2021 RCF Mock Draft: Discussion, Trades and Miscellany

[Hurl Bruce]

179. Atlanta Falcons @raiders91sc
180. San Francisco 49ers @Randolphkeys
181. Kansas City Chiefs @Coyote850

 

[The Human Q-Tip]

I think to assume that all pro days are skewed by the exact same amount is foolish.

I don't think that's part of the assumption. For example, perhaps pro day times vary anywhere from -.02 and - .08 compared to combine times because of the bias introduced by manual timing. Adding a +0.5 correction to all times - even though pro days are not all skewed by the exact same amount - would provide a result closer to more accurate combine times in the vast majority of those cases.

That's useful information if you are in a situation where, for whatever reason, you are comparing Player A's combine time to Player B's pro day time.

 

[Jack Brickman]

I don't think that's part of the assumption. For example, perhaps pro day times vary anywhere from -.02 and - .08 compared to combine times because of the bias introduced by manual timing. Adding a +0.5 correction to all times - even though pro days are not all skewed by the exact same amount - would provide a result closer to more accurate combine times in the vast majority of cases.

That's useful information if you are in a situation where, for whatever reason, you are comparing Player A's combine time to Player B's pro day time.

Exactly. There's no perfect way to account for the individual differences between each pro day, but we have years worth of data of guys who ran at both their pro day and the Combine and we know that the average difference is .05 seconds. The only fair way to move forward is to add that to every player's pro day time, even if it might be slightly inaccurate in some instances.

And as you said, the goal here is to accurately compare players who only ran at a pro day to guys who ran at the Combine, which provides us with the most accurate possible data for forty times. If you don't add that time, this year would wildly skew the data going forward, and would also mess up potential player comps. The entire purpose of analytics is to attempt to predict outcomes and provide historical context, and this year would really screw that up if we took all these pro day times as gospel.

 

[Randolphkeys]

There's no perfect way to account for the individual differences between each pro day

And then...

...the goal here is to accurately compare players who only ran at a pro day to guys who ran at the Combine.

That's the problem with the original certainty and lack of background to the opinions you were throwing around. Two pages later, at least you have given background to why your grouping of data looks nothing like the data everyone else was using. That, I suppose, is progress.

 

[Jack Brickman]

That's the problem with the original certainty and lack of background to the opinions you were throwing around. Two pages later, at least you have given background to why your grouping of data looks nothing like the data everyone else was using. That, I suppose, is progress.

When, from a historical standpoint, Combine times are .05 seconds slower than pro day times, the only way to be as accurate as possible when there's no Combine one year is to add .05 seconds to everyone's pro day times. I'm not sure why this is so complicated. The goal is to provide context to all of these players and to accurately compare them to players in the past and in the future, and you can't just take pro day times as gospel when, again, they are, on average, faster than Combine times. If you're developing a database that attempts to predict what these players will be with the highest degree of accuracy, you need to take that information into account, especially when 90% of the times in said database are based on Combine times.

Player Profiler does that. I don't know if anyone else does, nor do I really care. You can disagree with their method. That's fine. But their method has years of data behind it, whereas you just seem to be contrarian because you're annoyed I criticized your fifth round mock draft pick. And their site is, in my opinion, the best resource for athletic and production profiles for skill position players out there. They're committed to being as accurate as possible, and they wouldn't be as accurate as possible if they just input pro day times without considering that they are indisputably less accurate than Combine times.

And all this because of Michael fucking Carter, a guy who will probably be totally irrelevant a year from now.

 

[Out of the Rafters at the Q]

I don't think that's part of the assumption. For example, perhaps pro day times vary anywhere from -.02 and - .08 compared to combine times because of the bias introduced by manual timing. Adding a +0.5 correction to all times - even though pro days are not all skewed by the exact same amount - would provide a result closer to more accurate combine times in the vast majority of those cases.

That's useful information if you are in a situation where, for whatever reason, you are comparing Player A's combine time to Player B's pro day time.

1) I reject the assumption that "pro day times vary anywhere from -.02 and - .08 compared to combine times because of the bias introduced by manual timing" (yes I know it's just a hypothetical, but it's used as an assumption for the conclusion you make one sentence later).
2) This artificially skews the data and it would be worth far more to just leave it raw.
3) To your last line, in this case you understand that pro days are typically faster than combine times. But then you also have the added information that this was a UNC pro day, who might not allocate as many resources to football as an Ohio State, and combine that with the slower-than-expected times across the board, and use your own judgment.

 

[Jack Brickman]

2) This artificially skews the data and it would be worth far more to just leave it raw.

In a vacuum, maybe. But if the point is to compare this year's prospects with prospects from the previous decade to try to get a handle over how to project their NFL careers, you need to account for the fact that basically all of those prospects from previous years tested at the more accurate Combine.

