ArmChair's Dynasty Rankings (1 Viewer)

2nd post here. Let's see how many people I can get to hate me right off the bat.

[SIZE=10pt]I started this a few weeks ago. It is far from perfect. Please keep that in mind and read what I’m doing if you're at all interested. I’m fully comfortable and open to tweaks. Once you finish reading, if you’d like to ask what a player’s ranking would be if “x” happened, post it. I’ll randomly choose some to plug in. You will have issues with some of this, I know that (Demarco, for example, but see bolded immediately below)[/SIZE]

[SIZE=10pt]As of this time, this is based on 2014 stats through week 14 (13 games played), current team, coach, and depth chart seating.[/SIZE]

[SIZE=10pt]I am one who believes going based on last year's, or even this year's production leaves room for error and focuses you in on the highs, not necessarily the lows. I wanted to create something that takes current performance and use several models to project future performance. There are dynasty rankings all over the interweb, and I just got tired of not knowing what they were based on. Not knowing if they are dynasty 3 years out, emphasis on the 1st year, etc. This was "inspired" by ZWK.[/SIZE]

[SIZE=10pt]I do not have an overall ranking at this time, just positional. This is based on half ppr (25 yards per point for passing, 10 per for rushing. TDs are standard scoring)[/SIZE]

[SIZE=10pt]What I'm doing.[/SIZE]

[SIZE=10pt]1. Taking this year's players stats and snap counts.

2. Taking this year's total team snap counts.

3. Progressing, or in many cases, regressing to the mean on a few metrics. Here's a few examples as of week 14 of 2014.[/SIZE]

[SIZE=10pt]Aaron Rodgers has played 92% of his team's snaps. He receives 61 snaps per game. He touches the ball (not literally, but run/pass) 61% of the snaps he plays. He touches the ball (run/pass) 37 times per game. For QBs I want to know what their stats (points) would be if they played 95% of their team's snaps. I also want to know what their stats (points) would be if they touched the ball (run/pass) 58% of their snaps played, so I drop/increase the % touches of their snaps played to 58%. The 95% and the 58% figures are my regression or progression to a more normal season. Based on these metrics, and a projected future snap count of the team, I can determine if healthy, and in a normal season, how many snaps and touches a QB would have.[/SIZE]

[SIZE=10pt]Matt Forte has played 91% of his team's snaps. He receives 63 snaps per game. He touches the ball 37% of the snaps he plays. He touches the ball 23 times per game. For RBs I want to know what their stats (points) would be if they played 70% of their team's snaps. I also want to know what their stats (points) would be if they touched the ball 40% of their snaps played, so I drop or increase the % touches of their snaps played to 40%. The 70% and the 40% figures are my regression or progression to a more normal season. Based on these metrics, and a future snap count of the team, I can determine if healthy, and in a normal season, how many snaps and touches a RB would have.[/SIZE]

[SIZE=10pt]Antonio Brown has played 95% of his team's snaps. He receives 70 snaps per game. He touches the ball 12% of the snaps he plays. He touches the ball 8 times per game. For WRs I want to know what their stats (points) would be if they played 90% of their team's snaps. I also want to know what their stats (points) would be if they touched the ball 10% of their snaps played, so I drop or increase the % touches of their snaps played to 10%. The 90% and the 10% figures are my regression or progression to a more normal season. Based on these metrics, and a future snap count of the team, I can determine if healthy, and in a normal season, how many snaps and touches a WR would have.[/SIZE]

[SIZE=10pt]Greg Olsen has played 94% of his team's snaps. He receives 67 snaps per game. He touches the ball 8% of the snaps he plays. He touches the ball 5 times per game. For TEs I want to know what their stats (points) would be if they played 80% of their team's snaps. I also want to know what their stats (points) would be if they touched the ball 8% of their snaps played, so I drop or increase the % touches of their snaps played to 8%. The 80% and the 8% figures are my regression or progression to a more normal season. Based on these metrics, and a future snap count of the team, I can determine if healthy, and in a normal season, how many snaps and touches a TE would have.[/SIZE]

[SIZE=10pt]4. Based on the above, I setup future snaps and touches.

