TheMathNinja
Footballguy
I've dealt too long with nonsense fantasy scoring systems, and so after much deliberation, I'm finally proposing something new in fantasy scoring to you all, the sharks. I'm hoping it sparks some awesome discussion, and maybe even a new alternate format for some leagues. We've all been on the wrong side of a fantasy game when the third-down RB on our opponent's team scores more points for a 1-yd TD run than our every-down guy did for marching the ball 60 yards down the field (it's tragic!). And I know none of you actually thinks a QB who goes 35-for-60 for 300 yards had as good of a game as the one who goes 35-for-40 for 300 yards. So I'm here to propose something new: a fantasy scoring system that is simple, gives familiar point values, and yet actually measures the real, on-the-field contribution a player makes to his team.
The good news is, contribution isn't impossible to measure. The guys at AdvancedNFLStats.com have done well in creating "Estimated Points Added" (EPA), a measure of how many points a player improves or un-improves his team's expected points on every play, based on data from thousands of games on how down-and-yardage situations translate into real in-game points. For example, 1st and 10 at my own 40 yard-line is worth an estimated 1.44 points, because that's the average net gain for teams in that position over many years. In this situation, an 8-yard run is worth +0.53 points, an incomplete pass is worth -0.53 points, and a fumble at the line of scrimmage is worth -4.10 points. For more details on the concept, see http://bit.ly/d0ybjq, and you can browse the raw data here: http://bit.ly/19KNGWk. Though the EPA value of 8-yard runs, incompletions, and fumbles are somewhat situation-dependent, they're much less so than you might think, and so very meaningful averages can be ascertained. And I've done just that (via dizzying quantities of Multiple Linear Regression models).
The problem is, most fantasy systems don't penalize the incompletion like they reward the 8-yard run, or penalize the fumble as if it undoes eight 8-yard runs, on average. What results is the glorification of highly inefficient players who post a lot of "production", but do so while often hurting their team (i.e. who rack up inordinate amounts of drops, sacks, incompletions, and 2-yard runs that don't get 1st downs en route to "production"). For example, Trent Richardson only improved his team's expected points on 39.8% of his touches in 2012 (notice 3.6 yards per carry), totaling -19.7 EPA over the 15 games he played. C.J. Spiller, on the other hand, improved his team's expected points on 51.9% of his touches, totaling +48.3 EPA through his team's first 15 games. Notice he outgained Richardson by 250 yards on 85 fewer carries and had 20 more receiving yards on 20 fewer targets. And yet, standard fantasy scoring had Richardson leading Spiller by 8.5 fantasy points after 15 weeks! This is because detrimental running plays are not penalized, and the standard 5-yard TD run (which takes your team from 5 expected points to achieving a full 7) is awarded with 6 points, though it's only worth about 1.6 (since you have to kickoff after, giving 0.4 points to the other team), while yards are awarded at roughly their EPA value.
So what system fixes these major issues? The answer, of course, is that there is no perfect system. Some systems can estimate EPA quite closely, but are neither intuitive, nor satisfactory at all (i.e. a fantasy system where Trent Richardson is worth negative points on the year). The current standard system, on the other hand, may "feel good", but doesn't give results true to the game of football. But I have devised a system that tries to find a happy medium in being simple, giving "familiar" point values per game (about 3 times EPA), and yet resembles actual Estimated Points Added. Here are the basics:
Rushing:
-0.5 points per rushing attempt
0.2 points per yard (1 point per 5 yards)
4 points per Rush TD
-6 points per fumble
Receiving:
-1 point per target
0.5 points per reception
0.2 points per yard
4 points per Receiving TD
-6 points per fumble
Passing:
-0.7 points per pass attempt
0.5 points per completion
0.7 points per 5 yards gained
-2 points per sack
4 points per Passing TD
-6 points per INT or Fumble
I realize fully that the passing numbers are a bit messy, but TD's (+4), turnovers (-6), and standard 5-yard gains (+0.5) are simple and consistent across positions. If you're skeptical at first, consider the results for the 2012 season, with the standard system scores in parenthesis:
QBs:
1. Tom Brady - 477 pts (329 standard)....EPA: 202.5
2. Drew Brees - 469 pts (337 standard)....EPA: 157.8
3. Peyton Manning - 464 pts (304 standard)...EPA: 171.5
4. Matt Ryan - 429 pts (291 standard)...EPA: 178.8
5. Robert Griffin III - 424 pts (304 standard)...EPA: 138.4
6. Aaron Rodgers - 416 (329 standard)...EPA: 162.0
Standard System Correlation to EPA: 72.6%
New System Correlation to EPA: 88.9%
RBs:
