the results I obtained aren't due to random chance. 755847[/snapback] Fine but what is your baseline for comparison - teams or players? Since it's already been established that some of the playoff teams have 2 or more high o line picks, and that some have one or none, doesn't it make sense to start with players and then group? You can't start with teams(in this case winning teams), transpose that data to players(in this case high o line picks), and then go back to teams and say: there is a greater chance that players from winning teams are good. That's where the no schit sherlock part comes in. You should start with something that qalifies players(in this case good offensive lineman) only - something that has nothing to do with teams. Let't try o lineman drafted 1st day, played for the drafting team for 4 years minimum, started at least 12 games during each of those four years, for the years 2002-2005 - keep in mind they could have been drafted in 1999 as long as they played, and started at least 12 games a season, during the 2002-05 timeframe. Assuming that the hundred or so head coaches/OC/O line coaches employed over the timeframe cancel each other out in terms of variance(bad decuisions, schemes, etc.), a guy who can meet all those criteria should be classified as "good" or at least - standardized. Now that we have a group that has been selected by a standard(I believe fair), let's see what we can find out about them - let's say you end up with 20. How many belong to one group(playoff teams) and how many to the other(non-playoff teams). Now we can say what % of them played on playoff teams during the time frame. Find out total # of wins for the group during the timeframe and divide by the number of o lineman = average number of wins per lineman. Divide that by 16 and see avg # wins per season. How many of them played on the same team, and how many of those teams won a lot/playoffs. There's prolly more but - Ya get it? These percentages and averages, because they are based on a standardized set of criteria, should give solid evidence if there is any causal effect between drafting o lineman high and teams winning. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 I think a lot of people were assuming you were trying to prove more than you were here....IMO the obvious answer(which I posted earlier) is "Der...of course it is a yes." And the answer will obviously be yes for the following questions. . . . To quote myself from the earlier post.... "Teams usually draft a position high because they are sorely lacking in that area. It is only logical that if the draftee succeeds, the upgrade for the team is noticeable. Add to that the fact that a higher percent of early picks(that succeed) will be 'very' good & the upgrade will be very noticeable to the team." The only useful thing that you might be able to show from using this sort of data that I can see is whether 1st day picks have a higher upside on average than later picks. 755879[/snapback] You seem like a very logical thinker, and part of me wants to agree with what you're writing here. But there are other factors to consider: 1. It's true teams draft for need, especially in the early rounds. Well, most teams anyway. TD drafted Travis Henry when he had Antowain Smith but no OL . . . well nevermind. Most teams do in fact draft for need. But while you're adding new talent on the offensive line via the early rounds of the draft, you may be losing an equal level of talent due to players getting old or hurt. Look at the decline the Bills' defense experienced in just one year's time. In any case, the first day draft choice offensive linemen I looked at had often been with their teams a number of years. I can see how plugging a good player into a hole would get you an extra win or two early on. But if a guy's been with his team eight or ten years, he's not just plugging a hole. He's a piece of a long-term strategy to build an offensive line through the early rounds of the draft. 2. One of the key factors in having a good offensive line is that thing women talk so much about--chemistry. Guys who stay together for long periods of time develop better chemistry than players on offensive lines which keep changing. If you obtain your offensive linemen through the draft, you can keep them with your team their whole careers. On the other hand, teams which build their lines through free agency could find it more difficult to have consistency. Free agent Chris Villarrial only gave the Bills one good year, and he'll soon need to be replaced. Trey Teague was in and out over just a few years. On the other hand, early round draft choices like Ruben Brown and John Fina spent nearly their whole careers here. Link to comment Share on other sites More sharing options...