5. I calculate their points per touch based on their current season production. As we all know, typically the less touches you get, the more your points per touch should be. Keeping this in mind, I determine their future points per touch. [/SIZE]

[SIZE=10pt]Examples of this:[/SIZE]


[SIZE=10pt]For QBs, if they have played less than 40% of their team's season snaps, then I automatically give them a .40 pts per touch value. If they have played in more 40%, I let their current points per touch value ride. For a guide, currently Rivers sits at .54 points per touch. Rodgers sits at .73 points per touch.[/SIZE]

[SIZE=10pt]For RBs, if they have played less than 35% of their team's season snaps, then I automatically give them a .5 pts per touch value. If they have played in more 40%, I let their current points per touch value ride. For a guide, currently Ingram sits at .65 points per touch. Charles sits at 1.03 points per touch.[/SIZE]

[SIZE=10pt]For WRs, if they have played less than 40% of their team's season snaps, then I automatically give them a 1.5 pts per touch value. If they have played in more than 40%, I let their current points per touch value ride. For a guide, currently Golden Tate sits at 2.07 per touch. D Thomas sits at 2.57 points per touch.[/SIZE]

[SIZE=10pt]For TEs, if they have played less than 40% of their team's season snaps, then I automatically give them a 1.35 pts per touch value. If they have played in more than 40%, I let their current points per touch value ride. For a guide, Tellus sits at 2.03 points per touch. Julius Thomas sits at 3.34 points per touch.[/SIZE]

[SIZE=10pt]6. Future Redraft Value[/SIZE][SIZE=10pt] - Using the "Future Points Per Touch" value, and their "Future Touches" (based on my progression/regression from item 3), I work to determine their future redraft value. To do this though, I need to eliminate the progression/regression that I also did in step 3 (this was a true performance progression/regression based on normalization of snap counts and percent touches of snaps). This is where the depth chart and my assigned factors come in. Obviously a QB1, RB1, WR1, or TE1 will play more snaps, which means more touches, etc. So, based on the current state of affairs I assign the player's [/SIZE]current[SIZE=10pt] depth chart seating to them. This can change week to week, it is what it is.[/SIZE]

[SIZE=10pt]My depth chart factors are critically important. I needed to go fairly deep, so I created several different levels of a depth chart. Sometimes a player is a really good rb, but he’s in a split time situation. Gio and Hill are placed here. RB1a, RB1b. RB1 is [/SIZE]solely[SIZE=10pt] dedicated for the Fosters, Murrays, Bells of the world. So forth and so on. I did the same thing with all positions.



For WRs, WR1 is reserved for the elite with little target competition. Calvin, AJ, Julio...they fit this mold. Sometimes in a situation, you have an elite, but the qb is so good that he shares the ball. Jordy as is Cobb. These are WR1as. I really struggled with where to put DT, but I do have him as a sole WR1 on his team, with Sanders occupying a WR1a. It is rare that I share values with a WR2. Jordan Matthews pre-Sanchez fit here.[/SIZE]

[SIZE=10pt]Based on their depth chart position I have placed, I use another regression to predict future snaps (if you recall earlier, I took them out). Clearly the higher you are on a current depth chart, the more snaps you play. The more snaps you play, the more touches you are likely to receive.[/SIZE]

[SIZE=10pt]My redraft value is calculated using the future points per touch value (performance regression included) * future touches (assume 16 game season and healthy) * depth chart factor (playing time regression included).

Oh by the way, if I want to work to try to get a player's value if that player would soon be a RB1, RB1a, RB1b, WR1, WR1a, WR1b, etc, then I'll plug in that value. I have a player like McKinnon sitting here.[/SIZE]

[SIZE=10pt]I automatically rank them for redraft (Column A on the Stats tabs) based on their Redraft Value that I calculated.
[/SIZE]

[SIZE=10pt]7. Dynasty Final - My dynasty rankings follow this process, and takes their Redraft Value (not ranking), and does a bunch of crazy stuff with other factors, including Age, Coaching Hot Seat, Coach Philosophy (run, balanced, pass), and Personal Evaluation.


For dynasty purposes, I begin applying degradation factors for QBs at age 32. RBs at 27, WRs and TEs at 30. These are set as sliding scales and clearly up for debate. But hey, these are my dynasty rankings and I want to look slightly past the next 2-3 years.[/SIZE]

[SIZE=10pt]Coach philosophy is set to perform another set of regression/progression. No two teams are really alike. A WR on a run heavy team should probably be valued less than a WR on a pass heavy team. The opposite goes for RBs, etc.

Personal Eval is what I think of a player and his abilities. Obviously this is subjective.