1. Adrian Peterson - 302 pts (297 standard)...EPA: 31.4
2. Doug Martin - 226 pts (248 standard)...EPA: 17.4
3. Marshawn Lynch - 220 pts (234 standard)...EPA: 19.6
4. C.J. Spiller - 217 pts (199 standard)...EPA: 40.5
5. Alfred Morris - 205 pts (230 standard)...EPA: 4.8
6. Frank Gore - 190 pts (190 standard)...EPA: 11.2
Standard System Correlation to EPA: 9.0%
New System Correlation to EPA: 24.0%
WRs:
1. Calvin Johnson - 254 pts (214 standard)...EPA: 107.3
2. Andre Johnson - 230 pts (176 standard)...EPA: 83.4
3. Dez Bryant - 229 pts (201 standard)...EPA: 59.1
4. Brandon Marshall - 219 pts (210 standard)...EPA: 51.4
5. Demaryius Thomas - 216 pts (191 standard)...EPA: 85.9
6. Roddy White - 206 pts (169 standard)...EPA: 71.2
Standard System Correlation to EPA: 49.8%
New System Correlation to EPA: 67.3%
The slight decrease in RB scores relative to WR's is largely due to their relative lack of efficiency (Yds/Att), and QB scores are higher because they aren't arbitrarily deducted points for the sake of balance; the goal for both QB's and WR's was a fantasy score worth about 3 times EPA. Come draft time, this system makes QB's, RB's, and WR's all essentially equal in value in standard roster systems, a huge bonus that finally ends the absurdity of RB's being taken with the first 10 picks of a fantasy draft. Using Value-Based Drafting with the 14th starter as a baseline, the best first round choices in a 14-team league in 2012 would have included 5 QB's, 5 RB's, and 4 WR's.
Using this system, now Aaron Rodgers' sacks, Drew Brees' picks, A.J. Green's drops, and Arian Foster's struggles to get beyond 3 yards (all of which hugely affect their contribution to their team's scoring chances) all have an appropriate impact on their fantasy value, as does Matt Ryan's stellar completion percentage, Andre Johnson's massive yard total, and C.J. Spiller's efficiency. I've also developed a scoring system based on the same concept for IDP's.
So what do you think? Is the added accuracy to a player's REAL football value worth some added complexity? Are you willing to properly penalize pass attempts, rush attempts, and targets in order to accurately value a player's points contributed? Are you fed up enough with the standard system's propensity to value guys who just touch the ball all day long rather than those really move it down field in order to make a switch to a system like this? I'd like your thoughts, because I'm running at least one league this year based on this system, and am interested in piloting a few others for those who are ready for something a little more accurate and informative.
The good news is, contribution isn't impossible to measure. The guys at AdvancedNFLStats.com have done well in creating "Estimated Points Added" (EPA), a measure of how many points a player improves or un-improves his team's expected points on every play, based on data from thousands of games on how down-and-yardage situations translate into real in-game points. For example, 1st and 10 at my own 40 yard-line is worth an estimated 1.44 points, because that's the average net gain for teams in that position over many years. In this situation, an 8-yard run is worth +0.53 points, an incomplete pass is worth -0.53 points, and a fumble at the line of scrimmage is worth -4.10 points. For more details on the concept, see http://bit.ly/d0ybjq, and you can browse the raw data here: http://bit.ly/19KNGWk. Though the EPA value of 8-yard runs, incompletions, and fumbles are somewhat situation-dependent, they're much less so than you might think, and so very meaningful averages can be ascertained. And I've done just that (via dizzying quantities of Multiple Linear Regression models).
The problem is, most fantasy systems don't penalize the incompletion like they reward the 8-yard run, or penalize the fumble as if it undoes eight 8-yard runs, on average. What results is the glorification of highly inefficient players who post a lot of "production", but do so while often hurting their team (i.e. who rack up inordinate amounts of drops, sacks, incompletions, and 2-yard runs that don't get 1st downs en route to "production"). For example, Trent Richardson only improved his team's expected points on 39.8% of his touches in 2012 (notice 3.6 yards per carry), totaling -19.7 EPA over the 15 games he played. C.J. Spiller, on the other hand, improved his team's expected points on 51.9% of his touches, totaling +48.3 EPA through his team's first 15 games. Notice he outgained Richardson by 250 yards on 85 fewer carries and had 20 more receiving yards on 20 fewer targets. And yet, standard fantasy scoring had Richardson leading Spiller by 8.5 fantasy points after 15 weeks! This is because detrimental running plays are not penalized, and the standard 5-yard TD run (which takes your team from 5 expected points to achieving a full 7) is awarded with 6 points, though it's only worth about 1.6 (since you have to kickoff after, giving 0.4 points to the other team), while yards are awarded at roughly their EPA value.
