2017 Running Back Report

[Cameron Giles]

The guy made an awesome call on David Johnson. Awesome. I don't mean to present an argument against him as a whole. By all means, listen to the guy who called his shot on David Johnson. I mean that sincerely. But don't backwards engineer a system and make ridiculous claims about it as a predictive model. That's a ridiculous thing to do with a sample size of zero. Literally zero of predictions have a change to come to fruition. Even DJ is only a third of the way to doing so.

I don't think any fantasy model is truthfully predictive, though I think some are more likely to give you positive results than not. DD's report does say:

"Any RB can have any outcome in the NFL, all I'm trying to do is show you the odds or likelihood of a particular outcome. There will be Rb's not on any of these lists that have done very well for themselves. Some things simply can't be predicted based solely on numbers."

The particular outcome is the transparent ability on paper to become long-term, three-down, workhorse backs in the NFL. It doesn't mean that Dalvin Cook or Alvin Kamara can't eventually become one. It just means, they don't have the clearest ability to become one. They lack some sort of trait of football talent and/or production, that is usually synonymous with great RB's.

[Concept Coop]

Nobody can predict the future.

Uh...

Fair enough. I don't think argument is a bad thing, but I'm one guy. I'll move on.

[Concept Coop]

The particular outcome is the transparent ability to become long-term, three-down, workhorse backs in the NFL. It doesn't mean that Dalvin Cook or Alvin Kamara can't eventually become one. It just means, they don't have the clearest ability to become one.

I really am moving on, but just wanted to respond to this before I do. Very specific claims were made. The testing set was included in the reported performance of the system as a predictive model. No details were provided. That's all I've questioned.

[ImaRounder]

The particular outcome is the transparent ability to become long-term, three-down, workhorse backs in the NFL. It doesn't mean that Dalvin Cook or Alvin Kamara can't eventually become one. It just means, they don't have the clearest ability to become one.

I really am moving on, but just wanted to respond to this before I do. Very specific claims were made. The testing set was included in the reported performance of the system as a predictive model. No details were provided. That's all I've questioned.

I don't think argument is a bad thing either, Coop. I also don't want you to feel like I'm attacking you. I think you have warranted questions. However, I do believe there is a better way to question things than to attack Delorean. To question is a great quality, to attack is a terrible one. That's all I was getting at.

[Goddard]

The best part about all this is that we will be able to revisit this in a year and see who was right or wrong. Not as an "I told you so," but more to see how accurate DD's report is and how little we all know lol.

[AussieMate]

Yeah coop I'm not sure what your trying to get at, I've stated a few times (which you obviously haven't read) that you don't have follow this as gospel, I like it, I'm giving it a run but... I've stated a few times I still like Gordon, I also like Mixon as my no1 along with fournette. I don't understand how this can possibly be so hard for people to get, He has a formula that he likes... He has shared it with us all... Use it how you will just like everyone else who has read through it. Once again, the formula put foreman in that percentile, not DD, not me. It is obviously an audacious claim, let's find out in a few years instead of getting worked up like DD or anyone else have stolen free choice from you. I don't post alot on here and this is one of the reasons, share your opinion people but try not to make it personal. I honestly can't tell if people are trolling sometimes but this is the internet.

[Servo]

I can tell you that Rotoviz has been using player statistics to spit out predictions that have held water so far:

Previous Classes:

Future Success: Henry, Zeke, Gurley, Ajayi, Coleman, Gordon, David Cobb, Howard, DJ, Prosise

Avoid: Karlos, Dixon, Ervin, Booker, Buck Allen, Collins, Lasco, Zenner, Duke Johnson, Langford, Perkins, Artis-Payne, Abdullah, Matt Jones, Yeldon, Drake, Josh Robinson.

SURE, these predictions are young but if something can give me an idea of who to avoid wasting time, energy and most importantly, draft capital on based on production, measurables, etc, then why not pay attention?

[zelldawg26]

^^ Hope this is right about Perkins. Targeting Gallman in the late 2nd, early 3rd.

[Dynasty DeLorean]

There ain't nothing different from DD trying to spread the word if his report than the Jesus freaks spreading the word of God... at the very least the report is entertaining.

No different than reading any writers for DLF or any other site, they're all full of it.

Whew. With friends like these, who needs enemies?

[FiremanEd]

Correct me if I'm wrong, but I feel like this thread goes this or a very similar route every year. Admittedly I enjoy it and am glad it always makes a return. If nothing else it makes following the career of D'Onte Foreman that much more exciting! I don't have shares this time around (unlike with DJ), but still an interesting follow.

[dlf_jules]

Coop and others have touched on an important point: most predictive models are trained on one set of data and tested on another. Or, if not, the analyst acknowledges that and treats the results accordingly. If we combine the training and testing sets we're just saying, "These are things that have defined success in the past." If we separate them, we can say, "Here's what the model found would predict success. And here's how effective it was at predicting success on a completely different set of players."

Now, DeLorean has partially overcome that issue by publishing his predictions each year. In effect, he's using each new crop of rookies as his testing data. The trouble there is that anytime he tweaks his methods to accommodate new information, it's erasing some of the value of the testing.

No doubt DeLorean is putting a ton of work into this, and I'm glad to see it on the forums.

[sloth8u]

where does bob turbin rank of this test?

[Reljac]

Correct me if I'm wrong, but I feel like this thread goes this or a very similar route every year. Admittedly I enjoy it and am glad it always makes a return. If nothing else it makes following the career of D'Onte Foreman that much more exciting! I don't have shares this time around (unlike with DJ), but still an interesting follow.

I was thinking the same thing. Of course we don't see all those people complaining about the methods or models doing anything productive and building a better one. Always much easier to cast stones than build castles.

[FiremanEd]

Correct me if I'm wrong, but I feel like this thread goes this or a very similar route every year. Admittedly I enjoy it and am glad it always makes a return. If nothing else it makes following the career of D'Onte Foreman that much more exciting! I don't have shares this time around (unlike with DJ), but still an interesting follow.

I was thinking the same thing. Of course we don't see all those people complaining about the methods or models doing anything productive and building a better one. Always much easier to cast stones than build castles.

I have no issue with people poking holes in the approach. I feel it is better than following blindly like many no doubt will just by looking at the initial table and flashy names. How much you put into the analysis is for each to assess on their own. That said, while I agree with the identifying flaws in the logic/approach, I do think that it can sometimes come off more abrasive than may be necessary, but that's sometimes part of a good discussion too - passion on the topic to make your point. All within acceptable boundaries though.

DD has no obligation to go into the depth of the algorithm, though I know he has before and likely just doesn't have the desire now knowing some will never be satisfied anyway and only posted this out of the strong demand/request for it. That's completely fine too. I don't think everyone should just accept it as gospel though is the point, much like ignoring the tape and focusing on the Combine as your measure of success it would be a mistake. 80% and 0% are bold claims...we will see if Foreman falls into the 20% that the initial table of participants no doubt suggests he likely should to balance the scales. 0% could just means he won't be David Cobb after all too unless I missed a piece of definition, which is a low bar regardless.

[schiewerma21]

Why is there so much arguing? Lol. DD has the ease of changing his model in hindsight as he goes. He had Orleans darkwa in his stud column years ago and many other players in the top columns that have been moved (gurley another one this year)

It doesn't matter what any of us say. DD opinion is just as good as any and I appreciate him putting his stuff out here for us to see another viewpoint. He misses like everyone else but also has proven to hit big which is key in dynasty. If you don't agree don't read it. And certainly don't post 30 times trying to prove your point that quite frankly DD doesn't care about.