At some point, I will add a contract factor to this, and also a "measurables" factor. I will also include NFL draft position in this, as well as taking an average of all of the expert’s dynasty rankings. I’m also still thinking through how to make a value dependent on other talent around him, but specifically QB (Manning). I’m not sure how just yet as there’s that long-time debate on the link between the two. [/SIZE]

[SIZE=10pt]This is a work in progress and far from perfect. I have yet to try to delve into player's past performances, but this, for the most part takes situation, performance, opinion, and injury/suspension issues into the equation. But there are unique situations, like Tyler Eifert. He has yet to really play this year, so I have no stats to go on. I used his games played last year to determine where he would sit and his current team’s play count to fill out the rest. Peterson, I left as is for this year.[/SIZE]

[SIZE=10pt]There are probably players that have been left off of this. Please let me know if one stands out that should be on it. Injuries are the most likely reason they are not on this. I do not yet know how I will incorporate next year’s rookies into this before they begin playing (maybe I'll try to use ZWK's draft rankings). Same for guys like Duron Carter.[/SIZE]

[SIZE=10pt]Enjoy. It’s just food for thought and I was bored.[/SIZE]

Please see post 29 on this thread if you have chimed in, or would like to chime in and have input on how to perform a calculation. The post is here. http://forums.footballguys.com/forum/index.php?showtopic=722641&p=17634581

Welp, I suck. Let me figure out how to post tables or come up with another option. :(

Ok, I still suck, but do believe I've been able to upload to google.

Rankings temporarily removed.

You could link to a google spreadsheet?

slackjawedyokel said:

You could link to a google spreadsheet?

Click to expand...

Yep. Got it. Thanks for the suggestion. Showing that I'm a rook.

Interesting. Flattered to have been the "inspiration."

It looks like your procedure for making the rankings is relatively systematic, but I don't have that clear a picture of how your method led to these rankings. Maybe you could pick a player (or two at the same position) and sketch out how your method led to them being ranked where you have them?

Looking through your rankings, some players seem weirdly out of place. For example: Palmer (QB13), Bowe (WR37), Marqise Lee (WR62), Avant (WR68). If you want to get 2 birds with one stone, you could use some of these guys as the example players.

CJ Anderson #3 RB in dynasty?

ZWK, sure, I can do that.

Jrod and Workdog, I indicated that it was a mostly static look at this time based stats to-date, current coach, depth chart, and age. For it to be accurate, the depth chart has to be accurate. I agree with you on those, but am not comfortable projecting future depth chart for them until after the season or maybe even FA/draft. In addressing the players mentioned, I will give you their ranking based on some depth chart movement below. If you think I should, then I can rework the depth chart to manually account for what I think happens for all players. As mentioned though, this is far from perfect. One of my biggest issues is working through the points per touch regression that one should have given a smaller sample size. For instance, Stedman Bailey. I don't know what to do with him just yet. On the field this year, he's shown well at 2.26 per touch. But since he hasn't played enough, I give him 1.50 per. I don't think this is right, but not sure what to do about it yet aside from manually changing a future points per touch value. I'd like to be able to do a week by week regression on snap count and touches per game, but I don't think I want to work that hard.

CJ Anderson

Currently have CJ as the RB1 on his own team. This has changed, what, 4 times this year? But this only takes facts into the equation, for the most part. I have evaluated him as Good. Excellent and Great are higher levels. It is an extremely small sample size, and it does assume based on him being RB1 on his own team, in addition to Denver's current snap pace over the season and that regressing next year. He is on pace to play 433 snaps this year (that includes him missing time, rbs ahead of him, etc.). Over a full season, I project him (if he stayed RB1 on his own team) to play 816 snaps. This gives him 50 snaps per game. He currently touches the ball 41% of the time he is in the game. 40% of the time is a good number to throw at it and indicate how many times a true RB1 on his own team would touch the ball in his given snaps. This gives him 20 touches per game. Based on his 141 touches this year, he is scoring .92 points per touch. Charles was 1.04! He is 23, so no age factors hurt him. If I move him to RB1a on his own team, he moves down to 8 in redraft, and 11 in dynasty. If I give him RB1b, it is 17 and 20 respectively. RB2, it goes to 30 and 30.