So what system fixes these major issues? The answer, of course, is that there is no perfect system. Some systems can estimate EPA quite closely, but are neither intuitive, nor satisfactory at all (i.e. a fantasy system where Trent Richardson is worth negative points on the year). The current standard system, on the other hand, may "feel good", but doesn't give results true to the game of football. But I have devised a system that tries to find a happy medium in being simple, giving "familiar" point values per game (about 3 times EPA), and yet resembles actual Estimated Points Added. Here are the basics:
Rushing:
-0.5 points per rushing attempt
0.2 points per yard (1 point per 5 yards)
4 points per Rush TD
-6 points per fumble
Receiving:
-1 point per target
0.5 points per reception
0.2 points per yard
4 points per Receiving TD
-6 points per fumble
Passing:
-0.7 points per pass attempt
0.5 points per completion
0.7 points per 5 yards gained
-2 points per sack
4 points per Passing TD
-6 points per INT or Fumble
I realize fully that the passing numbers are a bit messy, but TD's (+4), turnovers (-6), and standard 5-yard gains (+0.5) are simple and consistent across positions. If you're skeptical at first, consider the results for the 2012 season, with the standard system scores in parenthesis:
QBs:
1. Tom Brady - 477 pts (329 standard)....EPA: 202.5
2. Drew Brees - 469 pts (337 standard)....EPA: 157.8
3. Peyton Manning - 464 pts (304 standard)...EPA: 171.5
4. Matt Ryan - 429 pts (291 standard)...EPA: 178.8
5. Robert Griffin III - 424 pts (304 standard)...EPA: 138.4
6. Aaron Rodgers - 416 (329 standard)...EPA: 162.0
Standard System Correlation to EPA: 72.6%
New System Correlation to EPA: 88.9%
RBs:
1. Adrian Peterson - 302 pts (297 standard)...EPA: 31.4
2. Doug Martin - 226 pts (248 standard)...EPA: 17.4
3. Marshawn Lynch - 220 pts (234 standard)...EPA: 19.6
4. C.J. Spiller - 217 pts (199 standard)...EPA: 40.5
5. Alfred Morris - 205 pts (230 standard)...EPA: 4.8
6. Frank Gore - 190 pts (190 standard)...EPA: 11.2
Standard System Correlation to EPA: 9.0%
New System Correlation to EPA: 24.0%
WRs:
1. Calvin Johnson - 254 pts (214 standard)...EPA: 107.3
2. Andre Johnson - 230 pts (176 standard)...EPA: 83.4
3. Dez Bryant - 229 pts (201 standard)...EPA: 59.1
4. Brandon Marshall - 219 pts (210 standard)...EPA: 51.4
5. Demaryius Thomas - 216 pts (191 standard)...EPA: 85.9
6. Roddy White - 206 pts (169 standard)...EPA: 71.2
Standard System Correlation to EPA: 49.8%
New System Correlation to EPA: 67.3%
The slight decrease in RB scores relative to WR's is largely due to their relative lack of efficiency (Yds/Att), and QB scores are higher because they aren't arbitrarily deducted points for the sake of balance; the goal for both QB's and WR's was a fantasy score worth about 3 times EPA. Come draft time, this system makes QB's, RB's, and WR's all essentially equal in value in standard roster systems, a huge bonus that finally ends the absurdity of RB's being taken with the first 10 picks of a fantasy draft. Using Value-Based Drafting with the 14th starter as a baseline, the best first round choices in a 14-team league in 2012 would have included 5 QB's, 5 RB's, and 4 WR's.
Using this system, now Aaron Rodgers' sacks, Drew Brees' picks, A.J. Green's drops, and Arian Foster's struggles to get beyond 3 yards (all of which hugely affect their contribution to their team's scoring chances) all have an appropriate impact on their fantasy value, as does Matt Ryan's stellar completion percentage, Andre Johnson's massive yard total, and C.J. Spiller's efficiency. I've also developed a scoring system based on the same concept for IDP's.
So what do you think? Is the added accuracy to a player's REAL football value worth some added complexity? Are you willing to properly penalize pass attempts, rush attempts, and targets in order to accurately value a player's points contributed? Are you fed up enough with the standard system's propensity to value guys who just touch the ball all day long rather than those really move it down field in order to make a switch to a system like this? I'd like your thoughts, because I'm running at least one league this year based on this system, and am interested in piloting a few others for those who are ready for something a little more accurate and informative.
Last edited by a moderator: