Subscriber Contest is LIVE (2 Viewers)

[BassNBrew]

Not sure how many combinations there are, but I'm sure the number is astronomical.

According to my (*very* quick, so possibly erroneous) calculations, there are about 2.27 x 10^30 combinations. If your computer could test out a million combinations per second, it would take you a little more than 720296800 million centuries to look at all possible combinations.

Not picking any Raiders would certainly shorten that timeframe.

 

[massraider]

I could sit here for hours and play with this.

Is there anything on the main page, where we can see last years winner, and what the roster was?

Very cool.

I have several plans for that 25 grand. Most involve , and maybe something to watch .

QB-41

RB-80

WR-103

TE-20

K-2

D-4

 

[Angry Beavers]

thus far - still time to tweak:

Your team has been successfully entered.

QB - Tom Brady - 24

QB - Drew Brees - 14

QB - Philip Rivers - 10

RB - Steven Jackson - 54

RB - Frank Gore - 10

RB - Chris Brown - 7

RB - LenDale White - 7

RB - Ladell Betts - 3

WR - Chad Johnson - 46

WR - Derrick Mason - 26

WR - Keenan McCardell - 10

WR - Antonio Bryant - 9

WR - Brandon Stokley - 2

WR - Braylon Edwards - 2

WR - Vincent Jackson - 1

TE - Kellen Winslow Jr - 10

TE - Ben Troupe - 8

PK - Jay Feely - 2

PK - Jeff Reed - 2

TD - Arizona Cardinals - 2

TD - Tennessee Titans - 1

Total value: 250

 

[RustyFA]

What is the scoring? Does only one starter score at QB, 2 at RB etc. per week? Thanks for the answers.

 

[apalmer]

What is the scoring? Does only one starter score at QB, 2 at RB etc. per week? Thanks for the answers.

First step in any contest: Read the Rules.

Each week your top QB, top 2 RBs, top 3 WRs, top TE, top PK and top Defense will count towards your weekly total. You do not have to select who will play. You will receive the maximum score based on your roster.

 

[RustyFA]

What is the scoring? Does only one starter score at QB, 2 at RB etc. per week? Thanks for the answers.

First step in any contest: Read the Rules.

Each week your top QB, top 2 RBs, top 3 WRs, top TE, top PK and top Defense will count towards your weekly total. You do not have to select who will play. You will receive the maximum score based on your roster.

Thanks, found it last night. As I'm sure you realize it wasn't with the contest stuff on the subscriber contest page.

 

[Particle Swarm]

If your computer could test out a million combinations per second, it would take you a little more than 720296800 million centuries to look at all possible combinations.

This makes for an interesting search problem. To solve this problem (by which I mean a good solution, as optimal solutions are hard to come by in NP-complete and NP-hard problems), you need a search algorithm, an objective function, and projections. Naturally, the whole exercise is only as good as the projections.I'm starting with simulated annealing as the algorithm. It is simpler to deal with than some other metaheuristics (although I prefer Ant Colony Optimization).

For projections, I'll just use Dodds' most recent. Once the program is done, adjusting the projections is really straightforward, if tedious.

The objective function quantifies the "goodness" of any potential solution. For a total points format, you simply want to maximize points. For a survivor format, you want to minimize the chance of being eliminated, which is probably done by maximizing the points scored in the worst week. This means the projections need to be prorated across weeks, and I am thinking that strength of schedule should be taken into account.

For example, I have a player projected to score, say, 218 points over the season. I have a strength of schedule for his team and his position. I normalize the SoS numbers so that each week is a percentage of the total points, and I can compute a score for each player for each week by dividing 218 by 17 weeks, then multiplying that number by the SoS-derived percentages (which total 100%).

Now, I want to generate solutions (rosters) such that the positions meet the rules, and the cost does not exceed 250. Given that roster, I can compute the number of points for each week, assuming the top N scorers at each position, based on the scoring rules. I really want to maximize the minimum week, but I also want to score as many points as I can over most weeks. I also should consider that some bye weeks will have a lower average score than others and non-bye weeks, so I really should generate some sort of "competitiveness" value for each week, and measure the projected weekly scores against this. But I haven't yet done that.

I also want to account for depth. If you simply take the top scoring player(s) each week, then sometimes the backup will never register. For example, Peyton Mannig will likely be the projected keeper score at QB for all weeks other than his bye week. That means the 3rd QB would never be selected in this process, and therefore the program will probably pick someone like Grossman. To account for this, I also want to include the total weekly score of the entire roster, weighted lower than the scores of the projected starting roster. This will, I hope, give me a stronger chance to survive in the face of injuries and bad games for the studs.

A sample result, using projections from 8/10 (IIRC) is this:

