More 2020 Hype: "If You Need A RB or WR, this is the draft for you."

[Cult of Dionysus]

I'm still holding out hope for Riley Ridley.

[OhCruelestRanter]

I'm still holding out hope for Riley Ridley.

He’s butt.

[DJB]

Thanks for running the test I appreciate it. As expected, the R4 and later was weighted really significantly. I'm assuming to get all the way up to that original 94%, the R4 and later was at some crazy high miss percent in the 97-99% range.

However, I'm not surprised, and this data is exactly actually validates exactly what I was saying. The number that your model's miss rate should be compared to is not the true hit rate of WR's, but the true miss rate (we need to be comparing the same "miss" parameter, and I mistyped true hit rate in my post above- my bad). Essentially, we naeed to see how much of an improvement your models miss rate is compared to the regular miss rate of any WR in these rounds without any limitations of DR/BA/or any other stat. The true miss rate is defined as (100 percent - the hit rate).

For example, R1 hit rates of 24 percent for example in 2019 (based on the link Cult of Dionysous linked) correspond to a 76 percent miss rate, which is really close to the 82 percent miss rate for 30 DR that your model predicts. So with the proper weighting, which you now implemented, the difference between your model's miss rate and the real miss rate of any WR with any statistics such as DR/BA in 2019 R1 for example is only 6 percent, or negligible. This would need to be run for all the years/rounds obviously.

Ultimately, the difference between your models miss rate and the real NFL miss rate in these rounds is not *that* significant. Definitely not to the point where we should exclude context. There is a really high chance the small difference is accounted for in other factors of why low DR might have been caused in the first place, like poor usage, bad QB play, bad coaching, injury, etc. Keep in mind I'm not trying to come across as harsh, but only because I don't want others to misinterpret the data. 82% accurate miss rate sounds really great until you realize the normal miss rates in these rounds without any of the DR/BA thresholds set in place is also quite similar. This makes a lot of sense as well because we all know the rate of WR that get drafted and actually hit is really low, and so naturally the rate of WR that get drafted and miss is really high (just like your models miss rate).

I appreciate this but at the same time the definitions of a hit from that twitter post and my own spreadsheet are different. my definition is a top 24 season. It's unfair to judge 2019 already when it can take 3 years for a wr to break out.

I'd be interested to see the stats of how many drafted have a top 24 season, compared to my rule oit method, and it's very doable and easy I think as I already have the data. might take a few hours but in would be willing to bet it's better than the nfl hit rate.

Just as I suspected... the science of ruling out misses is far more accurate than the hit rate in the NFL. And a DR of 25 or 30 has no significant difference on hit percentage...

When I compare apples to apples for what constitutes a hit:

Round 1
Draft hit Rate: 50%
Rule out 87.5%
Rule in 63%

Round 2
Draft Hit Rate: 40%
Rule out: 80%
Rule in: 50%

Round 3
Draft Hit Rate: 20%
Rule out: 91%
Rule in: 27%
*This is the only round that's remotely close.

Conclusion:
You can seriously increase the chance of drafting a hit of round 1-3 WR by first ruling out the likely misses

Great post by you guys.

But now I want to know. Who are the likely misses based off those criteria?

[stoneghost28]

Just say Michael Thomas... Not some 80s no names WR lol

I think I responded to this earlier, but I just wanted to expand on it just w/the point that I think having a deeper range of comps beyond guys from the aughts or this past decade can be actually helpful, even if the league has changed a ton, going back to the eighties. I think you probably start running into major issues w/the seventies when you find plenty of teams at 50/50 or even lower pass to run ratios (the Steelers in particular stuck out to me, which is why evaluating Stallworth and Swann was so damn hard, the volume of that passing game was often literally half of what you see today in many instances) but I still think it's helpful.

Some guys off the top of my head worth remembering for comps include: guys like Toon, and Morgan, the Marks Brothers, Wesley Walker, Eddie Brown, Roy Green, Harold Carmichael and Mike Quick, Wes Chandler and Charlie Joiner, James Lofton and Stanley Sharpe, Tony Hill, Anthony Carter, Gary Clark, Art Monk, Harold Moore, Dwight Clark, Freddie Solomon, John Taylor, Ellard etc just lots and lots of guys. We shouldn't confine comps to the last couple of years when we find matches. It was a silly comp at the time, especially when considering the comps AJ Brown is getting now (Andre Johnson from a lot of people) but when I watched him, his style of play reminded me a lot of the ferocity that Gary Clark played with circa '85-'92 when you always got the sense he was willing to break his own bones to reach the end zone. Looking at Brown, he just played with a similar ferocity to me. Physically, they weren't similar at all, but just the hunger and edge they played with seemed so on point which is why I ended up with AJ Brown on almost all of my 14 teams.