Is it a perfect system? Of course not. There is no perfect system in a year where there's no Combine and all we have to go on is pro days. Hell, there's no perfect system to begin with when you're dealing with so many unaccountable human variables like worth ethic, landing spot, opportunity, etc. Is it better than simply inputting pro day times into a database full of Combine times and expecting to get accurate projections? Yes.

At any rate, I've explained the reasoning behind this enough for one lifetime. Agree to disagree, I guess.

 

[The Human Q-Tip]

1) I reject the assumption that "pro day times vary anywhere from -.02 and - .08 compared to combine times because of the bias introduced by manual timing" (yes I know it's just a hypothetical, but it's used as an assumption for the conclusion you make one sentence later).
2) This artificially skews the data and it would be worth far more to just leave it raw.
3) To your last line, in this case you understand that pro days are typically faster than combine times. But then you also have the added information that this was a UNC pro day, who might not allocate as many resources to football as an Ohio State, and combine that with the slower-than-expected times across the board, and use your own judgment.

Upon what basis do you make that judgement if you don't know what the average variance even is?

 

[Randolphkeys]

No reason to take this in a personal direction. I really didn't think it was about Micheal Carter either. This podcast you referenced generates completely different numerical data than the websites everyone else reads leading up to the draft. Furthermore, their rejection of the standard every other draft website is using can then be applied to all the agility drills, complicating data for the entire draft. If you don't know why that led to a lot of confusion, good luck to you.

 

[Jack Brickman]

No reason to take this in a personal direction. I really didn't think it was about Micheal Carter either. This podcast you referenced generates completely different numerical data than the websites everyone else reads leading up to the draft. Furthermore, their rejection of the standard every other draft website is using can then be applied to all the agility drills, complicating data for the entire draft. If you don't know why that led to a lot of confusion, good luck to you.

Other drills at the Combine don't use the same technology the forty uses to clock the start time from the moment the player begins to move, and thus do not require adjustment. I assume there's a reason for this, but I couldn't tell you what it is.

Furthermore, it's not just a podcast. Player Profiler is a web platform used by most of the best analysts in the fantasy football community, and the podcast regularly hosts those analysts, who agree with their methodology in this regard. It's been discussed dozens of times on various podcasts under the Rotounderworld umbrella with various analysts who are employed by other sites. I'd wager more than a few NFL front office people use it as well, as it's very easy to use and presents you with an abundance of information in a clean, easy to read format.

As I said in my previous reply, if we were simply taking this year in a vacuum, then sure, the average variance between Combine and pro day times really doesn't matter, as we're only comparing 2021 prospects with each other. But if you want to compare these prospects from a historical context to try to find their best comparable players in an attempt to project their NFL careers, then you need to properly adjust for the fact that there was no Combine this year, and so average forty times will be slightly faster.

ETA: I guess technically the site also has athletic metrics for non-skill position players (I just looked up Myles Garrett, for example, who ironically has Clowney as his best comparable player), but there's far more production data for the skill guys available.

 

[Randolphkeys]

No, this Player Profiler metric isn't the standard.

CBS doesn't use it on their player profiles, NFL.com doesn't use it, Pro Football Focus doesn't use it, Tony Pauline and Pro Football Network doesn't use it... I really don't see why I have to defend the idea that most websites that people have been reading have used a standard metric that has existed in track and field for centuries. Start stopwatch, stop stopwatch.

This is like an episode of Rick and Morty at this point. Posting on RCF should not require living in the Player Profiler parallel universe.

But now we know why your data doesn't match everyone else's, and that's good enough to move on.

 

[Jack Brickman]

No, this Player Profiler metric isn't the standard.

Never said it was. But if you want to keep arguing a point I never made, you do you.

Good talk.

 

[Out of the Rafters at the Q]

Upon what basis do you make that judgement if you don't know what the average variance even is?

You're the one making the assumption. You want to manipulate the data by adding 0.05 seconds to each score. The null hypothesis is to not do anything to the data. You have to provide a reason why.

 

[Out of the Rafters at the Q]

In a vacuum, maybe. But if the point is to compare this year's prospects with prospects from the previous decade to try to get a handle over how to project their NFL careers, you need to account for the fact that basically all of those prospects from previous years tested at the more accurate Combine.

Since we're talking about one specific data point, wouldn't the best way to handle it be "He ran 4.50 at the North Carolina pro day"?

From there, we can make judgments about how that score should be adjusted to fit a historic combine score (if any adjustment at all is required).

Heck, maybe the best thing to do is to just give that a mark of uncertainty--rather than apply a specific adjustment to it.

I just think it's disingenuous to see a 4.50, and then claim it's really a 4.59 without really having anything to back that up.

 

[Randolphkeys]

You're the one making the assumption. You want to manipulate the data by adding 0.05 seconds to each score. The null hypothesis is to not do anything to the data. You have to provide a reason why.

What the CBS draft profiles have done is just allow a .05 variance for all NFL prospects this year. So, for example, they say Micheal Carter is between a 4.50 and 4.55.

I think this is the most responsible way to do it, because just adding .05 doesn't solve the human error.