ROSTER

player 44 Jake Delhomme QB Carolina Panthers 15 269

player 55 Brad Johnson QB Minnesota Vikings 10 216

player 58 Philip Rivers QB San Diego Chargers 10 240

player 130 Kevin Faulk RB New England Patriots 3 103

player 106 Reggie Bush RB New Orleans Saints 22 224

player 96 Kevin Jones RB Detriot Lions 38 239

player 144 Duce Staley RB Pittsburgh Steelers 2 85

player 137 Cedric Houston RB New York Jets 2 119

player 172 Hines Ward WR Pittsburgh Steelers 36 242

player 171 Donald Driver WR Green Bay Packers 36 246

player 192 Terry Glenn WR Dallas Cowboys 13 188

player 183 TJ Houshmanzadeh WR Cincinnati Bengals 23 217

player 240 Doug Gabriel WR Oakland Raiders 1 127

player 225 Travis Taylor WR Minnesota Vikings 2 121

player 223 Chad Jackson WR New England Patriots 2 107

player 303 Daniel Graham TE New England Patriots 1 56

player 264 Todd Heap TE Baltimore Ravens 22 186

player 314 Jason Elam PK Denver Broncos 3 124

player 311 Adam Vinatieri PK New England Patriots 4 138

player 16 Miami Dolphins DE Miami Dolphins 3 115

player 21 New York Jets DE New York Jets 2 105

=== cost 250 score 3467

week 1 score: 113.542218006487 209.935146098421

week 2 score: 125.759584560618 218.723049315972

week 3 score: 116.055384735941 175.828208642926

week 4 score: 122.631580612989 190.044605801122

week 5 score: 117.116793518391 197.057008849534

week 6 score: 118.298466239062 156.791264292928

week 7 score: 113.527399623348 202.396996955737

week 8 score: 113.066615360575 202.811870744064

week 9 score: 117.844960625356 187.956291730224

week 10 score: 122.956959539292 215.704284368622

week 11 score: 128.954883792368 229.079297598075

week 12 score: 114.593992812954 211.289220880133

week 13 score: 123.447349995104 216.661571519475

week 14 score: 127.271928807473 221.549957178136

week 15 score: 120.158917963307 216.82558038784

week 16 score: 118.001591786891 208.139462947631

(The player data includes the cost and the projected points scored for the season as the last two values. The weekly data is the projected score for the top players (1 QB, 2 RB, 3 WR, etc.) and the projected score for the entire roster.)Since simulated annealing uses randomness to arrive at a solution, the rosters may well have an eyebrow-raising component. Ignoring that for now, it looks to me like I should implement the competitiveness value I mentioned above, so each week could be judged in the context of bye weeks. It also looks like I need to use a gradient to emphasize later weeks more, given that the teams surviving to those weeks will be stronger, and thus my roster will need to be correspondingly stronger then to survive the cut. I am also thinking of watering down the effect of SoS, perhaps by making only 50% of a player's weekly score based on SoS and the other 50% being 1/16 of the total projected points for the season.

I would be happy to post the source code for this, but it contains Clayton's Ultimate SoS numbers and David's projections, both from the subscriber area, so I would need to get FBGs permission first. Meanwhile, it is just another exercise in entertainment, since I don't know how to account for luck in my objective function! It should result in a better roster than I would be able to select by any other means, but based on my performance in the contest 2 years ago, that isn't saying much.

 

[Bristol]

If your computer could test out a million combinations per second, it would take you a little more than 720296800 million centuries to look at all possible combinations.

This makes for an interesting search problem. To solve this problem (by which I mean a good solution, as optimal solutions are hard to come by in NP-complete and NP-hard problems), you need a search algorithm, an objective function, and projections. Naturally, the whole exercise is only as good as the projections.I'm starting with simulated annealing as the algorithm. It is simpler to deal with than some other metaheuristics (although I prefer Ant Colony Optimization).

For projections, I'll just use Dodds' most recent. Once the program is done, adjusting the projections is really straightforward, if tedious.

The objective function quantifies the "goodness" of any potential solution. For a total points format, you simply want to maximize points. For a survivor format, you want to minimize the chance of being eliminated, which is probably done by maximizing the points scored in the worst week. This means the projections need to be prorated across weeks, and I am thinking that strength of schedule should be taken into account.

For example, I have a player projected to score, say, 218 points over the season. I have a strength of schedule for his team and his position. I normalize the SoS numbers so that each week is a percentage of the total points, and I can compute a score for each player for each week by dividing 218 by 17 weeks, then multiplying that number by the SoS-derived percentages (which total 100%).

Now, I want to generate solutions (rosters) such that the positions meet the rules, and the cost does not exceed 250. Given that roster, I can compute the number of points for each week, assuming the top N scorers at each position, based on the scoring rules. I really want to maximize the minimum week, but I also want to score as many points as I can over most weeks. I also should consider that some bye weeks will have a lower average score than others and non-bye weeks, so I really should generate some sort of "competitiveness" value for each week, and measure the projected weekly scores against this. But I haven't yet done that.

I also want to account for depth. If you simply take the top scoring player(s) each week, then sometimes the backup will never register. For example, Peyton Mannig will likely be the projected keeper score at QB for all weeks other than his bye week. That means the 3rd QB would never be selected in this process, and therefore the program will probably pick someone like Grossman. To account for this, I also want to include the total weekly score of the entire roster, weighted lower than the scores of the projected starting roster. This will, I hope, give me a stronger chance to survive in the face of injuries and bad games for the studs.

A sample result, using projections from 8/10 (IIRC) is this:

