You are projecting, as I said. You have to project, because he's never completed a full season at the levels you are projecting him to be. He doesn't have a history of playing full seasons, never mind a career of playing at a top 10 level. You still have to make a bigger assumption for him, then say, a Matt Ryan putting up great numbers. Tanny might leapfrog Ryan into great territory. But he hasn't done it yet, and we're hoping he does. But it's still a projection. Until then, his career in total has been middling, and it isn't a great huge sin to say so, unless you're on a Miami board.
Sigh. You JD, not me, are framing your argument that a player must put together multiple seasons of 16 games above 100 rating. You do that here: That is your logic. I didn't invent that. Tannehill, by your own words, can't be crowned unless he plays 16 games above 100 rating. So I'll ask again, why doesn't that hold for other QBs. Brady did not have 16 straight games above 100, let alone 32 like Tannehill is required to do. Wilson did not have 16 straight games above 100 like Tannehill is required to do. If your point doesn't compute when its brought out to its logical conclusion, the fault probably doesn't lie with me.
I'm not projecting b/c I have to. I'm projecting b/c we're talking about where Tannehill's development is now. There's nothing magic about 16 games or any other number of games when you're trying to project where a player is now. It's just as ridiculous to pretend that career stats matter more than recent stats and circumstances. If I'm projecting Peyton for this year I'm not going to use his career numbers. Reality is that he looked different at the end of the season than he did last year or even early on last year. And circumstances with Peyton are that his age and injury history add a risk that he's going over the cliff. If you're making assumptions based on career stats over the recent stuff than you're making a far bigger assumption. I'm not saying it's a sin to look at historical numbers. I'm saying it's poor and inaccurate evaluation.
I don't agree AT ALL with just washing away Weeks 1, 2 and 3 as if they didn't happen. Aside from there being philosophical problems with that, it's just bad statistical practice. Your claim of those statistics being a statistical anomaly is not correct. The individual YPAs do not have a different enough average nor variance to rule out the null hypothesis. Football is played in challenging circumstances. First few games in a new system is not some black swan factor that almost never occurs. There's always some excuse. Next it'll be that Stills, Jennings, Parker and Cameron are all new and it's taking time for them all to gel together and that'll be the excuse du jour for results that constitute a "statistical anomaly***". Then it'll be some shakeups and/or injuries on the offensive line that constitute the newest "statistical anomaly***". I don't buy that Tannehill is sub-mediocre. But let's at least use honest practices in the evaluation. *** conclusions not based on valid statistical practices or methodology
That's not what he's saying. He's saying that shrinking a sample size is just another means of making excuses. The guys you listed have put up good cumulative numbers when considering a full 16 games. Tannehill has not. In order to make Tannehill's stats good, you have to lop off three bad games. If you did that for the other guys their numbers would increase too. A full season is what matters because all the games matter, and until Tannehill actually performs at that level, it is indeed projection.
There's a difference between roster turnover that happens every year and a system change that doesn't. If Lazor were to leave before the following season I would modify my season projections downward in anticipation that an adjustment period would occur. But it makes no sense to include something that isn't occurring in my projection for next season.
No one, that I saw, was arguing that it is not a projection. Of course its a projection. Its a projection to say Rodgers will be good this coming year. The argument is that people are projecting good things for Tannehill and others, like JD , are saying those positive projections are too positive. Also, Brady will decline at some point. Just because he's been great doesn't mean he'll always be. So with that in mind, is it fair for me to use the 16 game benchmark to determine if he's no longer great? If so, then how do we determine it....most games below 100 in a season? Total QB rating for the season? What if he gets injured? What if he loses his weapons to injury? Are there other factors? If there are other factors, do they exist for Tannehill too? As far as "lopping off 3 games", its not like we're just removing his 3 worst games and calling it a day. We are simply pointing out there's a clear & legitimate line of demarkation. Its not an excuse, its an explanation. If Luck was hurt and his performance suffered for 3 straight games, would it make sense to let those 3 games color his ability of the entire year or the coming year when making projections? Of course not.
It's not like Tannehill had terrible numbers for the year, even counting those first three games. The only thing that he really had that wasn't great was his ypa. If people are being honest, there's no way they can sit here and say that the first three games were even remotely similar to the rest of the season. If you guys are actually suggesting that implementing a new system is always seamless, and the QB and receivers don't have an adjustment period, then there is no reason for us to continue discussing this. Not only that, but the drops were everywhere in the first few games. We lead the league in drops.
I don't think anyone is upset at how positive or negative anyone's projections are. What isn't jiving is the notion of reducing the sample size. When you begin to speak deterministically about a system that is confounded and which you are measuring statistically, you're not going to get a lot of people readily accepting what you're putting forth. As CKP pointed out, you are on very thin ice when you start electing to erase data points, no matter where they are amongst other data points or what your reasoning may be. Your argument boils down to, "I believe I can make this assumption about throwing away games because ____." Right now it's just your gut feeling and that's why people are taking issue with your point. If you can show numerically how your assumption is valid, then do it.
