using weekly variation in yearly projections (1 Viewer)

continuing a discussion from this thread - I was curious if there is a way to compare weekly variations between players...that is, assume you have two players rated equally during the pre-season, one might want to know player X is more consistent than player Y. I am still not sure if consistency is a good thing; that's a topic for further discussion.

Regardless, what I want to know is if a players variation can be statistically predicted for season N+1, based on seasons N, N-1, N-2, etc. If I can demonstrate that a weekly variance is similar to what he has done in the past, I can get a handle on how reliable this player will be this season.

So, I looked at some data going back to 2003 (I've been in the same MFL league since then and historical weekly data is readily available). Using some excel, I made a table with all players names in column A, and weekly production for 2003 week1, 2003 week2, 2003 week3, etc all the way through 2011 week16. Next, I ran a simple f-test, using 2003-2010 as population A and 2011 as population B. Obviously, this analysis doesn't work for rookies.

I found that out of 172 players, the 2011 variance was not significantly different for 148. That is, for 86% of all players I looked at, the 2011 standard deviation could be approximated by that players historical standard deviation. I think that's pretty cool.

Where this is applicable is discriminating between similarly ranked players. Say I have Marcus Colston and Brandon Marshall similarly rated (Dodds projects them just over a point apart). Which one do you draft? Well, Marshall has a career stdev of 8.4, and Colston has a career stdev of 7.62. From that, we can infer that in 2012, Colston is more likely to be consistent than Marshall will be.

now, this is not 100% reliable - it was good only 86% of the time for 2011. I still need to see how reliable it was heading into 2010, 2009, etc. And, as always, past results are no guarantee of future earnings and all that. I just thought this was interesting, maybe an analysis can spark someone more gifted with the statistics than I to work this a little better.

This is exactly what I was looking for, but adding a handicap to a players value due to variance. For VBD I'm not just looking at the highest scores for the year, but the most consistantly highest scorers. Intuitively people who play a lot do it, (I hate you jermichael Finley) but I think there has to be a way to handicap players for inconsistencies to further evaluate across positions.

'moleculo said:

continuing a discussion from this thread - I was curious if there is a way to compare weekly variations between players...that is, assume you have two players rated equally during the pre-season, one might want to know player X is more consistent than player Y. I am still not sure if consistency is a good thing; that's a topic for further discussion.

Regardless, what I want to know is if a players variation can be statistically predicted for season N+1, based on seasons N, N-1, N-2, etc. If I can demonstrate that a weekly variance is similar to what he has done in the past, I can get a handle on how reliable this player will be this season.

So, I looked at some data going back to 2003 (I've been in the same MFL league since then and historical weekly data is readily available). Using some excel, I made a table with all players names in column A, and weekly production for 2003 week1, 2003 week2, 2003 week3, etc all the way through 2011 week16. Next, I ran a simple f-test, using 2003-2010 as population A and 2011 as population B. Obviously, this analysis doesn't work for rookies.

I found that out of 172 players, the 2011 variance was not significantly different for 148. That is, for 86% of all players I looked at, the 2011 standard deviation could be approximated by that players historical standard deviation. I think that's pretty cool.

Where this is applicable is discriminating between similarly ranked players. Say I have Marcus Colston and Brandon Marshall similarly rated (Dodds projects them just over a point apart). Which one do you draft? Well, Marshall has a career stdev of 8.4, and Colston has a career stdev of 7.62. From that, we can infer that in 2012, Colston is more likely to be consistent than Marshall will be.

now, this is not 100% reliable - it was good only 86% of the time for 2011. I still need to see how reliable it was heading into 2010, 2009, etc. And, as always, past results are no guarantee of future earnings and all that. I just thought this was interesting, maybe an analysis can spark someone more gifted with the statistics than I to work this a little better.

Click to expand...

Colston has played with Brees - their situations have been very different. What does Marshall's production with poor QBing have to do with how consistant he will be in 2012?