Palmer

Currently I have Palmer QB1 on his own team. I don't think Stanton or Thomas replace him. Palmer this year in this scoring system had .52 points per touch. Brees has .54 per touch. Palmer is a year younger. Both are with pass happy coaches. Brees is an excellent talent, Palmer was at one point, but I simply have him as good for this sheet. They are 2 rankings apart in my spreadsheet I know, it sounds crazy. Palmer touched the ball 57% time he was in the game, Brees touches it 61%. Earlier I mentioned that I will regress this down to 58%, so Palmer increases in attempts, and Brees decreases. Also remember to keep in mind that being a year younger means one more year to be playing. I know, large assumption. But for this exercise that is the case. This allows Palmer to hang in there close to the top 10. Finally, for next year, I project Brees to have 361 points, and Palmer to have 301 at this time, in this scoring system.

Tate/Alshon

Please remember this is half PPR, so PPR guys aren't just dismissed in this system. Tate is a PPR guy, but also also greatly benefited from Calvin being out a good portion of the year. Tate has 2.07 per touch so far this year, Alshon 2.30. Yada yada yada, snap count normalization, etc. Tate has 87 touches on the year, Alshon 79. Tate has performed better than him in leagues this year (with this scoring system). I also had Tate as a WR1a on his own team, but I did give Calvin WR1 status. While working through this for you, I moved Tate to WR1b, which put him at dynasty WR20.

Lee

Even if I give him WR1a depth chart, he only moves up to WR37 in dynasty. This is based on performance when he has touched the ball this year, so just keep that in mind. I do believe this is where my personal eval can change it. If I gave him an eval of Great, he goes to WR28.

Way too low on Carr.

opensourcebeer said:

Way too low on Carr.

Click to expand...

Thanks. I've got 67 formulas in this thing. I need these items pointed out to fine tune. No reason for him to be that low.

So I had him as an Average talent for now until I see more. If I move him up to "Good", he goes to QB22. He gets .34 points per touch as it stands now with his season, which isn't good at all. In fact, that points per touch puts him at QB35 for those qbs that have played over 30% of their team's snaps this year. His coach is also on the hot seat and currently a run oriented offense. This hurts him for now, but with a new coach who has a balanced or pass happy philosophy, it will help him.

armchairffqb said:

opensourcebeer said:

Way too low on Carr.

Click to expand...

Thanks. I've got 67 formulas in this thing. I need these items pointed out to fine tune. No reason for him to be that low.

So I had him as an Average talent for now until I see more. If I move him up to "Good", he goes to QB22. He gets .34 points per touch as it stands now with his season, which isn't good at all. In fact, that points per touch puts him at QB35 for those qbs that have played over 30% of their team's snaps this year. His coach is also on the hot seat and currently a run oriented offense. This hurts him for now, but with a new coach who has a balanced or pass happy philosophy, it will help him.

Click to expand...

I can see where you're coming from a formula standpoint. But at some point you have to take into account that guys like Stanton, Henne, Gabbert, Anderson, and Gradkkowski - all of which you have ranked for dynasty purposes above Carr - are NFL backup caliber players.

Even if you look at guys that have had decent production this year (Orton/Fitz), I'd never consider trading Carr for either of the guys. The potential ceiling with Carr far out weighs the known ceiling of marginal, journeymen NFL QBs..

opensourcebeer said:

armchairffqb said:

opensourcebeer said:

Way too low on Carr.

Click to expand...

Thanks. I've got 67 formulas in this thing. I need these items pointed out to fine tune. No reason for him to be that low.

So I had him as an Average talent for now until I see more. If I move him up to "Good", he goes to QB22. He gets .34 points per touch as it stands now with his season, which isn't good at all. In fact, that points per touch puts him at QB35 for those qbs that have played over 30% of their team's snaps this year. His coach is also on the hot seat and currently a run oriented offense. This hurts him for now, but with a new coach who has a balanced or pass happy philosophy, it will help him.

Click to expand...

I can see where you're coming from a formula standpoint. But at some point you have to take into account that guys like Stanton, Henne, Gabbert, Anderson, and Gradkkowski - all of which you have ranked for dynasty purposes above Carr - are NFL backup caliber players.

Even if you look at guys that have had decent production this year (Orton/Fitz), I'd never consider trading Carr for either of the guys. The potential ceiling with Carr far out weighs the known ceiling of marginal, journeymen NFL QBs..

Click to expand...

Rankings updated. Tweaked the formula a bit. There's still going to be one-offs. For guys like Carr, and Lee, the personal eval column is critical. Do I believe Lee and Patterson are extremely talented? Absolutely. Has their on-field play shown that? Not yet. So, I have to determine if my initial evaluation of them while in college is correct, or do I use their play this year to determine what I think of them as a player?

