This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!
I started Deshaun Jackson weeks 1-3. My league is surely much different than yours so don't assume that because YOU aren't interested or willing to start what looks like a marginal player that other FFballers are as uninterested as you are.
Andy --It's not really an issue of whether your method does or doesn't benefit FBG in any way.The issue is whether your method does a good job of reflecting the accuracy of the projections. Period.The example I gave shows that the method has what I would consider to be a pretty significant flaw, the result of which is a pretty poor reflection of the accuracy of the projections.As I asked at the end of that earlier post, would FBGs endorse the conclusion shown in my example if this was someone else's study? I would certainly hope not, because the conclusion stinks.As to your point about "punishing" a site despite it correctly predicting a top 4 performance -- *all* the study participants will be "punished" when a guy blows up with 5 TDs.The thing to realize is that the site that had him down for 300 and 3 will be "punished" less than those that had him down for 250 and 2, and thus the results will still reflect the fact that the site with the 300/3 projection did a better job than the rest. Proper credit will still be awarded for the best projection, even if it's way low due to a "blowup" game.And in the grand scheme, the "blowup" games will be a small fraction of all observations, and thus their impact will be minimal when they're averaged in with all the "normal" games.
davearm, you may be right about this. I'm still not sure which is the best way forward here.I was more adamant yesterday about sticking with the proposal, but have softened after re-reading yours and other arguments.I still think that the upper echelon guys should be weighted somehow. If we have 21 QB's in someone's top 12, a good ranking for the guy who ranked 18th and isn't started in the majority of leagues is going to count more than the guy who will be started in 99.9% of leagues and finishes as a top performer and goes crazy. eg Drew Brees throws for 400 yards and 4 TD's, but is projected for 270 and 2.5 TD's is going to count less than Gus Frerotte throwing for 175 yds and 1 TD after being projected for 185 and 1 TD. Because 1 guy had Frerotte in his top 12.Frerotte will be starting in 2% of leagues, Brees 99%I know the top 1/3 thing isn't perfect, but at least it rewards guys projected as must starts. Not sure if there is an easy solution to this, but it's probably best to start simple and move forward from there in future years. We'll all be wiser once this gets put into practice.Thanks for keeping the discussion positive.
First, kudos to you guys for doing this -- That's serious guts and accountability to put your rankings up against the other popular sites, not to mention all the extra work involved. Also, similar to a point made by an earlier poster, even if FBG weekly rankings (both Dodds & Bloom) were to finish near the bottom of the pack (unlikely, but anything is possible), I would still think that all the content FBG provides would be more than worth the price I pay as a subscriber.As to the scoring method, I'm not a statistics guru, and it's entirely possible that what I'm suggesting would be way too complicated to implement, but if you want to counteract the negatives you cite for option 1, couldn't some simple incremental weighting method be used that would apply to all players rather than having some arbitrary cutoff as method 2 uses that potentially produces the seriously skewed results described earlier? Once you translate all the sites' projections into FBG scoring, you'll have effective rankings for each site. If you average them all together to get the a priori consensus 1st - Xth rankings, couldn't you apply some sort of incremental weighting to give the accuracy of the projections for consensus QB 1 slightly greater weight than that for consensus QB 2, and so on?If that's not possible (or some other system can't be worked out), my vote is for option 1 over option 2.
I would lean toward #1. I could make an argument that method #2 favors more conservative projectors versus chronic overprojectors. Let's say we both have Manning in our top 4 for our projections, I have him projected to throw for 300 and 3 td's, you have him at 250 and 2.2 td's. If he throws for 4 td's and finishes top 4, we both get the same score. If he does worse, I perform worse. It's a tie-lose situation. If you think footballguys are going to be toward the conservative end, then I could argue that the methodology favors footballguys.The negatives of #1 aren't necessarily that the lower ranked players are weighted more heavily--its that the players with the widest range of projections are more significant. This could be the big names, or it could be the guys where there is a difference of opinion as to whether someone is a fantasy start or not. I earlier made a reference to a Simpson's paradox problem, that's not completely accurate, because everyone will be evaluated on the same sample size of players at each position. More accurately, I meant that because there will be more WR than other positions each week, projection performance at WR might drive the overall rankings. Then again, maybe it won't, if a position like QB has more variability even if there are fewer number of QB's being evaluated. You might want to report weekly final reports by summarizing by each position, and then also overall. Heck, this way, someone could say they were best at projecting quarterbacks, even if they weren't first overall. More people happy to be involved.
No offense, and correct me if I'm wrong, but I don't think anyone who has posted in this thread has been in favor of #2 except you, Andy. Why do you keep asking for more input? Are you just hoping someone else will agree with that approach? I think it has been clearly shown by others in this thread why that would be an inferior approach to #1. I'm actually pretty surprised that FBG would even be considering a method like #2, for many of the reasons already posted in this thread.
