Using a tree partition model on WR data since 2009 (1 Viewer)

[Spin]

Ok, so here is my task that I started on, and would appreciate input, criticism, more ideas, things I may have missed.

I took the top 40 wrs/te from 2009 through 2012 and using a k-fold cross-validation (k = 5) attempted to isolate the variables that lead to high fantasy points. This is for 1pt ppr, TEs considered as well.

Variables considered:

TEAM, Position, REC, TAR, YDS, Average Yards per catch, TDs, Longest Reception, Number of catches 20 yards +, Yards per Game, Fumbles, Yards after Catch, 1st Downs, Year, Team total YDS, Team Total Yds/game, Team Pass Yards, Team Pass Yards / G, Team rush yards, Team Rush yards / g, Team Pts Team Pts / g, Team Pass att, Team Comp, Team Comp PCT, Team Pass Yards / A, Team Long Pass play, Team Pass TD, Team Pass INT, Team Sacks against, Team Sack Yard Lost, Team Passer Rating, Team Rush ATT, Team YDS / A, Team Long Rush, Team Rush TD, Team Fumble, Team Fumble Lost

So, as expected Yards have by far the biggest impact on fantasy points.

837 was the first cutoff:

<837 yards: Count - 346, Mean score - 127.28, std dev - 30.32

>837 yards: Count - 134, Mean score - 232.35, std dev - 43.29

Continuing on, looking for just the creme of the crop

837<yards<1092: count - 76, mean score - 203.79, std dev - 23.17

>1092 yards: count - 58, mean score - 269.78, std dev - 33.98

Ok, so we know yards = points. Let's look at the biggest indicators for the top of the top.

Yards > 1508 - count - 7, mean score - 324.78, std dev - 26.5

For the guys with yards over 1508, no other variable stood out, so let's look at those between 1092 and 1508

As far as the individual stats go, Yards -> Tds -> Rec

Yards < 1508, Tds >= 10, Rec >= 90 yielded: count - 7, mean 302.65, std dev 12.71

What I found the interesting was the next breakdown. Yards per catch was the next significant variable, with 14.7 being the breaking point. But wrs/tes that averaged less than 14.7 outscored those that averaged more than 14.7:

Yards < 1508, TDS < 10, Avg Yards per catch < 14.7: count - 6, mean 280.56, std dev 6.6

Yards < 1508, tds < 10, Avg yards per catch > 14.7: count - 6, mean 263.84, std dev 4.98

My assumption is that having a low avg yards per catch, yields higher receptions. But there is no statistical breaking point on a particular reception number, but just looking at this group, the top 6 all had > 100 receptions with less than 14.7 yards per catch. (Welker twice (2010, 2012), Wayne twice(2010, 2012), Steve Smith NYG(2009), and Roddy White(2011))

So eliminating individual stats, lets just look at team data.

The following splits yielded the highest margin:

Team Pass Yards/g > 222.9, Team Pass Tds >= 25, Team Pass yards per attempt >= 7.9, 402>=Team Comp < 450. - Gave a mean of 229.

If the team had greater then 450 completions, but still the other breaking points (pass yards/g > 222.9, pass tds>= 25, pass yards per attempt >= 7.9) - the mean dropped to 172.64.

My only thought as to why having more than 450 completions would actually lead to lower fantasy points is maybe the teams with that many completions simply spread the ball out too much to yield higher points to a single player? Not sure.

So what are the thoughts? Is this a waste of my time? Is this just saying what we already know? Is anyone interested in another point range other than the top? Are the any obvious variables I could be missing? I originally thought that Targets would be one of the biggest indicators, but it continually failed to showed up. Forcing it to look JUST at targets, it gives a split at 106 targets. <106 - mean 126, std dev 33.64. >=106 targets - mean 227, std dev 66.43.

 

[DropKick]

Ok, so here is my task that I started on, and would appreciate input, criticism, more ideas, things I may have missed.

I took the top 40 wrs/te from 2009 through 2012 and using a k-fold cross-validation (k = 5) attempted to isolate the variables that lead to high fantasy points. This is for 1pt ppr, TEs considered as well.