I agree. That is the most difficult part of this.

Michael Floyd is an interesting case study. Where would you put him in dynasty? And, where would you draft him in redraft after this year?

Ranking Logan Thomas ahead of Teddy Bridgewater makes me sad.

Donnybrook said:

Ranking Logan Thomas ahead of Teddy Bridgewater makes me sad.

Click to expand...

Think you might have the older version before the tweaks. I have him slightly behind those two.

I would keep in mind though, that Bridge, Mett, and Bortles have .37, .40, and .33 per touch this year, which is not good at all. QBs arguably improve the most during the course of their careers, but they aren't off to a great start (like most rook qbs). Will be interesting to see how Johnny looks this weekend.

=

First of all you have done some nice work here. Making projections for rankings is an ambitious task. There are many different ways to approach doing this.

I think what you are doing is closer to a possible redraft ranking than it is a dynasty ranking right now based on the data you have put into it.

You main method of projection being points/snap with snaps extrapolated from 2014 utilization.

There are problems with projecting this way however as no single season is like another. There are factors that go into a player getting offensive snaps and touches that will not be repeated the following season.

For example injuries either to the player in question or to other supporting cast that affected the snaps for each player but may change in following seasons. A coaching change is another thing that can affect this, so I think it is a good idea to keep track of coaches and coaching philosophies as part of your projection/ranking

Points/Snap to me is something to be used for small sample size weekly projections. Using this type of projection as you have for full seasons is going to cause a lot of anomolies

A larger sample size could improve the accuracy of your projections. I usually try to use a players career stats over the past 3 seasons whenever possible. If a player has four years in the league then I will usually throw out rookie stats. Sometimes I throw out rookie stats even if they do not, the trick here is to try to have your data be consistent with data used for players whos careers do not offer as large a sample size.

I also think it can be problematic to project using stats based on points scored. As this is only correct for the scoring system you apply it to. If you project snaps targets and touches then why not use the receptions, yards and TD numbers for your projection so that all of it can be translated into multiple scoring formats instead?

I think it is easier to check your work when you are dealing with actual stats rather than points derived from those stats.

It looks like you're trying to project a single future for each player, when what you really want is a probability distribution over possible futures.

CJ Anderson is his team's RB1 right now, and he might remain their RB1 for years to come, or he might tumble back down the depth chart. Stedman Bailey has been a 2.26 per touch guy on the field in a small sample size; he might actually be a 2.26 per touch guy, or he might just be a standard 1.50 per touch guy, or he might be something else. Marqise Lee is just a rookie, and we don't know yet if he'll turn out to be a great receiver, or a good one, or an average one, or worse. Lee has been his team's WR3 or WR4 most of this year (playing behind Shorts, Robinson, and Hurns when they were healthy), but if he pans out in terms of talent then he'll probably take on the role of WR1 or WR1a. Emmanuel Sanders might be Peyton Manning's WR1a for the next few years, or he might leave Denver in a year, or he might find himself on the post-Manning Broncos. And so on.

For many players (especially young ones), most of their expected fantasy value comes from a relatively optimistic possibility which isn't all that likely. More likely than not, Carr is an average QB (or worse) and will never be a top 12 fantasy QB. But there is maybe a 10% chance (to make up a number) that he turns out to be a great QB who is consistently a fantasy QB1, and so he's worth owning to see if that 1 in 10 possibility comes to pass. We could debate over whether he's currently the 28th best QB in the NFL or the 18th best, but that actually doesn't matter that much (neither one would make him a fantasy starter) - the important thing is how likely it is that he becomes one of the 10 best. That's why I'd like to own him over guys like Eli Manning or Alex Smith, who have established themselves as competent NFL passers but are very unlikely to become a consistent fantasy QB1 because there is much less uncertainty about how good they are - they have a long track record of being outside that range.

It's very hard to set up a systematic, quantitative way of doing dynasty rankings which captures the range of possibilities in the right way, instead of just focusing on the median case or projecting the current situation to continue. I haven't tried to do it; I use my intuitions for my dynasty rankings (while trying to ask myself the right questions, and looking into relevant numbers to help inform my judgments). But that seems to be the main thing that your current method is missing.

You need to establish if there is any correlation between the stats/benchmarks you've chosen and future production. Otherwise, you are answering the question "What if...?" not "What will...?"