One thing you could do is weight guys based on how many of the sites have that particular player within the cutoff (top 12 in the case of QBs).So if you've got 10 sites participating, and all 10 have Brees in their top 12 QBs that week, then everybody's score for Brees gets a weight of 1.0 (10/10).If only 2 sites in 10 have Frerotte in their top 12, then everybody's score for Frerotte counts (as before), but those scores receive a weight of just 0.2 (2/10).Take a weighted average of all projections, and voila -- the Brees projections are 5x more meaningful to the bottom line than the Frerotte projections that week.Now having said all that, I don't really think such a weighting system is necessary or desirable. Often it's the borderline starts like Frerotte that are more important to fantasy players than the no-brainers like Brees that are going to start no matter what anyone is projecting. That is, knowing which site can most accurately predict which of the Frerotte-caliber players are going to have start-worthy weeks is arguably more valuable info than knowing which site is best at pinning down the must-start studs' numbers.And one other thing. As you try and nail down the methodology, why not use Dodds' and Bloom's projections for all 17 weeks of 2007 to test the various models you've got on the drawing board? You've got the empirical data needed to evaluate whichever model constructs you think are worth considering.
Fair enough. If we go ahead with this, we'll use option 1).It's a good base, but I imagine in future years the scoring methodology will need to become more sophisticated to weight players that most people actually start vs players that are started by few.Anyway, onto the latest. As suggested by the if comment we've hit a couple of hurdles with this.We're determining if they can be overcome or whether they make the exercise pointless.More soon...
Andy,I think you've made the correct decision.Just to reiterate a point that I think still keeps getting lost. If you have a Tony Romo, LT or any top 10 position player, your are generally going to play them regardless of projection. It honestly doesn't matter if your Romo projections are better than somebody else's because it's not going to change my mind to start Romo. Maybe I don't win by as much as you predicted, but it doesn't matter. Where it matters is in those RB3 vs. WR3 decisions for my flex. That is where I win/lose every week.Thanks for listening.Andy Hicks said:
I always and enjoy and respect Cal Bears takes. I do want to add to this one. I agree with the Romo analogy. But if you have a top 9 and 11 QB or a 13, 19, 23 RB because you are stronger in other areas... that starter may not be so clear and you play it week by week. So i would just say maybe emphasize the middle areas more then the top 5's but still would like to see when a ADP, Romo or whomever has a tank game, why it might have occurred or why they should not have been ranked that high that particular week.
On one hand, I get what you're saying. On the other, you don't think there are teams that ended up with Manning, Rodgers, and Cutler? Those are top 10 guys a lot of weeks and I need to know which one (or 2) to start.Same with RBs. I have Peterson, Gore, and Bush in a keeper league. There are just as many WDIS questions with top guys as there are at the bottom.In fact, I could argue the ones at the bottom are less impt, the difference between the 20th and 40th best WR is smaller than that between the 1st and 5th best.
I'd definitely prefer to see the pay sites included if at all possible, but I agree 100% that I trust you guys not to fudge the data, and if you can legally get away with posting the end results for some sites even if you have to omit the details backing up those results for those particular sites, that sounds great to me.
You guys should also track the AVERAGE across all of these projections. I wouldn't be suprised if thats what ended up the most accurate.
Agree that this would be interesting. This feels a little like picking stocks where over the course of a single season (or even a few seasons), one system may prove to be better. But it would be interesting to see over a long period of time if the "index fund" performs better than most. Sticking with the mutual fund analogy, it would be helpful to understand each system's philosophy -- for example, which systems favor a "stick with your studs" strategy vs. a "matchups are everything" strategy?(HULK) said:
Not sure yet.Depends how much time David has. Not sure if should get released week to week or at the conclusion of the season/commencement of next season.Weeks 4 & 5 when we were testing it out were pretty good for FBG.Haven't sorted through the week 6 stuff yettytyty said:
:subscribe:
great thanks!
where can the data be found?
In a FFA debate I once asked Joe if he really though his benefits derived no benefits from the government. When he said not really I asked him what he would do if I set up a web site that promised to provide all of my members all of the Footballguys.com premium content within 10 minutes of being posted here for $5 a year. The answer is why you can't really do this, especially for premium sites. When the sites were mostly free there was a site that did this for you.
Anyone? Andy? Dodds? Bueller? Bueller?
Wow, the silence in this thread is deafening.May I suggest that some update needs to come forth? The cynics and conspiracy theorists may start to believe that the analysis has been done, the results weren't too flattering to FBG's, and the project has thus been cancelled prior to publication. Vacuums always command attention. People will fill them with their own conjecture when not given an alternative.I'm not one of the cynics...but I'm just saying....
Really?Andy posted on 10/12 saying it was a go. These quotes are from 9/23.
Oh come on. You're absolutely one of the cynics. The opening line gave it away.Reasonable people would believe that the thing was an overambitious project shot down by the fact that nobody else wanted their premium content out there for the masses.
link to post added.
Not sure what good week 4 , 6 or 16 projections are good for now.