I think it is perhaps a fallacy to assume that past consistency implies future consistency. There are many variables that can go into a players "consistency". I think that data is perhaps more relevant for players who stay in the same situation (team, offense, supporting cast). Overall, I would say that not only is it difficult to quantify consistency with respect to expected future performance, but the difference is in many cases too negligible to matter with respect to the fantasy game itself. In the example you used, Marshall has a stnd dev of 8.4 and Colston 7.62. While I suppose it would be fair to call Colston in this case "more consistent" I think that the data just shows that neither player is relatively consistent. The difference of .78 I would consider negligible in the grand scheme of things.

Just my $.02

'amicsta said:

I think it is perhaps a fallacy to assume that past consistency implies future consistency. There are many variables that can go into a players "consistency". I think that data is perhaps more relevant for players who stay in the same situation (team, offense, supporting cast). Overall, I would say that not only is it difficult to quantify consistency with respect to expected future performance, but the difference is in many cases too negligible to matter with respect to the fantasy game itself. In the example you used, Marshall has a stnd dev of 8.4 and Colston 7.62. While I suppose it would be fair to call Colston in this case "more consistent" I think that the data just shows that neither player is relatively consistent. The difference of .78 I would consider negligible in the grand scheme of things.Just my $.02

Click to expand...

consider this: stdev is calculated comparing games vs games...that's not a yearly metric. End of season totals is a yearly metric. Dodds has Colston @ 152.3, Marshall @ 153.5. On a per game basis, thats 9.52 PPG and 9.59 PPG, or a difference of 0.075 points per game. With that in mind, I think the stdev difference of 0.78 is plenty significant.Really, that's what I'm trying to do here - come up with a purely rational way to separate similarly projected players.

'moleculo said:

'amicsta said:

I think it is perhaps a fallacy to assume that past consistency implies future consistency. There are many variables that can go into a players "consistency". I think that data is perhaps more relevant for players who stay in the same situation (team, offense, supporting cast). Overall, I would say that not only is it difficult to quantify consistency with respect to expected future performance, but the difference is in many cases too negligible to matter with respect to the fantasy game itself. In the example you used, Marshall has a stnd dev of 8.4 and Colston 7.62. While I suppose it would be fair to call Colston in this case "more consistent" I think that the data just shows that neither player is relatively consistent. The difference of .78 I would consider negligible in the grand scheme of things.Just my $.02

Click to expand...

consider this: stdev is calculated comparing games vs games...that's not a yearly metric. End of season totals is a yearly metric. Dodds has Colston @ 152.3, Marshall @ 153.5. On a per game basis, thats 9.52 PPG and 9.59 PPG, or a difference of 0.075 points per game. With that in mind, I think the stdev difference of 0.78 is plenty significant.Really, that's what I'm trying to do here - come up with a purely rational way to separate similarly projected players.

Click to expand...

You cant get standard deviation without first having an average, which comes from a yearly total. The two are not mutually exclusive statistics. That aside, i still do not believe that gap is significant. .78 ppg is 12.48 total points over the course of a season, which could easily be scored in a myriad of ways. If Marshall were to score those additional points, then on average, even if marshall followed his std dev in the negative direction, hed still only tie colston if he hit his std dev in the positive direction. What would make this a complete study would be if you could determine how likely this outcome is to happening (scoring those additional points). From a common sense standpoint, i would say the variability from projections to actual points scored is high enough to render this insignificant. Establishing confidence intervals i think would be the answer, but figuring out what statistic to use would vary greatly the results. Id be interested in playing with these numbers myself if you could link the excel data somehow.

WR by the way is a fairly inconsistent position by its nature. We can tie consistency usually to play style (possesion guys are usually more consistent then homerun hitters) in some fashion just as reliably perhaps as the statistics could show. Just something else to consider

'amicsta said:

WR by the way is a fairly inconsistent position by its nature. We can tie consistency usually to play style (possession guys are usually more consistent then homerun hitters) in some fashion just as reliably perhaps as the statistics could show. Just something else to consider

Click to expand...

I'd be curious to see if this assumption is borne out by the statistics, but I think it would as it seems to be common sense. I would think that high reception guys would probably be more consistent than lower reception, big play type guys. Desean Jackson is a guy who would fit squarely into that second category. My guess would be that guys like DJax, Torrey Smith, Mike Wallace, etc. would have less consistency than Antonio Brown, Anquan Boldin, Wes Welker, etc. because the former are more reliant on big plays and TDs (more random) to score their points while the latter are more reception/yardage guys (less random). Colston and Marshall would both seem to be pretty similar players and I would expect that they are both pretty similar in consistency.I think the 2nd big factor in a players' consistency would be the number of other top targets on their team. For example, Marshall was pretty consistently solid last year with few huge games but almost always getting into double digits. I think part of that was the fact that he was the only real weapon on his team. He was going to get his 8-10 targets almost every single week. All of the Packers WRs (and Finley) were pretty inconsistent because there could be a different go-to guy depending on the matchup which led to some huge weeks and some real stinkers as well. They might have some 8-10 target games which with a much better QB were huge, but they might also have some 2-3 target games due to the many other weapons on the team.Based on this, I'd expect guys like AJ Green, Andre Johnson and Percy Harvin to be pretty consistent. Maybe a bit less consistency for the Green Bay guys, maybe New England this year with the addition of Lloyd and the Giants if Randle emerges along with Cruz/Nicks to be less consistent with huge games mixed in with single digit scores.

I think the metric you guys want to use is the crank score. He's a FBGuys article from Waldman a few years back... http://subscribers.footballguys.com/2009/09waldman_gut159.php

'bengalbuck said:

'amicsta said:

WR by the way is a fairly inconsistent position by its nature. We can tie consistency usually to play style (possession guys are usually more consistent then homerun hitters) in some fashion just as reliably perhaps as the statistics could show. Just something else to consider

Click to expand...

I'd be curious to see if this assumption is borne out by the statistics, but I think it would as it seems to be common sense. I would think that high reception guys would probably be more consistent than lower reception, big play type guys. Desean Jackson is a guy who would fit squarely into that second category. My guess would be that guys like DJax, Torrey Smith, Mike Wallace, etc. would have less consistency than Antonio Brown, Anquan Boldin, Wes Welker, etc. because the former are more reliant on big plays and TDs (more random) to score their points while the latter are more reception/yardage guys (less random). Colston and Marshall would both seem to be pretty similar players and I would expect that they are both pretty similar in consistency.I think the 2nd big factor in a players' consistency would be the number of other top targets on their team. For example, Marshall was pretty consistently solid last year with few huge games but almost always getting into double digits. I think part of that was the fact that he was the only real weapon on his team. He was going to get his 8-10 targets almost every single week. All of the Packers WRs (and Finley) were pretty inconsistent because there could be a different go-to guy depending on the matchup which led to some huge weeks and some real stinkers as well. They might have some 8-10 target games which with a much better QB were huge, but they might also have some 2-3 target games due to the many other weapons on the team.Based on this, I'd expect guys like AJ Green, Andre Johnson and Percy Harvin to be pretty consistent. Maybe a bit less consistency for the Green Bay guys, maybe New England this year with the addition of Lloyd and the Giants if Randle emerges along with Cruz/Nicks to be less consistent with huge games mixed in with single digit scores.

Click to expand...

I've always assumed that inconsistency is higher for guys that get their scoring through TD's... that's kind of a quick 6 points, that may or may not be consistent on a weekly basis. I'd rather have a guy with 4 receptions for 80 yards vs a guy with 2 receptions, 40 yards, and a TD. in either case, the overall score is the same, but IMO the 4-80 guy is more likely to score higher the following week.Maybe the way to generate the desired effect, then, is to forget about trying to use standard variation, and simply discount projected TD's across the board? If my league scores all TD's as 6, maybe when I plug in the parameters into my VBD app, I'll put in 3. The impact that would have is to drive yardage/receptions guys higher up in the rankings and TD heavy guys down. I'll have to play with that a little more.

'Late225 said:

I think the metric you guys want to use is the crank score. He's a FBGuys article from Waldman a few years back... http://subscribers.footballguys.com/2009/09waldman_gut159.php

Click to expand...

thanks, I hadn't seen that. I'd like to see where Waldman went with that - seems to me that knowing a standard deviation is required to come up with a crank score...if you have a projected st dev and a projected average, it's pretty easy to come with a probability that a player scores higher than a threshold. How you come up with that stdev is a different animal though...sounds like he wanted to use AVT, which I don't think is appropriate.