OCinBuffalo Posted August 30, 2006 Share Posted August 30, 2006 1. It's true teams draft for need, especially in the early rounds. Well, most teams anyway..... 755905[/snapback] This depends on your definition of "need". I am sure you have needed to take a piss, but I also sure you have NEEDED to take a piss. We NEEDED to shore up our d this year and it looks like we did so far. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 Fine but what is your baseline for comparison - teams or players? Since it's already been established that some of the playoff teams have 2 or more high o line picks, and that some have one or none, doesn't it make sense to start with players and then group? You can't start with teams(in this case winning teams), transpose that data to players(in this case high o line picks), and then go back to teams and say: there is a greater chance that players from winning teams are good. That's where the no schit sherlock part comes in. The analysis you've suggested looks interesting, and I wish you the best in conducting it. As for why I did my analysis the way I did it, I'll offer the following analogy. Suppose you had several basketball teams consisting of seventh graders. Then one day, you decided to allow ninth graders to join these teams. Some of the seventh grade teams got as few as zero ninth graders, while others got as many as four. If the ninth graders tended to be better than the seventh graders, you'd expect to see a correlation between the number of ninth graders on a team and that team's number of wins. Without attempting to measure the performance of any ninth grader in particular, you could still conclude (correctly) that in general, adding ninth graders to your team would tend to increase the number of wins. Likewise, obtaining your starting offensive linemen through first day picks tends to increase your team's number of wins. Link to comment Share on other sites More sharing options...
Dibs Posted August 30, 2006 Share Posted August 30, 2006 In any case, the first day draft choice offensive linemen I looked at had often been with their teams a number of years. I can see how plugging a good player into a hole would get you an extra win or two early on. But if a guy's been with his team eight or ten years, he's not just plugging a hole. He's a piece of a long-term strategy to build an offensive line through the early rounds of the draft. 755905[/snapback] That's right(& supports my point). Before him, the team was (on average) in need of someone at that OL position. Not only is that position filled successfully in your scenarios but filled successfully with 1st day talent so (on average) one can expect a higher upside to the player. Now we have a (on average?) close to pro bowl player where previously there was a (on average) player you really wanted to replace. All you are proving is the equivalent of 'guys like lots of sex'. Do the same comparisons with other positions & I'm sure you will find basically the same conclusions. Just quickly, take the QB position. Manning, Palmer, Rothlesburger, McNabb, Culpepper...& going back...Elway, Marino, Kelly, Montana. These guys(and all the others I can't be bothered to list) will logically show that if you succeed in getting a starting caliber QB in the 1st day of the draft, your team (on average) will win more games than before......& guys like sex a lot. If you only take successful 1st day draft picks & plot how those teams improved, you are simply saying..... "When teams draft well in the 1st day of the draft, they (on average) improve on the field." Link to comment Share on other sites More sharing options...
OCinBuffalo Posted August 30, 2006 Share Posted August 30, 2006 The analysis you've suggested looks interesting, and I wish you the best in conducting it. As for why I did my analysis the way I did it, I'll offer the following analogy. Suppose you had basketball teams consisting of seventh graders. Then one day, you decided to allow ninth graders to join these teams. Some of the seventh grade teams got as few as zero ninth graders, while others got as many as four. If the ninth graders tended to be better than the seventh graders, you'd expect to see a correlation between the number of ninth graders on a team and that team's number of wins. Without attempting to measure the performance of any ninth grader in particular, you could still conclude (correctly) that in general, adding ninth graders to your team would tend to increase the number of wins. Likewise, obtaining your starting offensive linemen through first day picks tends to increase your team's number of wins. 755919[/snapback] I was waiting for u to tell me to do it myself two posts ago. Anyway, your analogy is oversimplified in the sense that it does not acount for "meaningful" participation of the ninth graders, or the seventh graders for that matter. What if, like the NFL, some of those ninth grader get hurt and can't play? How about, like the NFL, some of those ninth graders do well the first year but get detention the next and can't play? What if a team with 4 ninth graders ends up benching them all and still wins the championship? Did they win because of ninth graders? The original question posed is not about general relationships but DEFINITIVE evidence that high o line drafting =(this means is equal to not might or maybe) wins/playoffs. By not holding a standard you are ignoring seventh grade participation that might prove to be just as valid a corellation to improved play as ninth grade participation. Without valid comparison - all you have is conjecture. Link to comment Share on other sites More sharing options...
Orton's Arm Posted August 30, 2006 Author Share Posted August 30, 2006 This depends on your definition of "need". I am sure you have needed to take a piss, but I also sure you have NEEDED to take a piss. 755912[/snapback] Now this post is a clear case of congruence between a man's words and his avatar! Link to comment Share on other sites More sharing options...
Recommended Posts