Anyway, using comps that include guys like Eric Martin can work. You just need to dig a little. History's always relevant.

[stoneghost28]

Whats the deal with Cam Akers? I see so many mocks with him being the No.1/No.2 running back and I just can't fathom anyone taking him Before the big 3. He's meant to be athletically very good but I've heard bad marks against his vision which to me is just a big no-no (Trent Richardson).

Is he a riser that could sneak up or is it just more clickbait mocks.

At this point I won't be touching him early, the big 3 all seem to have plus vision and athleticism which is why they are being valued so highly, what does Akers do that is better than any of them?

I've heard one guy reference him as belonging in the big 3, and I think there may be some hidden individuals, like myself, here and there that wonder if he doesn't belong in the grouping, but I haven't seen anyone emphatically say #1 or #2 w/him, just, he could belong in that tier.

He was the #1 RB coming out of High School and one of the top 10-20ish prospects period. He ended up on a horrific team w/particularly horrific blocking. You can get him on a discount, but he looks like he could be like a more extreme version of Reagor, w/just a ton of volatility in terms of real rookie draft selections. For now I get a feeling he'll slide in somewhere between 5th and 8th in terms of ADP by mid summer, probably around sixth, after the big 3, and 2 WR's. For me, that makes him a steal in the same vein of guys like Chubb in '18, Brown and Sanders in '19, and Cook in '17. If I'm wrong, I'm wrong, but he will be a locked in target for picks of mine in that zone. My only concern will be if he somehow falls out of the top 75 or so and/or lands in a horrific situation, but even then, so long as it isn't a committee w/legit competition that I respect, I'll probably pull the trigger anyway.

[stoneghost28]

Isn’t it common knowledge that Edelman and Reek are the only players with multiple 1000 yard seasons with a dominator rating below 25%, and that’s because one of them was a QB (unless someone else has met the threshold in the last few years)?

I’d say that’s pretty useful.

Tyreek is the only wr in the last 10 years to finish in the top 24 and not meet the dominator and breakout age thresholds.

There are less than 10 drafted WRs who missed on one of the criteria and had a top 24 season. All but 1 (Tyreek) only missed on breakout age, so those guys still had a good dominator rating. Furthermore, the majority of those wrs only spent 1-3 seasons in the top 24.

Wes Welker was a qb in college so he did not have a dominator rating, but he would be the other if we counted him as a zero in both categories.

Anyone saying these stats arent useful at all is seriously giving themselves a handicap when drafting wrs.

If I could tell you I could identify rookie wrs who had a 94% of not being a top 24, and you could eliminate them from your draft board, thus improving your chances of drafting a hit, wouldnt you say that's useful?

I was just focusing on breakout age alone last year in arguing w/fellow redskins fans about McLaurin and my concerns about him. It kept me from drafting him anywhere and by the time I began to think he was going to hit anyway, it was just too expensive to get him (September of '19) and I took solace in the fact that I had a ton of AJ Brown shares (less solace in a lot of Harry shares though I still think he'll be fine, but it will take some painful patience to get him there). Anyway, I found a similar issue w/breakout age. Not scientific anymore, just counting, but somewhere between 85-90% of hits have great breakout age's. I counted less than seven or eight guys going into the '19 season in the league top 50 that were like sub 45th or so percentile in breakout age.

Now McLaurin will break that rule you have above too. He just missed this past year, I believe he was around 25th in total points, and in the top 30 or so in PPG, and that's with Case Keenum and Dwayne Haskins at QB, w/no weapons to speak of on offense to deflect attention and a middling at best OL, and injuries.

I think the important thing to remember w/these cases is that they are strong trends, and they aren't bust proof, somebodies always gonna squeak through, but if you have a choice between a methodology that has a track record of being accurate 85-90% of the time, or something nebulous like just watching tape, I'm definitely going to go with the former. It will mean I'll miss on some guys (Kareem Hunt from the RB position, Terry McLaurin at the WR position) but it means I'll hit more in the long run anyway. It's an edge, that's all, but it's definitely an edge, at least based upon what I've seen, and certainly a better one than watching tape.

[nathanq42]

Whats the deal with Cam Akers? I see so many mocks with him being the No.1/No.2 running back and I just can't fathom anyone taking him Before the big 3. He's meant to be athletically very good but I've heard bad marks against his vision which to me is just a big no-no (Trent Richardson).

Is he a riser that could sneak up or is it just more clickbait mocks.