ROSTER

player 44 Jake Delhomme QB Carolina Panthers 15 269

player 55 Brad Johnson QB Minnesota Vikings 10 216

player 58 Philip Rivers QB San Diego Chargers 10 240

player 130 Kevin Faulk RB New England Patriots 3 103

player 106 Reggie Bush RB New Orleans Saints 22 224

player 96 Kevin Jones RB Detriot Lions 38 239

player 144 Duce Staley RB Pittsburgh Steelers 2 85

player 137 Cedric Houston RB New York Jets 2 119

player 172 Hines Ward WR Pittsburgh Steelers 36 242

player 171 Donald Driver WR Green Bay Packers 36 246

player 192 Terry Glenn WR Dallas Cowboys 13 188

player 183 TJ Houshmanzadeh WR Cincinnati Bengals 23 217

player 240 Doug Gabriel WR Oakland Raiders 1 127

player 225 Travis Taylor WR Minnesota Vikings 2 121

player 223 Chad Jackson WR New England Patriots 2 107

player 303 Daniel Graham TE New England Patriots 1 56

player 264 Todd Heap TE Baltimore Ravens 22 186

player 314 Jason Elam PK Denver Broncos 3 124

player 311 Adam Vinatieri PK New England Patriots 4 138

player 16 Miami Dolphins DE Miami Dolphins 3 115

player 21 New York Jets DE New York Jets 2 105

=== cost 250 score 3467

week 1 score: 113.542218006487 209.935146098421

week 2 score: 125.759584560618 218.723049315972

week 3 score: 116.055384735941 175.828208642926

week 4 score: 122.631580612989 190.044605801122

week 5 score: 117.116793518391 197.057008849534

week 6 score: 118.298466239062 156.791264292928

week 7 score: 113.527399623348 202.396996955737

week 8 score: 113.066615360575 202.811870744064

week 9 score: 117.844960625356 187.956291730224

week 10 score: 122.956959539292 215.704284368622

week 11 score: 128.954883792368 229.079297598075

week 12 score: 114.593992812954 211.289220880133

week 13 score: 123.447349995104 216.661571519475

week 14 score: 127.271928807473 221.549957178136

week 15 score: 120.158917963307 216.82558038784

week 16 score: 118.001591786891 208.139462947631

(The player data includes the cost and the projected points scored for the season as the last two values. The weekly data is the projected score for the top players (1 QB, 2 RB, 3 WR, etc.) and the projected score for the entire roster.)Since simulated annealing uses randomness to arrive at a solution, the rosters may well have an eyebrow-raising component. Ignoring that for now, it looks to me like I should implement the competitiveness value I mentioned above, so each week could be judged in the context of bye weeks. It also looks like I need to use a gradient to emphasize later weeks more, given that the teams surviving to those weeks will be stronger, and thus my roster will need to be correspondingly stronger then to survive the cut. I am also thinking of watering down the effect of SoS, perhaps by making only 50% of a player's weekly score based on SoS and the other 50% being 1/16 of the total projected points for the season.

I would be happy to post the source code for this, but it contains Clayton's Ultimate SoS numbers and David's projections, both from the subscriber area, so I would need to get FBGs permission first. Meanwhile, it is just another exercise in entertainment, since I don't know how to account for luck in my objective function! It should result in a better roster than I would be able to select by any other means, but based on my performance in the contest 2 years ago, that isn't saying much.

Go ahead and run with that and see how far it takes you. My guess is not past week one.

 

[SeniorVBDStudent]

Fourth iteration...

QB 39

RB 98

WR 88

TE 16

PK 3

DEF 6

 

[Dope]

15th iteration

QB 43

RB 103

WR 84

TE 10

K 5

DEF 5

 

[T-BONE]

If your computer could test out a million combinations per second, it would take you a little more than 720296800 million centuries to look at all possible combinations.

This makes for an interesting search problem. To solve this problem (by which I mean a good solution, as optimal solutions are hard to come by in NP-complete and NP-hard problems), you need a search algorithm, an objective function, and projections. Naturally, the whole exercise is only as good as the projections.I'm starting with simulated annealing as the algorithm. It is simpler to deal with than some other metaheuristics (although I prefer Ant Colony Optimization).

For projections, I'll just use Dodds' most recent. Once the program is done, adjusting the projections is really straightforward, if tedious.

The objective function quantifies the "goodness" of any potential solution. For a total points format, you simply want to maximize points. For a survivor format, you want to minimize the chance of being eliminated, which is probably done by maximizing the points scored in the worst week. This means the projections need to be prorated across weeks, and I am thinking that strength of schedule should be taken into account.

For example, I have a player projected to score, say, 218 points over the season. I have a strength of schedule for his team and his position. I normalize the SoS numbers so that each week is a percentage of the total points, and I can compute a score for each player for each week by dividing 218 by 17 weeks, then multiplying that number by the SoS-derived percentages (which total 100%).

Now, I want to generate solutions (rosters) such that the positions meet the rules, and the cost does not exceed 250. Given that roster, I can compute the number of points for each week, assuming the top N scorers at each position, based on the scoring rules. I really want to maximize the minimum week, but I also want to score as many points as I can over most weeks. I also should consider that some bye weeks will have a lower average score than others and non-bye weeks, so I really should generate some sort of "competitiveness" value for each week, and measure the projected weekly scores against this. But I haven't yet done that.

I also want to account for depth. If you simply take the top scoring player(s) each week, then sometimes the backup will never register. For example, Peyton Mannig will likely be the projected keeper score at QB for all weeks other than his bye week. That means the 3rd QB would never be selected in this process, and therefore the program will probably pick someone like Grossman. To account for this, I also want to include the total weekly score of the entire roster, weighted lower than the scores of the projected starting roster. This will, I hope, give me a stronger chance to survive in the face of injuries and bad games for the studs.

A sample result, using projections from 8/10 (IIRC) is this:

ROSTER

player 44 Jake Delhomme QB Carolina Panthers 15 269

player 55 Brad Johnson QB Minnesota Vikings 10 216

player 58 Philip Rivers QB San Diego Chargers 10 240

player 130 Kevin Faulk RB New England Patriots 3 103

player 106 Reggie Bush RB New Orleans Saints 22 224

player 96 Kevin Jones RB Detriot Lions 38 239

player 144 Duce Staley RB Pittsburgh Steelers 2 85

player 137 Cedric Houston RB New York Jets 2 119

player 172 Hines Ward WR Pittsburgh Steelers 36 242

player 171 Donald Driver WR Green Bay Packers 36 246

player 192 Terry Glenn WR Dallas Cowboys 13 188

player 183 TJ Houshmanzadeh WR Cincinnati Bengals 23 217

player 240 Doug Gabriel WR Oakland Raiders 1 127

player 225 Travis Taylor WR Minnesota Vikings 2 121

player 223 Chad Jackson WR New England Patriots 2 107

player 303 Daniel Graham TE New England Patriots 1 56

player 264 Todd Heap TE Baltimore Ravens 22 186

player 314 Jason Elam PK Denver Broncos 3 124

player 311 Adam Vinatieri PK New England Patriots 4 138

player 16 Miami Dolphins DE Miami Dolphins 3 115

player 21 New York Jets DE New York Jets 2 105

=== cost 250 score 3467

week 1 score: 113.542218006487 209.935146098421

week 2 score: 125.759584560618 218.723049315972

week 3 score: 116.055384735941 175.828208642926

week 4 score: 122.631580612989 190.044605801122

week 5 score: 117.116793518391 197.057008849534

week 6 score: 118.298466239062 156.791264292928

week 7 score: 113.527399623348 202.396996955737

week 8 score: 113.066615360575 202.811870744064

week 9 score: 117.844960625356 187.956291730224

week 10 score: 122.956959539292 215.704284368622

week 11 score: 128.954883792368 229.079297598075

week 12 score: 114.593992812954 211.289220880133

week 13 score: 123.447349995104 216.661571519475

week 14 score: 127.271928807473 221.549957178136

week 15 score: 120.158917963307 216.82558038784

week 16 score: 118.001591786891 208.139462947631

(The player data includes the cost and the projected points scored for the season as the last two values. The weekly data is the projected score for the top players (1 QB, 2 RB, 3 WR, etc.) and the projected score for the entire roster.)Since simulated annealing uses randomness to arrive at a solution, the rosters may well have an eyebrow-raising component. Ignoring that for now, it looks to me like I should implement the competitiveness value I mentioned above, so each week could be judged in the context of bye weeks. It also looks like I need to use a gradient to emphasize later weeks more, given that the teams surviving to those weeks will be stronger, and thus my roster will need to be correspondingly stronger then to survive the cut. I am also thinking of watering down the effect of SoS, perhaps by making only 50% of a player's weekly score based on SoS and the other 50% being 1/16 of the total projected points for the season.

I would be happy to post the source code for this, but it contains Clayton's Ultimate SoS numbers and David's projections, both from the subscriber area, so I would need to get FBGs permission first. Meanwhile, it is just another exercise in entertainment, since I don't know how to account for luck in my objective function! It should result in a better roster than I would be able to select by any other means, but based on my performance in the contest 2 years ago, that isn't saying much.

That's great ! Sure we can never put luck in a function... but we can use the algorithm to help the decision making process. Besides the fact that it's very nice to use maths to help you in a hobbie like ff.

Would you mind sending me a PM with the source code? I'm a subscriber.

TIA

 

[pizzatyme]

After 6 modifications, here is how my roster $$$ are spent.

QB-37

RB-102

WR-97

TE-7

PK-2

Def-5

 

[CrossEyed]

Anyone else thinking Warner/Leinart and another QB might not be a bad way to go?

 

[-OZ-]

QB- 42

RB- 114

WR- 75

TE- 11

PK- 3

Def- 5

 

[Bankerguy]

QB- 41

RB- 101

WR- 83

TE- 15

K- 4

DT- 6

 

[Robbie Cooper]

QB- 45

RB- 99

WR- 82

TE- 10

PK- 3

Def- 11

 

[joffer]

QB - 6

RB - 231

WR - 7

TE - 2

PK - 2

DT - 2

 

[TheFanatic]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

 

[ravnzfan]

I've toyed w/this a few times...I'm sure I'll run it a couple more before Monday

Position--$$$---Dodd's projections

QB----46----807

RB----82----762

WR----97----844

TE----10-----124

K------4------242

D-----11------277

250..........3056

 

[red]

QB 44

RB 108

WR 81

TE 11

K 3

D 3

 

[TheFanatic]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

Actually, never mind. With a minimum of 3 QB's you have to choose from three of these guys:A Feeley - 2J Martin NO - 2C Nall BUF - 2K Clemens NYJ - 2And with 7 WR's you have to pick 7 of these guys Gabriel OAK - 1T Pinkston PHI - 1D Carter CAR - 1A Davis BUF - 1W Welker MIA - 1D Hall KC - 1B Johnson ARI - 1R Curry OAK - 1M Hass NO - 1D Boston TB - 1K Colbert CAR - 1C Rogers DET - 1V Jackson SD - 1B Berrian CHI - 1B Finneran ATL - 1D Henderson NO - 1M Robinson MIN - 1S Vines DET - 1P Warrick SEA - 1P Crayton DAL - 1C Roby TEN - 1D Watts DEN - 1I don't even see 7 WR2's out of that list. Helluva TE crew to choose 2 from:J Kleinsasser MIN - 1D Byrd STL - 1S Alexander DEN - 1K Mangum CAR - 1B Fletcher IND - 1D Martin GB - 1M Schobel PHI - 1M Lewis JAX - 1D Jolley NYJ - 1K Wilson KC - 1S Heiden CLE - 1A Becht TB - 1G Wrighster JAX - 1D Lee GB - 1R Kelly CIN - 1B Scaife TEN - 1I Mili SEA - 1K Brady JAX - 1D Graham NE - 1C Anderson OAK - 1D Wilcox BAL - 1M Pollard DET - 1E Johnson SF - 1L Pope ARI - 1C Baker NYJ - 1T Scheffler DEN - 1Someone with a more well rounded team is going to stomp you as soon as one of your big RB's is on a bye and another one has a down week. I don't see you scoring more than a few points a week outside of your RB's.....You better hope all of your RB's have cake opponents the weeks the other ones are on byes. But with as much as D's change from year to year, how will you actually be able to predict that....