'moleculo said:

consider this: stdev is calculated comparing games vs games...that's not a yearly metric. End of season totals is a yearly metric. Dodds has Colston @ 152.3, Marshall @ 153.5. On a per game basis, thats 9.52 PPG and 9.59 PPG, or a difference of 0.075 points per game. With that in mind, I think the stdev difference of 0.78 is plenty significant.Really, that's what I'm trying to do here - come up with a purely rational way to separate similarly projected players.

Click to expand...

What I find interesting is that even if this proves to me right, and Colston is more "reliable" or whatever word you want to use.....that you'd then select him above Marshall. Maybe I view it a bit different, but I'd rather have Marshall and then start him vs. weaker defenses to capitalize on those high games that he will have. My MFL league (like I assume all do) has a "efficiency rating" where you're scored on how close to your "perfect lineup" you get each week, and I'm typically top 2 in my 10 team league.As such, for my 2 RB spots - I've typically always had a "stud" (currently Rice), and then a handful of other guys (currently Redman, K Smith, F Jackson, Spiller, J Stew, Ingram) and I start the one with the best matchup.

'matttyl said:

'moleculo said:

consider this: stdev is calculated comparing games vs games...that's not a yearly metric. End of season totals is a yearly metric. Dodds has Colston @ 152.3, Marshall @ 153.5. On a per game basis, thats 9.52 PPG and 9.59 PPG, or a difference of 0.075 points per game. With that in mind, I think the stdev difference of 0.78 is plenty significant.Really, that's what I'm trying to do here - come up with a purely rational way to separate similarly projected players.

Click to expand...

What I find interesting is that even if this proves to me right, and Colston is more "reliable" or whatever word you want to use.....that you'd then select him above Marshall. Maybe I view it a bit different, but I'd rather have Marshall and then start him vs. weaker defenses to capitalize on those high games that he will have. My MFL league (like I assume all do) has a "efficiency rating" where you're scored on how close to your "perfect lineup" you get each week, and I'm typically top 2 in my 10 team league.As such, for my 2 RB spots - I've typically always had a "stud" (currently Rice), and then a handful of other guys (currently Redman, K Smith, F Jackson, Spiller, J Stew, Ingram) and I start the one with the best matchup.

Click to expand...

I don't think the point is to rank one player over the other because he has a higher/lower st.dev. The point is that it's just an extra piece of information. Personally, I think the most use comes from Who do i start questions, where you can factor in the strenght of your team versus your opponents. If you think you have a stronger team then you might choose colston (assuming you're torn on Colston vs Marshall after taking into account other variables). If you think you have a weaker team, then you might choose Marshall to try to narrow the gap. I think it'd be a nice tool to have, but shouldn't be a major deciding factor.

'amicsta said:

You cant get standard deviation without first having an average, which comes from a yearly total. The two are not mutually exclusive statistics.

That aside, i still do not believe that gap is significant. .78 ppg is 12.48 total points over the course of a season, which could easily be scored in a myriad of ways. If Marshall were to score those additional points, then on average, even if marshall followed his std dev in the negative direction, hed still only tie colston if he hit his std dev in the positive direction. What would make this a complete study would be if you could determine how likely this outcome is to happening (scoring those additional points). From a common sense standpoint, i would say the variability from projections to actual points scored is high enough to render this insignificant. Establishing confidence intervals i think would be the answer, but figuring out what statistic to use would vary greatly the results. Id be interested in playing with these numbers myself if you could link the excel data somehow.

Click to expand...

That would be just one standard deviation. If we assume normal distribution (which I don't think it's the best fit but good enough for the discussion) the spread is actaully 3 times that which puts it at 37.5 (or at the minumum 2 times that if you only want to go to 95% of all possibilities). So Marshall would have a ceiling 25-37.5 points higher and a floor 25-37.5 points lower than Colston.

'Modog814 said:

'amicsta said:

You cant get standard deviation without first having an average, which comes from a yearly total. The two are not mutually exclusive statistics.

That aside, i still do not believe that gap is significant. .78 ppg is 12.48 total points over the course of a season, which could easily be scored in a myriad of ways. If Marshall were to score those additional points, then on average, even if marshall followed his std dev in the negative direction, hed still only tie colston if he hit his std dev in the positive direction. What would make this a complete study would be if you could determine how likely this outcome is to happening (scoring those additional points). From a common sense standpoint, i would say the variability from projections to actual points scored is high enough to render this insignificant. Establishing confidence intervals i think would be the answer, but figuring out what statistic to use would vary greatly the results. Id be interested in playing with these numbers myself if you could link the excel data somehow.

Click to expand...

That would be just one standard deviation. If we assume normal distribution (which I don't think it's the best fit but good enough for the discussion) the spread is actaully 3 times that which puts it at 37.5 (or at the minumum 2 times that if you only want to go to 95% of all possibilities). So Marshall would have a ceiling 25-37.5 points higher and a floor 25-37.5 points lower than Colston.

Click to expand...

exactly. of course, that assumes that Marshalls career stdev will be a good estimate for his 2012 stdev...that's step one, and right now, that's what I'm trying to prove.Another thing to consider, and this is what Modog814 was referring to, is I'm not sure if consistency is desirable. The Waldman article that Late225 posted claims that higher consistancy players will win more in the long run. However, I could see an arguement where in a deep, competitive league, the odds of winning are pretty slim to start with; playing it safe would be a great way to be average an barely miss the playoffs. The way you win is with high variability guys, all coming up big at the same time.

I suspect the optimal mix is somewhere in the middle.

'matttyl said:

'moleculo said:

consider this: stdev is calculated comparing games vs games...that's not a yearly metric. End of season totals is a yearly metric. Dodds has Colston @ 152.3, Marshall @ 153.5. On a per game basis, thats 9.52 PPG and 9.59 PPG, or a difference of 0.075 points per game. With that in mind, I think the stdev difference of 0.78 is plenty significant.

Really, that's what I'm trying to do here - come up with a purely rational way to separate similarly projected players.

Click to expand...

What I find interesting is that even if this proves to me right, and Colston is more "reliable" or whatever word you want to use.....that you'd then select him above Marshall. Maybe I view it a bit different, but I'd rather have Marshall and then start him vs. weaker defenses to capitalize on those high games that he will have. My MFL league (like I assume all do) has a "efficiency rating" where you're scored on how close to your "perfect lineup" you get each week, and I'm typically top 2 in my 10 team league.As such, for my 2 RB spots - I've typically always had a "stud" (currently Rice), and then a handful of other guys (currently Redman, K Smith, F Jackson, Spiller, J Stew, Ingram) and I start the one with the best matchup.

Click to expand...

I think this has more to do with the quality of your depth than it does your "skill" at selecting starters. My MFL league has this as well and mine usually isn't very great because I have a lot of quality bench players. Looking at last year, I had the best record, was #2 in overall scoring, but only 8/12 in efficiency rating for the year.

'Modog814 said:

'amicsta said:

You cant get standard deviation without first having an average, which comes from a yearly total. The two are not mutually exclusive statistics.

That aside, i still do not believe that gap is significant. .78 ppg is 12.48 total points over the course of a season, which could easily be scored in a myriad of ways. If Marshall were to score those additional points, then on average, even if marshall followed his std dev in the negative direction, hed still only tie colston if he hit his std dev in the positive direction. What would make this a complete study would be if you could determine how likely this outcome is to happening (scoring those additional points). From a common sense standpoint, i would say the variability from projections to actual points scored is high enough to render this insignificant. Establishing confidence intervals i think would be the answer, but figuring out what statistic to use would vary greatly the results. Id be interested in playing with these numbers myself if you could link the excel data somehow.

Click to expand...

That would be just one standard deviation. If we assume normal distribution (which I don't think it's the best fit but good enough for the discussion) the spread is actaully 3 times that which puts it at 37.5 (or at the minumum 2 times that if you only want to go to 95% of all possibilities). So Marshall would have a ceiling 25-37.5 points higher and a floor 25-37.5 points lower than Colston.

Click to expand...

Oh shoot, yea you're right. That's what I get for trying to crunch numbers when I'm half asleep haha.

I think you really have to find a mix to make this factor matter during your draft. I personally think 2 players of similar ranking and only a small margin of variance is not good reason to pick one over the other. My tie breakers are usually a Strength of Schedule or better overall team value. I'd rather have a low rank player on GB or NE than I slightly higher ranked player on JAX or CLE. However the best factor for deciding on players is actually research and study. The more you read on player personnel, coaches and overall team chemistry the easier it is to decide. FBG does an excellent job with spotlights.

Regardless, I'm working on creating an algorithm in excel that calculates key factors to give me one deciding number. It projects current year total pts, weighs a 5 year average for consistency (based on average per week), weighs their current age by position ((You'd be surprised how big the drop off is -- I encourage you all to do it. There is a magic number for every position based on age. You can use the historic stat dominator to find it out)), and weighs a 5 year VBD rank. In the end I'm hoping to create a scaleable calculator that will work for all my leagues, most importantly my keeper and dynasty leagues. Redraft is cake... cuz the only stat that matters is the projection and the dVBD.

I'm getting really close on this formula... I used an early version of it last year and I dominated the league... and when I say dominate... I mean it. Only lost 1 game. I barely even went to the wire, my guys were consistent and high reward. I even lost Frank Gore and Peyton Manning and it didn't even phase my team. I would say early tests are giving me the results I was hoping for. I think we should start a board just for the fantasy math geeks. Share formulas... isn't that the point of this site... dominate the competition. It's pretty difficult to find this info on the web.

I created a massive excel file that calculates everything off one page. I simply fill in the draft board in real time and it calculates everything from game-to-game match up, strength of position, auction values, and creates a cheat sheet depth chart that updates during my draft. Because it tracks the draft board it automatically counts my opponents players so I know their needs vs mine. This helps me get more value early instead of regretting a pick that I could have made in the next round. It's primitive only in that it's in excel... but the draft dominator feels limited to me... I still use it a ton for the export options. But I just want more settings so I can have more control and that won't happen any time soon... so I decide to create my own monster draft dominator. You need to know your excel to use it, as you would need to modify the file to fit your needs. My current league is 2QB/IDP/10 team.

If there are any mathematicians out there... we could create something very unique. As unique as what VBD was when it first came out. (I just read this... damn I'm a geek!)

'Jasonb said:

I think you really have to find a mix to make this factor matter during your draft. I personally think 2 players of similar ranking and only a small margin of variance is not good reason to pick one over the other. My tie breakers are usually a Strength of Schedule or better overall team value. I'd rather have a low rank player on GB or NE than I slightly higher ranked player on JAX or CLE. However the best factor for deciding on players is actually research and study. The more you read on player personnel, coaches and overall team chemistry the easier it is to decide. FBG does an excellent job with spotlights. Regardless, I'm working on creating an algorithm in excel that calculates key factors to give me one deciding number. It projects current year total pts, weighs a 5 year average for consistency (based on average per week), weighs their current age by position ((You'd be surprised how big the drop off is -- I encourage you all to do it. There is a magic number for every position based on age. You can use the historic stat dominator to find it out)), and weighs a 5 year VBD rank. In the end I'm hoping to create a scaleable calculator that will work for all my leagues, most importantly my keeper and dynasty leagues. Redraft is cake... cuz the only stat that matters is the projection and the dVBD.I'm getting really close on this formula... I used an early version of it last year and I dominated the league... and when I say dominate... I mean it. Only lost 1 game. I barely even went to the wire, my guys were consistent and high reward. I even lost Frank Gore and Peyton Manning and it didn't even phase my team. I would say early tests are giving me the results I was hoping for. I think we should start a board just for the fantasy math geeks. Share formulas... isn't that the point of this site... dominate the competition. It's pretty difficult to find this info on the web.I created a massive excel file that calculates everything off one page. I simply fill in the draft board in real time and it calculates everything from game-to-game match up, strength of position, auction values, and creates a cheat sheet depth chart that updates during my draft. Because it tracks the draft board it automatically counts my opponents players so I know their needs vs mine. This helps me get more value early instead of regretting a pick that I could have made in the next round. It's primitive only in that it's in excel... but the draft dominator feels limited to me... I still use it a ton for the export options. But I just want more settings so I can have more control and that won't happen any time soon... so I decide to create my own monster draft dominator. You need to know your excel to use it, as you would need to modify the file to fit your needs. My current league is 2QB/IDP/10 team.If there are any mathematicians out there... we could create something very unique. As unique as what VBD was when it first came out. (I just read this... damn I'm a geek!)