At this point I won't be touching him early, the big 3 all seem to have plus vision and athleticism which is why they are being valued so highly, what does Akers do that is better than any of them?

I've heard one guy reference him as belonging in the big 3, and I think there may be some hidden individuals, like myself, here and there that wonder if he doesn't belong in the grouping, but I haven't seen anyone emphatically say #1 or #2 w/him, just, he could belong in that tier.

He was the #1 RB coming out of High School and one of the top 10-20ish prospects period. He ended up on a horrific team w/particularly horrific blocking. You can get him on a discount, but he looks like he could be like a more extreme version of Reagor, w/just a ton of volatility in terms of real rookie draft selections. For now I get a feeling he'll slide in somewhere between 5th and 8th in terms of ADP by mid summer, probably around sixth, after the big 3, and 2 WR's. For me, that makes him a steal in the same vein of guys like Chubb in '18, Brown and Sanders in '19, and Cook in '17. If I'm wrong, I'm wrong, but he will be a locked in target for picks of mine in that zone. My only concern will be if he somehow falls out of the top 75 or so and/or lands in a horrific situation, but even then, so long as it isn't a committee w/legit competition that I respect, I'll probably pull the trigger anyway.

Same for me. Him and reagor are very draft capital and scheme dependant. Reagor in the third to the Jets... Might lean shenault at that point. Akers in the third to Detroit? Maybe CEH gets the nod for RB 4

[stoneghost28]

Tyreek is the only wr in the last 10 years to finish in the top 24 and not meet the dominator and breakout age thresholds.

terry mclaurin?

also, does it work in reverse? how many players with great DR and BA go bust?

McLaurin didnt finish top 24 in ppr through week 16


And it is not useful as a predictive tool for hits. only misses.

So you're taking your 10 wrs, shaving off 3 who are very likely busts, and making your choice from 1/10 to 1/7.

This is a damn good point that I'm generally too stupid to note. Thanks.

[stoneghost28]

Everybody trade Deebo and McLaurin right now. They didn’t check the two boxes

I think trading McLaurin is a great move. Deebo too. Neither are prototypical wr1s.
A case could be made to hold for Deebo as he was very close to breakout age threshold

Ehhh, I'm not with that, I don't think Deebo's ever going to be a top 10-15 guy, but I think he will be a steal and a top 20-30 guy depending upon how that offense evolves based upon the '19 draft capital he cost.

McLaurin is absolutely being groomed as the #1 for the team, the problem is the team is a total disaster. I see no reason to trade McLaurin. Barring injury he should only accrue more value over time. I look at those guys, when they sneak through, as outliers. There will always be some, no system is outlier proof, and McLaurin is one. So be it. The system you're using is one I use as well and it works far more often than it fails and it works better at sniffing out frauds than any other methodology so I continue to use and eat the rare outlier I miss out on.

[stoneghost28]

mock it all you want... the data is the data and ot goes back 10 years and counting.

a bit on the defensive, aren't we?
i was merely being facetious alluding to the fact that if two guys can't tick a box (physically, as on a piece of paper) they probably won't have enough brains to be successful football players. didn't have much to do with stats.

McLaurin didnt finish top 24 in ppr through week 16

right. he was 25th. top 24 in both standard and half-ppr. as a rookie. for the redskins, looking good. i'll put a flag on him.
it doesn't even go against your work, as if it is one or two over ten years doesn't really make a difference.

And it is not useful as a predictive tool for hits. only misses.

So you're taking your 10 wrs, shaving off 3 who are very likely busts, and making your choice from 1/10 to 1/7.

i'd like to expand on this point, asking if you have any numbers on the "reverse" (over the thresholds, still bust), and if the picture changes by adding draft capital, but you'll probably take that as a personal insult, so maybe we could talk about it after you've had a wank, or a spliff, or both.

Not defensive at all about it, and no need to be a tool about it with your last statement, but to answer your question,

Yes there are several WRs who are not predicted to be busts, but still bust. The spreadsheet constructed gives:
1. Predicted busts that bust (>94% accuracy)
2. Predicted busts that hit (<6%)

Then, for fun, it goes into 3. seeing how well it does at predicting hits. So, if you meet both thresholds (and the spreadsheet looks at two different groupings of thresholds), how predictive is it for hits. And it isnt predictive at all (<50%) This is the point most people seem to have trouble understanding: This is used as a way to rule people out, not rule them in. The best way to use this data is to take the likely misses off your draft board and then work with what's left. It narrows the field down, which increases your chance for drafting a hit.

I'd like to see draft capital factored in somehow however at this time figuring out how that works and what that looks like is the challenge. Higher draft picks are given far more opportunity than lower draft picks.