Variables considered:

TEAM, Position, REC, TAR, YDS, Average Yards per catch, TDs, Longest Reception, Number of catches 20 yards +, Yards per Game, Fumbles, Yards after Catch, 1st Downs, Year, Team total YDS, Team Total Yds/game, Team Pass Yards, Team Pass Yards / G, Team rush yards, Team Rush yards / g, Team Pts Team Pts / g, Team Pass att, Team Comp, Team Comp PCT, Team Pass Yards / A, Team Long Pass play, Team Pass TD, Team Pass INT, Team Sacks against, Team Sack Yard Lost, Team Passer Rating, Team Rush ATT, Team YDS / A, Team Long Rush, Team Rush TD, Team Fumble, Team Fumble Lost

So, as expected Yards have by far the biggest impact on fantasy points.

837 was the first cutoff:

<837 yards: Count - 346, Mean score - 127.28, std dev - 30.32

>837 yards: Count - 134, Mean score - 232.35, std dev - 43.29

Continuing on, looking for just the creme of the crop

837<yards<1092: count - 76, mean score - 203.79, std dev - 23.17

>1092 yards: count - 58, mean score - 269.78, std dev - 33.98

Ok, so we know yards = points. Let's look at the biggest indicators for the top of the top.

Yards > 1508 - count - 7, mean score - 324.78, std dev - 26.5

For the guys with yards over 1508, no other variable stood out, so let's look at those between 1092 and 1508

As far as the individual stats go, Yards -> Tds -> Rec

Yards < 1508, Tds >= 10, Rec >= 90 yielded: count - 7, mean 302.65, std dev 12.71

What I found the interesting was the next breakdown. Yards per catch was the next significant variable, with 14.7 being the breaking point. But wrs/tes that averaged less than 14.7 outscored those that averaged more than 14.7:

Yards < 1508, TDS < 10, Avg Yards per catch < 14.7: count - 6, mean 280.56, std dev 6.6

Yards < 1508, tds < 10, Avg yards per catch > 14.7: count - 6, mean 263.84, std dev 4.98

My assumption is that having a low avg yards per catch, yields higher receptions. But there is no statistical breaking point on a particular reception number, but just looking at this group, the top 6 all had > 100 receptions with less than 14.7 yards per catch. (Welker twice (2010, 2012), Wayne twice(2010, 2012), Steve Smith NYG(2009), and Roddy White(2011))

So eliminating individual stats, lets just look at team data.

The following splits yielded the highest margin:

Team Pass Yards/g > 222.9, Team Pass Tds >= 25, Team Pass yards per attempt >= 7.9, 402>=Team Comp < 450. - Gave a mean of 229.

If the team had greater then 450 completions, but still the other breaking points (pass yards/g > 222.9, pass tds>= 25, pass yards per attempt >= 7.9) - the mean dropped to 172.64.

My only thought as to why having more than 450 completions would actually lead to lower fantasy points is maybe the teams with that many completions simply spread the ball out too much to yield higher points to a single player? Not sure.

So what are the thoughts? Is this a waste of my time? Is this just saying what we already know? Is anyone interested in another point range other than the top? Are the any obvious variables I could be missing? I originally thought that Targets would be one of the biggest indicators, but it continually failed to showed up. Forcing it to look JUST at targets, it gives a split at 106 targets. <106 - mean 126, std dev 33.64. >=106 targets - mean 227, std dev 66.43.

Seems to me like you're using the numbers to "back in" to the scoring system. So, if two players have comparable yardage, the one with the lower Yards/Reception might score more in PPR. The opposite might be true if you award points based on yards/touch.How do you hope your analysis to impact the way you play the game?

 

[Spin]

Ok, so here is my task that I started on, and would appreciate input, criticism, more ideas, things I may have missed.

I took the top 40 wrs/te from 2009 through 2012 and using a k-fold cross-validation (k = 5) attempted to isolate the variables that lead to high fantasy points. This is for 1pt ppr, TEs considered as well.

Variables considered:

TEAM, Position, REC, TAR, YDS, Average Yards per catch, TDs, Longest Reception, Number of catches 20 yards +, Yards per Game, Fumbles, Yards after Catch, 1st Downs, Year, Team total YDS, Team Total Yds/game, Team Pass Yards, Team Pass Yards / G, Team rush yards, Team Rush yards / g, Team Pts Team Pts / g, Team Pass att, Team Comp, Team Comp PCT, Team Pass Yards / A, Team Long Pass play, Team Pass TD, Team Pass INT, Team Sacks against, Team Sack Yard Lost, Team Passer Rating, Team Rush ATT, Team YDS / A, Team Long Rush, Team Rush TD, Team Fumble, Team Fumble Lost

So, as expected Yards have by far the biggest impact on fantasy points.