 

[Pigskin Fanatic]

QB 27

RB 108

WR 94

TE 10

PK 6

TD 5

i like joffer's approach i think, but aren't ya spending too much for backups and not enough on starters?

 

[apalmer]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

Actually, never mind. With a minimum of 3 QB's you have to choose from three of these guys:A Feeley - 2J Martin NO - 2C Nall BUF - 2K Clemens NYJ - 2And with 7 WR's you have to pick 7 of these guys Gabriel OAK - 1T Pinkston PHI - 1D Carter CAR - 1A Davis BUF - 1W Welker MIA - 1D Hall KC - 1B Johnson ARI - 1R Curry OAK - 1M Hass NO - 1D Boston TB - 1K Colbert CAR - 1C Rogers DET - 1V Jackson SD - 1B Berrian CHI - 1B Finneran ATL - 1D Henderson NO - 1M Robinson MIN - 1S Vines DET - 1P Warrick SEA - 1P Crayton DAL - 1C Roby TEN - 1D Watts DEN - 1I don't even see 7 WR2's out of that list. Helluva TE crew to choose 2 from:J Kleinsasser MIN - 1D Byrd STL - 1S Alexander DEN - 1K Mangum CAR - 1B Fletcher IND - 1D Martin GB - 1M Schobel PHI - 1M Lewis JAX - 1D Jolley NYJ - 1K Wilson KC - 1S Heiden CLE - 1A Becht TB - 1G Wrighster JAX - 1D Lee GB - 1R Kelly CIN - 1B Scaife TEN - 1I Mili SEA - 1K Brady JAX - 1D Graham NE - 1C Anderson OAK - 1D Wilcox BAL - 1M Pollard DET - 1E Johnson SF - 1L Pope ARI - 1C Baker NYJ - 1T Scheffler DEN - 1Someone with a more well rounded team is going to stomp you as soon as one of your big RB's is on a bye and another one has a down week. I don't see you scoring more than a few points a week outside of your RB's.....You better hope all of your RB's have cake opponents the weeks the other ones are on byes. But with as much as D's change from year to year, how will you actually be able to predict that....

Won't even make it to the first bye week. Assume he's got LT, LJ and SA (217 of his 237). Give them each 150 total yards, 5 receptions and 2 TD's per week-- Pretty respectable numbers and about what you'd hope for. That's 32 points apiece, except that 32 of them are on your bench. You get 64 points. As I recall, the cutoff point most weeks was around 115. Where do you get 50 more points out of the rest of that team?

 

[LHUCKS]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

 

[TheFanatic]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

Actually, never mind. With a minimum of 3 QB's you have to choose from three of these guys:A Feeley - 2J Martin NO - 2C Nall BUF - 2K Clemens NYJ - 2And with 7 WR's you have to pick 7 of these guys Gabriel OAK - 1T Pinkston PHI - 1D Carter CAR - 1A Davis BUF - 1W Welker MIA - 1D Hall KC - 1B Johnson ARI - 1R Curry OAK - 1M Hass NO - 1D Boston TB - 1K Colbert CAR - 1C Rogers DET - 1V Jackson SD - 1B Berrian CHI - 1B Finneran ATL - 1D Henderson NO - 1M Robinson MIN - 1S Vines DET - 1P Warrick SEA - 1P Crayton DAL - 1C Roby TEN - 1D Watts DEN - 1I don't even see 7 WR2's out of that list. Helluva TE crew to choose 2 from:J Kleinsasser MIN - 1D Byrd STL - 1S Alexander DEN - 1K Mangum CAR - 1B Fletcher IND - 1D Martin GB - 1M Schobel PHI - 1M Lewis JAX - 1D Jolley NYJ - 1K Wilson KC - 1S Heiden CLE - 1A Becht TB - 1G Wrighster JAX - 1D Lee GB - 1R Kelly CIN - 1B Scaife TEN - 1I Mili SEA - 1K Brady JAX - 1D Graham NE - 1C Anderson OAK - 1D Wilcox BAL - 1M Pollard DET - 1E Johnson SF - 1L Pope ARI - 1C Baker NYJ - 1T Scheffler DEN - 1Someone with a more well rounded team is going to stomp you as soon as one of your big RB's is on a bye and another one has a down week. I don't see you scoring more than a few points a week outside of your RB's.....You better hope all of your RB's have cake opponents the weeks the other ones are on byes. But with as much as D's change from year to year, how will you actually be able to predict that....

Won't even make it to the first bye week. Assume he's got LT, LJ and SA (217 of his 237). Give them each 150 total yards, 5 receptions and 2 TD's per week-- Pretty respectable numbers and about what you'd hope for. That's 32 points apiece, except that 32 of them are on your bench. You get 64 points. As I recall, the cutoff point most weeks was around 115. Where do you get 50 more points out of the rest of that team?

Totally forgot that the guy will only get two of the seemingly 4-5 stud RB's on his squad.....