I can say, the only trend that stands out is that the majority of those not predicted to miss are drafted in the first 4 rounds. So, in my opinion, I think the NFL is very much aware of these same statistics and use them in their own draft evaluations.

However, not all hits are early draft picks; which is why this isnt very predictive for hits (this is where people have trouble understanding how to use it and what its saying, but for the most part peoppe who have responded here seem to understand)

Maybe I can break it down by 1st, 2nd, 3rd, 4-5, 6-7 Round and figure out where most of the predicted hits come from, which would improve the predictability of hits. I've played with a few things to see what can affect the predicted miss accuracy, but that doesnt adjust it enough to make it worth whole (and in my opinion 94+% is pretty damn good by itself).

In the end, I dont blame someone for drafting McLaurin, Deebo, Hardman, or even Ruggs this next year... but that person wont be me. I'd rather not hook ly wagon to a guy who has a <6% chance of panning out. I'll take missing on the Tyreek Hills and Michael Thomas', because in that process I'm also avoiding wasting a pick on far more guys that end up busting.

Nothing will ever be fool proof. In some professions (like the medical field), something with 75-80% predictability is considered amazing, so I take 94% to be pretty darn good. Like I said, I can accept the ones that sneak through and end up great because in general they are few and far between. Those that squeak through tend to only have 1-3 years as a top 24 wr... the longevity is even less likely with those guys too (which is why I think McLaurin and Deebo are excellent sells).

When I'm spending a 1st or 2nd round pick on a rookie wr, I want someone who will be a top 24 wr pretty consistently. Some like to try and shoot for the stars with every pick and that's fine. Different strokes and all.

That is the point Ive always had such a hard time articulating, and you've summarized it quite well. It really is the chief point.


On Draft Capital as a means of sorting guys:
It's why draft capital remains the best predictor of anything, and it's not even a fair sampling in that sense since the subjects being tested aren't being tested in an equivalent environment. Day 1 picks get more chances/reps than day 2 guys, both day 1 and day 2 guys get more chances/reps than day 3 guys, and everybody gets more chances than UDFA's. We aren't dealing w/an even playing field ever, which is why draft capital remains a more reliable way to project future success, at least part of it is just simply based upon the opportunity offered. The more a team is invested in you in terms of draft capital, and salary cap dollars etc the more chances you get. Your chances to break through can depend upon organizational philosophy (a team like Seattle will give UDFA's and day 3 guys more rope than a lot of teams for instance), and they can also depend upon organizational depth at the position and injury luck. One can go on and on....

[stoneghost28]

Honestly really liked how you presented your case. It may only be something small to think about, but often times it is really difficult to decide between WR's, and this definitely helps. Like you said the purpose is not to solely decide rankings by this method, which would be pretty ineffective, but to potentially cut a few prospects out of an already deep 2020 WR class.

One thing worth noting, and not scientific, just me combing through more than a decade of drafts after my redskins took Harmon (a guy I really liked last January (of '19), but got really concerned when he tested so poorly) was how few hits there are after round 4 at WR. They are extraordinarily rare. You get a good chunk in round 3, and round 4, and then the roof totally caves in at the position. Yes there are hits, but it's like McLaurin and that "crappy Dom/Breakout Age" 94% failure rate Goode noted. It's around that bad. If you are a WR and drafted in round 5, 6 or 7, the hit rate is just beyond appalling. I think I learned my lesson on this circa 2016-2017 rookie drafts where I occasionally threw fourth round rookie dart throws at WR's drafted in that Sarlac Pit zone of failure. After '17, I just prioritized TE, QB and RB fallers w/those picks or vets that fell into the rookie pile. So instead of going after WR's, I'd just draft Darnold's, Josh Oliver's, Kahale Warring's, Ian Thomas's, that sorta deal. It can lead to some roster clogging, but at the same time, I end up getting guys that make rosters and show signs of being eventual weapons or at least depth. Better than throwing a dart at the Ishmael Zamorra's of the world.

[stoneghost28]

The only analytics I believe in is that tape beats advanced metrics 9 out of 10 times.

Why?

Cuz the sample sizes are far too small and you aren't comparing apples to apples when you consider the fact that college teams can have such dramatically disparate personnel constructions, schemes, focuses.

Tape is garbage and no it doesn't. Period.

[stoneghost28]

The only analytics I believe in is that tape beats advanced metrics 9 out of 10 times.

Why?

Cuz the sample sizes are far too small and you aren't comparing apples to apples when you consider the fact that college teams can have such dramatically disparate personnel constructions, schemes, focuses.