ZWK said:

It's very hard to set up a systematic, quantitative way of doing dynasty rankings which captures the range of possibilities in the right way, instead of just focusing on the median case or projecting the current situation to continue. I haven't tried to do it; I use my intuitions for my dynasty rankings (while trying to ask myself the right questions, and looking into relevant numbers to help inform my judgments). But that seems to be the main thing that your current method is missing.

Click to expand...

I do agree with this. You've shown me your project before, and I admire the ambitiousness and the sheer scope involved, but I default towards skepticism of ranking-by-formula methods. I love the value of formulas in terms of challenging our preconceptions and drawing attention to facts that we are overlooking, but in terms of assembling the rankings themselves, I'm firmly on #TeamHeuristics. I think there's too much complexity to account for it all mathematically, and I worry that trying lends a patina of false confidence. I mentioned before the scientific concept of significant figures, how scientists have to be careful when doing math to preserve the amount of significant figures so as not to overstate their true confidence. I think, at their core, the end result of any ranking should reflect a fuzzy and vague process.

I also think that there's a lot of danger in perceiving formulas with subjective inputs as providing objective outputs. I think the Derek Carr and C.J. Anderson examples are particularly illustrative of this point. If you change Carr's talent level to "great", or Anderson's positional designation to "RB1a", their ranking changes, (often dramatically). These are subjective factors, and no matter how many objective manipulations you perform to them, the end result remains a fundamentally subjective output.

I really don't want to come off as too negative, because I think what you're doing here is very, very cool and potentially massively useful. It's like ZWK's "generic prospect rankings"- I'd never use them to actually rank rookies, but they provide a massive amount of value when it comes to informing my own process. And I think your regressions are really picking up on some contextual info that a savvy ranker could do a lot with probabilistically. Basically, I think what you're doing is massively value as a starting point to a ranking process, but my default reaction to it as an ending point to a ranking process is... skepticism.

Not that I'm right. I'm often mistrustful of things I don't fully comprehend, or things that are new. That's the human condition. It's possible that I'm not really apprehending everything involved, or that the accuracy and value added of your method would destroy my expectations. And by no means do I mean to discourage you from continuing- if you do continue to update and continue to refine your process, this is absolutely something that I would refer to frequently as a check on my own process. I think this is an awesome input that I could do a lot with, and I appreciate you sharing it.

Like the attempt but something is wrong with process. Steven Jackson as a top 20 back? Steve Smith is too high.

I agree with all of you. I had to absolutely mix in some subjective values into it to combine with the stats. But, no matter what I put for Patterson, he remains low because his performance has been terrible, along with his actual volume and projected volume as of this time. That one is for you, Adam.

The thing I really do like about this data (and I again, I mentioned it wasn't nearly perfect, but it is food for thought and uses a good amount of stats) is that I have snap regressions built in. I don't think it is reasonable to think that Forte, nor his replacement, would continue that pace of touches per snap.

I think more than anything, behind the scenes, I can see who is really producing based on volume, or who has unsustainable points per touch. And Adam, for you I can easily (and most of us can) sort through and see who should have tds come based on yards.

There's a lot more that I want to do, and I'm not sure I will. From what I'm gathering right now, no matter how in line many of the players are, there's just too many one-offs that kill the thing, or where my determination of their talent or depth chart skews it. Antoine Smith is actually one of these, and I'm sure you all know why.

I do think that I can take the average dynasty rankings of these players, remove my talent assertion and even depth chart models to let stats at least indicate translated talent.

I dunno. Probably shouldn't have posted it, but was bored and did want feedback.

Thanks to all that chimed in!

msudaisy26 said:

Like the attempt but something is wrong with process. Steven Jackson as a top 20 back? Steve Smith is too high.

Click to expand...

I agree. I need to increase the age factor I have applied for RBs.

I have another direction I could go with this relatively easily, but I pretty much hate it. My true dynasty value for guys like Manning, Smitty, and Jackson is pretty much 0. Redraft obviously that is incorrect, but based on so many years of data and age regression, you just can't say they are dynasty plays. Of course, that is when your individual goal in dynasty comes into play. Adam has Manning high, and it's hard to blame him if you're looking to win immediately. I always drafted dynasty to set myself up for stability and domination without a ton of year to year maintenance. I fail at that often, though.