Generally, it's fine to remove data from analysis if you can independently show that the conditions were different for that data. Question is, can you independently show (without using the data itself) that Tannehill was adjusting to a new system in a qualitatively different way during the first 3 games vs. afterwards? I agree the explanation is plausible, but if you can't show that independent of the data itself, it's best to do the analysis on the entire dataset and not remove data. The situation is different when you're projecting what will happen in the future. Each projection is based on some assumptions, and if you choose to assume that Tannehill was comfortable from game 4 so the analysis should begin from there, then there's nothing wrong with that. But at that point you're arguing based on plausibility (which is fine) and not based on statistical analysis alone. btw.. one way to "remove" data without actually removing it is to just find a different statistical analysis that's still justifiable. Easiest example is mean vs. median. The mean is sensitive to outliers while the median is not but both are commonly used and easy to justify. So if you can find a justifiable analysis that naturally "removes" the first 3 games without actually excluding those from the analysis, then you're also fine.
That's just convention though. The standard could just have easily been 8 games or 12 or 18. When I'm considering whether or not to trade for a player mid-season in fantasy I don't look at the whole season if it isn't relevant. My goal is to project how he will produce. The most accurate way to do that is to use his most recent numbers and consider any changed circumstances. If my goal is accuracy than it would be foolish to include data that is so different from the vast majority of the season.
It's also very important to note that the first 3 games may easily have just been bad games that had nothing to do with being ill-prepared from an offensive standpoint. Week 1 the Dolphins beat the Patriots and Tannehill looked fine so I doubt the data says otherwise. Weeks 2 and 3 had Miami facing two really good D-lines and Tannehill was pressured a lot. It's easy to see why Tannehill would've had bad games which had nothing to do with Bill Lazor and his offense. In fact, what we know actually argues that those two opponents (Buffalo and Kansas City) would've proven to be bad match-ups for the Dolphins late in the season as well. That's my guess as to what happened. I think the competition was above average and Tannehill played badly. Based on that line of thinking, it seems more than dubious to take our Buffalo and Kansas City (2 great defenses) and begin the analysis with....Oakland who played like garbage during a game played in London! It's just a guess, but I'd say the new offense had little to do with how things went down during weeks 2-4.
We are projecting he's going to have instant familiarity and chemistry with 3 new receivers as well. No hiccups, just cruise on. Keep in mind, it wasn't just "familiarity" with Lazor's offense that kickstarted his performance. It was Lazor changing the offense after 3 or 4 games to eliminate stuff Tanny doesn't do well. Basically limiting him to short and intermediate passes through the heart of the season.
I disagree with Tannehill playing badly against the Patriots. He had two dropped touchdowns and one that went out of bounds. I thought he had a fine game and a lot of the problems were because the offense wasn't clicking.
The "clear and legitimate line of demarkation" may very well be the opposition. His numbers were poor against good defenses (NE, KC, BUF) and suddenly started playing well against bad ones (OAK, GB, JAX). I guess that could be a coincidence that the lightbulb went on...or it's just that it's a hell of a lot easier to move the ball against the second batch of teams. this is one of the issues with small sample sizes - you just happen to remove the toughest competition.
I just get annoyed b/c before the season everybody was talking about how there would be an adjustment period. There were articles and quotes posted from QBs and OCs talking about it. I recall one from Bruce Arians saying it takes up to 8 games. The speculation went from about 2 or 3 games to 8 games. It wasn't just plausible, it was accepted. When people were making projections for the season many skewed the projections downward b/c of it. I know I lowered my projection b/c of it. I guessed it would take 4-5 games. I also said that the length of that adjustment period was critical in terms of whether or not we made the playoffs. In the end I settled for the team being in the playoff hunt at the end, but not making it in. The point is that I factored in anomalous circumstance for the purpose of accuracy. Nobody complained that I was being dishonest in my evaluation then. I didn't use a statistical model. I used my eyes and head to factor in an obviously anomalous circumstance that didn't occur the previous season but would be a factor in 2014. I'm doing the exact same thing now for the exact same reason and some are acting like I'm lying to get a more favorable projection.
Sure. Except he played pretty good from there on though. Like I said, there was a clear and legit line of demarkation. I'm not sure why we're all of a sudden not supposed to think learning a new offense is not a detriment anymore. Again, does this count for all QBs or just Tannehill?
WADR, its ludicrous that you of all people, are making this argument. You are one, if not THE, most vocal opposition to the importance of WRs and QBs working together in the offseason. You've repeatedly made the stand that Wallace working with Tannehill more was not important. How do you rectify that stance with this one?
Not really. We played two top 10 pass defenses in those first three games and then we played 5 in the last 13. And in fact we played the two same top 10 pass defenses twice, once before and once after.
Oh please, there you go again putting your words in other people's mouths. I said it's no big deal, I never said I'm opposed to it. Typical for your rants by the way. I also said Suh not working out here is no big deal. Aaron Rodgers agrees with me. http://www.packersnews.com/article/...aron-Rodgers-debunks-value-offseason-workouts Which has ABSOLUTELY NOTHING to do this current topic. Seriously, how do you come up with these connections man.
It just never ends with you. 1. I didn't say you were opposed to it. I said you were part of the opposition to the IMPORTANCE of it. Which you are. Learn to read. 2. If you are going to claim that Tannehill needs time to acclimated with his receivers, then please without your BS accusing me of ****, explain how in the hell working together in the offseason is not important to that......I'll wait. 3. It is EXACTLY in line with the topic. Tannehill needs time to acclimate to his receivers according to you. You don't think a WR working with his QB in the offseason is important. 4. I didn't bring up Suh. He is not analogous to the discussion. Tannehill getting used to Wallace is absolutely analogous to Tannehill getting used to his other WRs. Literally a 1 to 1 comparison. I get you're upset because I keep punching holes in your logic, but once again, that doesn't mean the problem is me.
Actually it's good that you brought this up. Magically, the period of acclimation to the offense is being deemed 3 games. Even though others including venerable coaches have said 8 games or even more than that. So why 3 games? One reason and one reason alone: it makes him look good. This is a dishonest statistical practice. It's the definition of data mining.
But your stance leaves no room for acclimation. For whatever reason Tannehill improved after the first 3 games. Many people say it takes UP TO 8 games to acclimate a QB to a new offense. What seems more plausible based on the facts present: - Tannehill acclimated faster based on his performance. - Tannehill didn't acclimate at all in spite of his performance.
I don't want to argue about Tannehill as a QB. I want to throw out a hypothetical. If the Dolphins are still mediocre and haven't made the playoffs by the time we can opt out of Tannehill's contract, who believes we should opt out? Who believes we will or won't opt out?
I don't think opting out of Tannehill's deal is a one-step process. I think the first thing they do is start bringing in younger players to groom and/or a veteran that seems especially equipped for whatever system the Dolphins are running. Personally I think that if Ryan Tannehill can keep his passer rating and/or QBR in the 10 to 14 range long term as he did in 2014, then we're not having opt out discussions unless the Dolphins stumble on a young player that gets a shot and impresses the hell out of them.
But it's good scouting. It's not about making him look good anymore than it was about making him look bad when I downgraded my projection the year before. At some point player evaluation is about more than crunching numbers. It's about using your experience and knowledge of the game to make a projection. Relying just on statistics and including information from a no longer existing situation would be less accurate.
Then don't crunch them. If you're going to use statistical practices, don't use poor and/or dishonest ones.
Its more dishonest/poor usage to use stats without context. Honestly, if Tannehill had been injured for those first three games, wouldn't it make sense to remove them from the equation being used to figure out if he's good or not?
Not sure how to answer that question without arguing about Tannehill as a QB. His contract is structured so that it's relatively easy (cheap) for us to opt out after the 2016 season, so that's 2 years he has to prove he is good enough or not. Now, with the talent and youth on this team, I just don't see how Tannehill can play well in 2015/2016 and the team does so poorly you would think about starting over (including the QB), but IF it came to that, you'd still have to analyze whether the QB is the problem or the problem lies elsewhere. Personally, I think that unless some freak injury occurs or all this improvement was a mirage, we're good enough at the QB position, even if Tannehill's hit his ceiling. A decent QB is hard enough to find, and it's time to work on fixing the other remaining problems.
All I ever did was say that Tannehill produced top 10 stats for the majority of the season and that based on that it was reasonable to say that was where he stood now.
Football is filled with excuses for removing data points until you get them to show what you want them to show.
It's never been about what I wanted. I just evaluated a player the same way I've done it for decades.
Simple question: Would you have proposed prior to the 2014 season, prior to knowing how he was going to perform, that Tannehill's first three (not four, not six, not two) games be tossed from statistical evaluation of his season? You don't have to answer that. I know the answer already.
Actually I did exactly that before the season. Not specifically three games, but I did say that for evaluation purposes we wouldn't know what we had until after the acclimation period. That we would have to evaluate Tannehill from a later point in the season to get an accurate read. This is not some after the fact justification. Knowing there would be an acclimation period, I lowered my projection before the season and knew I would have to base my evaluation for the next season from a later point in the year. It's literally the same thing I've done for decades and it's far more accurate than ignoring circumstances and pretending every season is exactly the same.
The heck with it. Let's just throw out the entire season, since a new offense was installed, and determine that Tannehill made no progress.