I lean this way as well. Stats and analytics have a place in the process. However, for me I want to make the decision based on what I see, so I don't draft the next Jamarcus Russell....

See and this is exactly why I don't trust tape people. Jamarcus blew up because people looked at him and his arm and ignored literally everything else. They saw a huge bowl game, a monster arm on tape, and stopped thinking. Russell had literally zero to do with analytics.

In analytics, you can run into bias w/how you decide to pick and choose model's etc, but the #'s are still the numbers.

With tape, you are literally inviting bias into every single thing you do, from falling in love with RB's that run over people, and show "toughness inside," to arguing w/one another about separation, footwork, beating the press and all manner of nebulous bull----. I'm attracted to tape because like everyone, I can look at players, and see something I like, find the comp, and run with it, but that's also the problem with it, people are flat out lying to themselves when they pretend the tape is revealing reliability consistent things that give you an edge. There isn't an edge. People have been looking at these draft results and tape for decades and there's literally no advantage in any of it beyond getting more hits if you have more picks, and that alone. That tells you everything. People are just deceiving themselves with it.

I'm sure there are ways we can do the same with #'s as well. But I think there's a modesty to the #'s that provides an edge that you simply cannot get with tape.

[FantasyFreak]

The only analytics I believe in is that tape beats advanced metrics 9 out of 10 times.

Why?

Cuz the sample sizes are far too small and you aren't comparing apples to apples when you consider the fact that college teams can have such dramatically disparate personnel constructions, schemes, focuses.

Tape is garbage and no it doesn't. Period.

This is the problem, with these types of this discussions. People are way too entrenched on one side or the other. Just because something doesn't work for one person, doesn't mean it doesn't for others. The idea that "tape is garbage", simply shows that it doesn't work for you, because you have found it to be less serving a method than another. Doesn't mean it's like that for everyone, and vice versa. Tape without any statistical reference is bordeline useless, and you can watch tape and not like a guy despite the stats saying he's good, and be right to pass on him, but using one or the other exclusively with not cross referencing aren't what any successful NFL franchises are doing, nor is it most likely what most successful Fantasy Players are doing.

[stoneghost28]

The only analytics I believe in is that tape beats advanced metrics 9 out of 10 times.

Why?

Cuz the sample sizes are far too small and you aren't comparing apples to apples when you consider the fact that college teams can have such dramatically disparate personnel constructions, schemes, focuses.

Tape is garbage and no it doesn't. Period.

This is the problem, with these types of this discussions. People are way too entrenched on one side or the other. Just because something doesn't work for one person, doesn't mean it doesn't for others. The idea that "tape is garbage", simply shows that it doesn't work for you, because you have found it to be less serving a method than another. Doesn't mean it's like that for everyone, and vice versa. Tape without any statistical reference is bordeline useless, and you can watch tape and not like a guy despite the stats saying he's good, and be right to pass on him, but using one or the other exclusively with not cross referencing aren't what any successful NFL franchises are doing, nor is it most likely what most successful Fantasy Players are doing.

I think you'd have more of a point if you'd read my following point in my next post.

When I say tape is garbage, and it doesn't work, I'm talking about what we've learned from what teams have been doing from decades, and it's nothing good. It's basically 100% pure randomness and zero modesty about what's actually gained, and zero ability to actually measure, accurately, what your projecting, and how effectively what your doing works. It's incredibly nebulous, and the results are in w/what tape, and the NFL model has done and it's clear. There's been zero improvement. Looking at literally decades of drafts, what you get is only one trend: More picks=more hits.

That is my point. If you get under the hood w/tape people, what you get is disagreement on the very basics of individual players in all manner of ways and no ways to test any of it reliably. Again, drenched in visual bias. I don't deny that people using tape only approaches can find things: Kareem Hunt as I referenced earlier was a point for tape guys. The problem I have is that you get a big enough sample size and it's just white noise. For every Kareem Hunt, you get a pile of contact balance busts. You don't get guys eliminated to narrow your choices like Goode suggested, you just get people going back and forth about contact balance, quick feet, running inside, vision, or not vision, just like WR's, beating the press, using hands well, separation, yada yada.

Eventually it comes down to this. You can evaluate analytics models and figure out how effectively they work, or not, and when trends, rule changes or anything else make them less efficacious, and dump them and look for something better. It's efficient. Watching tape? I just see bias, bias, bias, and yes, I listen and read tape guys because I want to have that area covered, so I can at least be open to a future Kareem Hunt, but I listen to check a box, not because I think it's reliable, or w/o bias because I absolutely don't think it's either of those things, but I like to be thorough, and I like to question myself, even when I say things are garbage, I'm still sniffing, lol.