Yep something in your method should be pushing Steven Jackson down, at least in the dynasty ranks. Any way to increase impact of age or anticipated playing years?

slackjawedyokel said:

Yep something in your method should be pushing Steven Jackson down, at least in the dynasty ranks. Any way to increase impact of age or anticipated playing years?

Click to expand...

I can easily do that, and am doing that now. But, I'd rather actually reflect the anticipated depth chart after the season. Guys like Cmike, Anderson, Smitty, Hyde, etc, would be greatly affected by this because at that point, I'm actually creating their work volume and combining that with their regressed performance,

Other guys like Allen, Sjax, and others need some TD regression built in. No clue how to really do that yet.

It was a start, hopefully a decent attempt at using some sort of stat and situation based rankings.

Lost me at ranking Greg Olsen #7 TE while slotting JT at #3..Thomas scores TDs doesnt catch many passes. Olsen does both..You're assuming Cobb re-signs with GB which is doubtful, they didnt pay Greg Jennings..therefore, Davante Adams belongs on that top 25 WR list. Alshon at 18 was another funny one..

armchairffqb said:

I agree with all of you. I had to absolutely mix in some subjective values into it to combine with the stats. But, no matter what I put for Patterson, he remains low because his performance has been terrible, along with his actual volume and projected volume as of this time. That one is for you, Adam.

The thing I really do like about this data (and I again, I mentioned it wasn't nearly perfect, but it is food for thought and uses a good amount of stats) is that I have snap regressions built in. I don't think it is reasonable to think that Forte, nor his replacement, would continue that pace of touches per snap.

I think more than anything, behind the scenes, I can see who is really producing based on volume, or who has unsustainable points per touch. And Adam, for you I can easily (and most of us can) sort through and see who should have tds come based on yards.

There's a lot more that I want to do, and I'm not sure I will. From what I'm gathering right now, no matter how in line many of the players are, there's just too many one-offs that kill the thing, or where my determination of their talent or depth chart skews it. Antoine Smith is actually one of these, and I'm sure you all know why.

I do think that I can take the average dynasty rankings of these players, remove my talent assertion and even depth chart models to let stats at least indicate translated talent.

I dunno. Probably shouldn't have posted it, but was bored and did want feedback.

Thanks to all that chimed in!

Click to expand...

I'm quite glad you posted it. Like I said, there's a lot of great stuff here, and I always think that whenever someone brings new ideas to a sharp community like this, everyone winds up a little better off- both the community and the idea-haver.

If it were me- and, to be clear, this is absolutely not me- I would more or less keep staying the course. I think you're producing something valuable by bundling all of that regression together. Like Jrodicus said, you should probably test your regression values to make sure they are maximally predictive, but that's just tweaking the values inside the framework, not overhauling the entire framework.

Yeah, sometimes you're going to get some crazy one-off problems like C.J. Anderson or Derek Carr or Antoine Smith. But if you're just using this as an input rather than a final product, that's not so much a problem. In fact, those one-offs can be the most valuable takeaway from formulas like this. I remember in 2004 when Football Outsiders first created a projection model attempting to predict year N+1 performance from teams based on year N performance. San Diego was coming off of a horrible 4-12 season, and one of their beat writers had quipped that he checked San Diego's schedule to see who the easiest team on it was, and that team was San Diego. Nevertheless, FO's numbers kept spitting out a projection that San Diego was going to win the AFC West. They kept tweaking the formula and trying to understand what was going on, and it kept saying that San Diego was going to rebound in a huge way. So eventually they just presented the data and said "look, this is what the formula is telling us, it probably looks a little crazy- it looks crazy to us, too!- but we're just telling you what it's telling us." And San Diego went 12-4 and won the AFC West.

I see that your methodology loves C.J. Anderson, and I start thinking "am I underrating C.J. Anderson?" I see that it hates Derek Carr and think "am I overrating Derek Carr?" I may not necessarily take what it says at face value- again, I think it's an input, but not a finished product- but it's certainly going to inform my expectations to some degree. And who knows, maybe some of those crazy-looking rankings are going to prove prescient and you'll curse yourself for trying so hard to "fix" them.

So yeah, if it were me, I'd plug in the numbers, let the chips fall where they may, and then say "okay, that's what the objective numbers say, and here's my subjective take where I incorporate those numbers but smooth out some of the rough edges." Like ZWK's generic prospect rankings on steroids.

I'm glad you stepped out and presented your work. You could always add a column for an adjuated ranking taking into account the intangible scoring. In a spreadsheet like this, you really can show rankings from various angles. Good work so far.

Perhaps, like with Fitzpatrick ahead of Carr, you need to look at the probability of either of them being a top 10 qb as opposed to who will be better next year. Someone could say Fitzpatrick is likely to outscore Carr (whether you agree or not, say it's hypothetically true) but that is kind of irrelevant because there's no upside to Fitzpatrick but there is for Carr. Somehow you need to factor in the upside, which is obviously somewhat subjective. Perhaps part of upside could be draft position (higher=more upside) years in the league (more= less upside), etc, etc. It may help smooth out some of these obvious misses.

Great input guys. I'll work on it today while watching my teams lose.

Adam, that actually helped confirm the direction I want to go. For me this was initially going to be done to help me identify under-the-radar guys who might just need the chance. Or the opposite and see who should really regress based on snap counts out of the norm.

Unreason, very good point. Perhaps that is what I should use as the personal eval slot, or in combination with it. Maybe even another set of rankings that would show the futuristic predictive depth chart, not a static "as it stands now with no real positional changes" view.

It would be difficult to do, but perhaps I plug in the final last year numbers and try to recall where I stood on player eval, and a few other various factors to produce what the dynasty rankings might have been coming into this year as a test.

Picking this back up. I've removed the rankings from the OP.

Really struggling with the talent eval factor. I wanted to try to place the factor based on their draft position, but that's really only good for rookies, and maybe 2nd/3rd year players. Using that for established players (Rodgers, Brees, etc) is pointless. Additionally, a WR taken in the 1st is the equivalent to a RB taken in the 2nd these days (rough guess on my part). A QB taken in the top 5-10 is quite different than a QB taken in the late 1st.

Since I haven't worked through that, I thought about using the PFF grades, but again, that's really only good for those who have enough snaps.

Suggestions are needed.

Option 1 - My assessment of their talent (Excellent, Great, Good, Average, Not Good, Bad, Terrible)

Option 2 - NFL Draft Round

Option 3 - PFF Rating

Option 4 - Average Professional Dynasty Ranking

Option 5 - Your suggestion

Keep in mind that I don't really want to do a ton of manual entries (inputting ages were enough for me!).

armchairffqb said:

Picking this back up. I've removed the rankings from the OP.

Really struggling with the talent eval factor. I wanted to try to place the factor based on their draft position, but that's really only good for rookies, and maybe 2nd/3rd year players. Using that for established players (Rodgers, Brees, etc) is pointless. Additionally, a WR taken in the 1st is the equivalent to a RB taken in the 2nd these days (rough guess on my part). A QB taken in the top 5-10 is quite different than a QB taken in the late 1st.

Since I haven't worked through that, I thought about using the PFF grades, but again, that's really only good for those who have enough snaps.

Suggestions are needed.

Option 1 - My assessment of their talent (Excellent, Great, Good, Average, Not Good, Bad, Terrible)

Option 2 - NFL Draft Round

Option 3 - PFF Rating

Option 4 - Average Professional Dynasty Ranking

Option 5 - Your suggestion

Keep in mind that I don't really want to do a ton of manual entries (inputting ages were enough for me!).

Click to expand...

I use option 1, heavily informed by option 2.

Here are some links about concepts of the fourth dimension such as quality years remaining, average career length, aging patterns

As far as the options you list.

Option 1 - This is perhaps the most challenging and important part of player evaluation. What I do is compile rankings of draftniks and other people willing to share those ideas. I then average those lists for a consensus. Prior to the combine, Post combine, then post NFL draft, which is where the draft position and player to team fit are considered.

Option 2 - There are a lot of studies using the draft as a guide to what are the odds of a player picked at a position to be successful, and how successful. But in the end good players come from every point of the draft. It certainly is a tie breaker for similar prospects, but I will not ignore a player that was ranked highly pre NFL draft just because they fell to a later round in the draft. You have several categories that you are cross ranking the players. Giving draft position an equal weight to say 6(or more) other categories would not be an overcorrection that throws out the other information too much.

Option 3 - I am not sure how you would do this, but again if this were one category out of many, perhaps that would be helpful.

Option 4 - I would try to complete your projections/rankings before considering others.

Option 5 - Another important concept is value above replacement. If you do end up projecting snaps/targets/yards/TD you can then calculate the expected points of players on the waiver wire based on the number of players rostered. This becomes the replacement level baseline which can help you use your roster spots more efficiently.

sub'd