837 was the first cutoff:

<837 yards: Count - 346, Mean score - 127.28, std dev - 30.32

>837 yards: Count - 134, Mean score - 232.35, std dev - 43.29

Continuing on, looking for just the creme of the crop

837<yards<1092: count - 76, mean score - 203.79, std dev - 23.17

>1092 yards: count - 58, mean score - 269.78, std dev - 33.98

Ok, so we know yards = points. Let's look at the biggest indicators for the top of the top.

Yards > 1508 - count - 7, mean score - 324.78, std dev - 26.5

For the guys with yards over 1508, no other variable stood out, so let's look at those between 1092 and 1508

As far as the individual stats go, Yards -> Tds -> Rec

Yards < 1508, Tds >= 10, Rec >= 90 yielded: count - 7, mean 302.65, std dev 12.71

What I found the interesting was the next breakdown. Yards per catch was the next significant variable, with 14.7 being the breaking point. But wrs/tes that averaged less than 14.7 outscored those that averaged more than 14.7:

Yards < 1508, TDS < 10, Avg Yards per catch < 14.7: count - 6, mean 280.56, std dev 6.6

Yards < 1508, tds < 10, Avg yards per catch > 14.7: count - 6, mean 263.84, std dev 4.98

My assumption is that having a low avg yards per catch, yields higher receptions. But there is no statistical breaking point on a particular reception number, but just looking at this group, the top 6 all had > 100 receptions with less than 14.7 yards per catch. (Welker twice (2010, 2012), Wayne twice(2010, 2012), Steve Smith NYG(2009), and Roddy White(2011))

So eliminating individual stats, lets just look at team data.

The following splits yielded the highest margin:

Team Pass Yards/g > 222.9, Team Pass Tds >= 25, Team Pass yards per attempt >= 7.9, 402>=Team Comp < 450. - Gave a mean of 229.

If the team had greater then 450 completions, but still the other breaking points (pass yards/g > 222.9, pass tds>= 25, pass yards per attempt >= 7.9) - the mean dropped to 172.64.

My only thought as to why having more than 450 completions would actually lead to lower fantasy points is maybe the teams with that many completions simply spread the ball out too much to yield higher points to a single player? Not sure.

So what are the thoughts? Is this a waste of my time? Is this just saying what we already know? Is anyone interested in another point range other than the top? Are the any obvious variables I could be missing? I originally thought that Targets would be one of the biggest indicators, but it continually failed to showed up. Forcing it to look JUST at targets, it gives a split at 106 targets. <106 - mean 126, std dev 33.64. >=106 targets - mean 227, std dev 66.43.

Seems to me like you're using the numbers to "back in" to the scoring system. So, if two players have comparable yardage, the one with the lower Yards/Reception might score more in PPR. The opposite might be true if you award points based on yards/touch.How do you hope your analysis to impact the way you play the game?

Hmm, good question. It started off as hoping to find a way to identify factors that I may be over looking to find hidden gems when doing projections etc. I had completely underestimated how big of an impact yardage really has. I guess it just shows how unpredictable TDs really are. Yards are king, the rest follows. And I get your point, two people with similar yardage, tds, but one averaged less yards per catch, will obviously have to have more catches to get to that yardage, and thus require more receptions, and in the meanwhile create more points. I suppose it just reaffirms that possession receivers hold more value then I had originally thought. Non-flashy players, like Amendola, are commonly overlooked, while being gold in ppr and available much much later.

Another interesting note, I had always assumed that a strong running game created more receiver stats. My thinking was a stronger rushing game would create more in the box, leaving more single coverage, and thus more stats. But what was what I found interesting was that if you isolate just the rushing stats, teams that average less than 4.4 yards per rush attempt and less than 371 rushing attempts have the highest mean (218). Which makes sense, if you suck at running, you'll obviously pass more. Or, you rush less, because you're that good at passing.

 

[Rhythmdoctor]

I think if you're trying to find hidden gems, you first need to identify what player attributes are consistent among elite players (if there are any). From there you'll want to tier your findings based on opportunity, offensive philosophy, coaching staff and a variety of other factors.The analysis you conducted sort of just narrows down the data to detect the obvious. You obviously have the passion to analyze data so you're more than halfway there.