Click to expand...

Sounds very interesting. Does it do the projections as well or is that an input? I've spent the last several years building a model that simulates every game based on run/pass tendencies depending on the situation (down, distance, score, time, ect.) So it essentially simulates every play of every games a predetermined amount of times. Personally (of course I'm probably biased) I think its as good as or better than anything out there. It's still a work in progress though as I think I can get more accurate. That's where I am, I believe that any drafting tool is only as good as the projections you feed into it. I'd be interested in hearing more about this tool you have though.

I'd be equally curious to see yours. It sounds like you created something to replace Strength of Schedule.

Projections have to be an input, because everyone believes differently. I really like Dodd's projections, he's conservative enough that you don't overvalue. Nothing worse than overvaluing... I'd rather under value and get more, than over value and get less. I don't do my own projections, I'm paying FBG so if they are wrong I have someone to blame.

My file has a tab that is designed like a draft board. As I enter the player's name a hidden fields references the owner who made pick. From that formula the entire file replicates the data. It took a long time to build this, but it works very well. Since I'm a designer, it looks beautiful. It has never crashed. I import projections modified from DD. I import Weekly Points from DD. I use the Weekly Points as a reference for picking bye week replacements. But I'd like to create a formula that will identify those players automatically. Vlookup or an Array formula to find best projected score for available player for the week my current player(s) are on bye week. That should be pretty simple to create. However all that is just a reference guide. The draft formula is something different.

Getting all the data crunched into one flowing and changing number is very difficult. What I want it to do in the end is provide a rank that is more precise than VBD, and gives me an Expected Value vs Projected Value across positions and not just for one position. Isn't that the biggest problem with VBD that it really isn't a great rank for all positions. Half the time it tells you to take a player you know you shouldn't and the ADP factor also gets in the way forcing you to second guess. And you can only use it for a about 8-10 rounds and then its positional drafting anyways. If I can add a 'need' factor in as well that would account for what I need, and what my competition has taken. Mix that in with a differential for ADP, so that i don't grab them too early but maybe no less than 1 round before, that guarantees I don't miss the boat. That could be a huge addition... liking that.

I'm intrigued by your projected games formula... something like that could really enhance how we look at SOS. My SOS is based on positions. So its ARI allowed XXX to QBs, which ranks 16. Simple. However, it can't account for changes to the defense for the current year. That would require a projected weight formula. And would still be a lot of guessing. PIT added 3 players to their starting defense, power rank of 6th overall. So they would get a higher weight. Does that make sense? I have never made SOS a huge part of my draft rank... harder to project results. But if you have created something that ranks close to what actually happens... that would be huge.

As nerdy as all this is... I think its fun to try to be a know it all. My leagues hate me cuz I ALWAYS place in the money. I have won 3 out of 4 years over 3 leagues, and never less than 3rd. I know it can't predict the future... but math is everywhere... and maybe, just maybe I can create something that will make my drafts even easier. Not to take anything away from the DD. AWESOME program, I literally use it every day to track all my projections and notes on players. But I just want it to do more. I'm not wasting anymore time than I would if I was just watching tv... might create lighting in a bottle.

I need a cool acroynm.... lol

Who the heck are we kidding here? When things are that close, you will do just as well to flip a freaking coin then to come up with some special "tiebreaker" type analysis.

This is the kind of thing that is really just a waste of time....but then again so is fantasy football in general, so cheers!