 

[apalmer]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

Actually, never mind. With a minimum of 3 QB's you have to choose from three of these guys:A Feeley - 2J Martin NO - 2C Nall BUF - 2K Clemens NYJ - 2And with 7 WR's you have to pick 7 of these guys Gabriel OAK - 1T Pinkston PHI - 1D Carter CAR - 1A Davis BUF - 1W Welker MIA - 1D Hall KC - 1B Johnson ARI - 1R Curry OAK - 1M Hass NO - 1D Boston TB - 1K Colbert CAR - 1C Rogers DET - 1V Jackson SD - 1B Berrian CHI - 1B Finneran ATL - 1D Henderson NO - 1M Robinson MIN - 1S Vines DET - 1P Warrick SEA - 1P Crayton DAL - 1C Roby TEN - 1D Watts DEN - 1I don't even see 7 WR2's out of that list. Helluva TE crew to choose 2 from:J Kleinsasser MIN - 1D Byrd STL - 1S Alexander DEN - 1K Mangum CAR - 1B Fletcher IND - 1D Martin GB - 1M Schobel PHI - 1M Lewis JAX - 1D Jolley NYJ - 1K Wilson KC - 1S Heiden CLE - 1A Becht TB - 1G Wrighster JAX - 1D Lee GB - 1R Kelly CIN - 1B Scaife TEN - 1I Mili SEA - 1K Brady JAX - 1D Graham NE - 1C Anderson OAK - 1D Wilcox BAL - 1M Pollard DET - 1E Johnson SF - 1L Pope ARI - 1C Baker NYJ - 1T Scheffler DEN - 1Someone with a more well rounded team is going to stomp you as soon as one of your big RB's is on a bye and another one has a down week. I don't see you scoring more than a few points a week outside of your RB's.....You better hope all of your RB's have cake opponents the weeks the other ones are on byes. But with as much as D's change from year to year, how will you actually be able to predict that....

Won't even make it to the first bye week. Assume he's got LT, LJ and SA (217 of his 237). Give them each 150 total yards, 5 receptions and 2 TD's per week-- Pretty respectable numbers and about what you'd hope for. That's 32 points apiece, except that 32 of them are on your bench. You get 64 points. As I recall, the cutoff point most weeks was around 115. Where do you get 50 more points out of the rest of that team?

Totally forgot that the guy will only get two of the seemingly 4-5 stud RB's on his squad.....

Looking back, it seems like my 115 point cutoff was too low. Most weeks it was closer to 135. Good luck joffer.

 

[ctriopelle]

If your computer could test out a million combinations per second, it would take you a little more than 720296800 million centuries to look at all possible combinations.

This makes for an interesting search problem. To solve this problem (by which I mean a good solution, as optimal solutions are hard to come by in NP-complete and NP-hard problems), you need a search algorithm, an objective function, and projections. Naturally, the whole exercise is only as good as the projections.I'm starting with simulated annealing as the algorithm. It is simpler to deal with than some other metaheuristics (although I prefer Ant Colony Optimization).

For projections, I'll just use Dodds' most recent. Once the program is done, adjusting the projections is really straightforward, if tedious.

The objective function quantifies the "goodness" of any potential solution. For a total points format, you simply want to maximize points. For a survivor format, you want to minimize the chance of being eliminated, which is probably done by maximizing the points scored in the worst week. This means the projections need to be prorated across weeks, and I am thinking that strength of schedule should be taken into account.

For example, I have a player projected to score, say, 218 points over the season. I have a strength of schedule for his team and his position. I normalize the SoS numbers so that each week is a percentage of the total points, and I can compute a score for each player for each week by dividing 218 by 17 weeks, then multiplying that number by the SoS-derived percentages (which total 100%).

Now, I want to generate solutions (rosters) such that the positions meet the rules, and the cost does not exceed 250. Given that roster, I can compute the number of points for each week, assuming the top N scorers at each position, based on the scoring rules. I really want to maximize the minimum week, but I also want to score as many points as I can over most weeks. I also should consider that some bye weeks will have a lower average score than others and non-bye weeks, so I really should generate some sort of "competitiveness" value for each week, and measure the projected weekly scores against this. But I haven't yet done that.

I also want to account for depth. If you simply take the top scoring player(s) each week, then sometimes the backup will never register. For example, Peyton Mannig will likely be the projected keeper score at QB for all weeks other than his bye week. That means the 3rd QB would never be selected in this process, and therefore the program will probably pick someone like Grossman. To account for this, I also want to include the total weekly score of the entire roster, weighted lower than the scores of the projected starting roster. This will, I hope, give me a stronger chance to survive in the face of injuries and bad games for the studs.

A sample result, using projections from 8/10 (IIRC) is this:

ROSTER

player 44 Jake Delhomme QB Carolina Panthers 15 269

player 55 Brad Johnson QB Minnesota Vikings 10 216

player 58 Philip Rivers QB San Diego Chargers 10 240

player 130 Kevin Faulk RB New England Patriots 3 103

player 106 Reggie Bush RB New Orleans Saints 22 224

player 96 Kevin Jones RB Detriot Lions 38 239

player 144 Duce Staley RB Pittsburgh Steelers 2 85

player 137 Cedric Houston RB New York Jets 2 119

player 172 Hines Ward WR Pittsburgh Steelers 36 242

player 171 Donald Driver WR Green Bay Packers 36 246

player 192 Terry Glenn WR Dallas Cowboys 13 188

player 183 TJ Houshmanzadeh WR Cincinnati Bengals 23 217

player 240 Doug Gabriel WR Oakland Raiders 1 127

player 225 Travis Taylor WR Minnesota Vikings 2 121

player 223 Chad Jackson WR New England Patriots 2 107

player 303 Daniel Graham TE New England Patriots 1 56

player 264 Todd Heap TE Baltimore Ravens 22 186

player 314 Jason Elam PK Denver Broncos 3 124

player 311 Adam Vinatieri PK New England Patriots 4 138

player 16 Miami Dolphins DE Miami Dolphins 3 115

player 21 New York Jets DE New York Jets 2 105

=== cost 250 score 3467

week 1 score: 113.542218006487 209.935146098421

week 2 score: 125.759584560618 218.723049315972

week 3 score: 116.055384735941 175.828208642926

week 4 score: 122.631580612989 190.044605801122

week 5 score: 117.116793518391 197.057008849534

week 6 score: 118.298466239062 156.791264292928

week 7 score: 113.527399623348 202.396996955737

week 8 score: 113.066615360575 202.811870744064

week 9 score: 117.844960625356 187.956291730224

week 10 score: 122.956959539292 215.704284368622

week 11 score: 128.954883792368 229.079297598075

week 12 score: 114.593992812954 211.289220880133

week 13 score: 123.447349995104 216.661571519475

week 14 score: 127.271928807473 221.549957178136

week 15 score: 120.158917963307 216.82558038784

week 16 score: 118.001591786891 208.139462947631

(The player data includes the cost and the projected points scored for the season as the last two values. The weekly data is the projected score for the top players (1 QB, 2 RB, 3 WR, etc.) and the projected score for the entire roster.)Since simulated annealing uses randomness to arrive at a solution, the rosters may well have an eyebrow-raising component. Ignoring that for now, it looks to me like I should implement the competitiveness value I mentioned above, so each week could be judged in the context of bye weeks. It also looks like I need to use a gradient to emphasize later weeks more, given that the teams surviving to those weeks will be stronger, and thus my roster will need to be correspondingly stronger then to survive the cut. I am also thinking of watering down the effect of SoS, perhaps by making only 50% of a player's weekly score based on SoS and the other 50% being 1/16 of the total projected points for the season.

I would be happy to post the source code for this, but it contains Clayton's Ultimate SoS numbers and David's projections, both from the subscriber area, so I would need to get FBGs permission first. Meanwhile, it is just another exercise in entertainment, since I don't know how to account for luck in my objective function! It should result in a better roster than I would be able to select by any other means, but based on my performance in the contest 2 years ago, that isn't saying much.

As the "unofficial scorekeeper" of the last time we did this type of contest (I usually has the results posted by my lunch hour on Tuesdays, several hours ahead of FBG, and I also cought a few scoring errors in their "official" results) I think this roster stands about a 50% chance of being eliminated in week 7 or 8. My results files have all been wiped from my home machine, but they are on my office PC. I intend on checking tomorrow how these results would have stood up in prior years. Unfortunately the "internet nazis" at work have shut down my ability to view the forums on the work pc , so this year I won't be able to post final scores from there (for now anyway, I may try to invoke "executive privilige" ).If anyone who has access to FFLM report generator can post the stat file to the message board showing the player name and his points per the FBG contest rules, I should still be able to post updates during the MNF game as to where the cut is and may project out. I already have the Excel format to make this work, I just need to dust off how I constructed it and get used to being very tired on Tuusday mornings!

-CTR

 

[Fiddles]

anyone know where the rules are posted for this

the link in post 1 says official rules can be found here but theres no clickable link

 

[apalmer]

anyone know where the rules are posted for this

the link in post 1 says official rules can be found here but theres no clickable link

The word "Here" is clickable.

 

[Sinrman]

Let's see, I have:

QB 47

RB 105

WR 69

TE 17

PK 3

DEF 9

 

[Sister Havana]

My team, $249:

QB - Jake Delhomme - 15

QB - Trent Green - 15

QB - Brad Johnson - 10

RB - Shaun Alexander - 69

RB - Frank Gore - 10

RB - Chris Brown - 7

RB - Samkon Gado - 7

RB - LenDale White - 7

WR - Steve Smith - 49

WR - Joe Horn - 16

WR - Joe Jurevicius - 8

WR - Jerry Porter - 7

WR - Ernest Wilford - 5

WR - Bobby Engram - 4

WR - Antwaan Randle El - 3

TE - Ben Troupe - 8

TE - Bubba Franks - 2

PK - David Akers - 2

PK - Robbie Gould - 1

TD - Kansas City Chiefs - 2

TD - Cincinnati Bengals - 2

 

[Dolfan]

Updated my roster for probably the last time:

QB: $36

RB: $104

WR: $89

TE: $16

PK: $2

TD: $3

Total: $250

 

[CommuterMan]

I keep changing mine and probably will do until the last minute... but for now this is how I've spent my cap $

QB - $58

RB - $105

WR - $67

TE - $12

PK - $3

Def - $5

How many of you are paying attention to bye-weeks in order to make sure you have a full roster each week so as not be be eliminated in the survivor format? To me this has been the most difficult aspect - making sure I have "quality" players out on the field each week...

 

[Gr00vus]

My squad expenditures:

QB - 44

RB - 120

WR - 70

TE - 8

PK - 4

TD - 4

 

[wharf rat]

I keep changing mine and probably will do until the last minute... but for now this is how I've spent my cap $QB - $58RB - $105WR - $67TE - $12PK - $3Def - $5How many of you are paying attention to bye-weeks in order to make sure you have a full roster each week so as not be be eliminated in the survivor format? To me this has been the most difficult aspect - making sure I have "quality" players out on the field each week...

Great minds think alike...My team is:QB - $54RB - $105WR - $66TE - $18PK - $2Def - $5

 

[Q-Bert]

Position--$$$---Dodd's projections

QB----38----530

RB----87----883

WR----99----1257

TE----17-----254

K------5------226

D-----4------189

250..........3340

Draft Dominator low score of 102 in week 6.

I think that the scoring rules have changed since the last time this was done. If I remember correctly TEs used to get 2 ppr. So I am not sure if the cutoff points from previous seasons is relevent.

 

[Rounders]

QB - 40

RB - 108

WR - 71

TE - 19

K - 6

DT - 5

TOTAL - 249

According to the DD - Total points of 1902.7, with a single week low of 102.4

So FBG better be right

 

[jon_mx]

qb 41

rb 119

wr 68

te 10

k 4

def 8

------------

may have to upgrade WR a bit. did not realize starting 3.

 

[lborges]

QB 48

RB 107

WR 70

TE 10

K 6

Def 9

 

[reg]

QB - 42

RB - 89

WR - 90

TE - 19

PK - 4

DEF - 6

----------

250

DD says 1937.2 total points, low week of 107.2

 

[texasyankee]

QB - 45

RB - 96

WR - 92

TE - 10

PK - 3

Def - 4

DD says 2004.3

I had it up to 2015.9 but got scared by injury possibilities

 

[Disco Stu]

QB - 43

RB - 82

WR - 92

TE - 21

PK - 4

TD - 7

----------

249

DD says 2092.2 total points, low week of 107.4



EDIT: gonna need some more work... too many W6 and W7 byes.

 

[Riptide]

QB - 46

RB - 90

WR - 87

TE - 16

PK - 5

TD - 6

----------

250

DD says 2092.8 total points, low week of 103.9 Week 3 Next lowest Week 114.9 Week 11

 

[Doug Drinen]

By the way, if you want a printout of your team, use this link:

http://subscribers.footballguys.com/contest/checkmyteam.php

 

[videoguy505]

By the way, if you want a printout of your team, use this link:

http://subscribers.footballguys.com/contest/checkmyteam.php

SCARY! I tried that and it read:

Total value: 500

Something's multiplying by two...

 

[Doug Drinen]

By the way, if you want a printout of your team, use this link:

http://subscribers.footballguys.com/contest/checkmyteam.php

SCARY! I tried that and it read:

Total value: 500

Something's multiplying by two...

Not good, but thanks for the info. Videoguy, could you PM me the email address under which you're signed up with FBG so I can look into this?

 

[nerangers]

I only started mine today...might do one more tweak before Tuesday.

QB - 42

RB - 74

WR - 115

TE - 10

K - 3

Def - 6

I think I went with a more balanced approach and did not score as well when using the DD.

Edit - ooops forgot to add the 1 PPR in the scoring - 2029.8 looks much better than 1609.2

 

[apalmer]

By the way, if you want a printout of your team, use this link:

http://subscribers.footballguys.com/contest/checkmyteam.php

SCARY! I tried that and it read:

Total value: 500

Something's multiplying by two...

Not good, but thanks for the info. Videoguy, could you PM me the email address under which you're signed up with FBG so I can look into this?

Doug, I got the same thing. Will PM in caase Videoguy didn't see the request.

 

[Bristol]

I think it's funny that people are using DD to pick their team for this contest. This is one area where using DD makes little or no sense. As with any contest, luck takes a bigger part. To use DD to try and predict what each player will score from week to week doesn't fly. Every week, any player can come out of nowhere to score a TD or have a big game. With a contest like this, you need to try to give yourself as many players that have the opportunity to score or have a big game and it's not always the better players. DD is not capable of predicting when those games will occur, and neither are we. But, you can limit your risk by spreading your money out on solid, but not spectacular players.

Just my

 

[BassNBrew]

QB - 6RB - 231WR - 7TE - 2PK - 2DT - 2

Wow, talk about scraping the bottom of the barrel at every position besides RB....I'm interested in seeing how this turns out.

Actually, never mind. With a minimum of 3 QB's you have to choose from three of these guys:A Feeley - 2J Martin NO - 2C Nall BUF - 2K Clemens NYJ - 2And with 7 WR's you have to pick 7 of these guys Gabriel OAK - 1T Pinkston PHI - 1D Carter CAR - 1A Davis BUF - 1W Welker MIA - 1D Hall KC - 1B Johnson ARI - 1R Curry OAK - 1M Hass NO - 1D Boston TB - 1K Colbert CAR - 1C Rogers DET - 1V Jackson SD - 1B Berrian CHI - 1B Finneran ATL - 1D Henderson NO - 1M Robinson MIN - 1S Vines DET - 1P Warrick SEA - 1P Crayton DAL - 1C Roby TEN - 1D Watts DEN - 1I don't even see 7 WR2's out of that list. Helluva TE crew to choose 2 from:J Kleinsasser MIN - 1D Byrd STL - 1S Alexander DEN - 1K Mangum CAR - 1B Fletcher IND - 1D Martin GB - 1M Schobel PHI - 1M Lewis JAX - 1D Jolley NYJ - 1K Wilson KC - 1S Heiden CLE - 1A Becht TB - 1G Wrighster JAX - 1D Lee GB - 1R Kelly CIN - 1B Scaife TEN - 1I Mili SEA - 1K Brady JAX - 1D Graham NE - 1C Anderson OAK - 1D Wilcox BAL - 1M Pollard DET - 1E Johnson SF - 1L Pope ARI - 1C Baker NYJ - 1T Scheffler DEN - 1Someone with a more well rounded team is going to stomp you as soon as one of your big RB's is on a bye and another one has a down week. I don't see you scoring more than a few points a week outside of your RB's.....You better hope all of your RB's have cake opponents the weeks the other ones are on byes. But with as much as D's change from year to year, how will you actually be able to predict that....

Won't even make it to the first bye week. Assume he's got LT, LJ and SA (217 of his 237). Give them each 150 total yards, 5 receptions and 2 TD's per week-- Pretty respectable numbers and about what you'd hope for. That's 32 points apiece, except that 32 of them are on your bench. You get 64 points. As I recall, the cutoff point most weeks was around 115. Where do you get 50 more points out of the rest of that team?

Totally forgot that the guy will only get two of the seemingly 4-5 stud RB's on his squad.....

Should we stop these guys yet or let them go since they're on a roll.