|
kiimo posted:The last two posts are making it clear why I haven't won a fantasy championship in over ten years. You should ask them where they placed in their leagues!
|
# ? Aug 17, 2015 21:09 |
|
|
# ? Jun 6, 2024 06:20 |
|
Forever_Peace posted:Alright I wrote up some quick excel formula for you to try that at least fixes the problem of over-dependence on the player at the top of each tier. Here's the raw data I'm working with: https://docs.google.com/spreadsheets/d/1Jz08tdIaz7WU9-9OfO8zpGPneSPdfGmNMPe8Z43f-rw/edit?usp=sharing One of the guys on Reddit uses something called a Gaussian mixture model. Either of you two know anything about what that could be? kiimo posted:The last two posts are making it clear why I haven't won a fantasy championship in over ten years. You haven't won a championship because FF is arbitrary and capricious.
|
# ? Aug 17, 2015 21:10 |
|
kiimo posted:The last two posts are making it clear why I haven't won a fantasy championship in over ten years. I've beaten at least one of those nerds heads up in the playoffs at least two years in a row. #kickersmatter
|
# ? Aug 17, 2015 21:15 |
|
Forever_Peace posted:That could work, but the tiers may come out a bit strange sometimes because you're comparing every player to the first player in a tier. Like, consider if the first player happens to be Martavis or Randle. You'd get a bigger than normal tier because the initial variance happened to be high. You might also get some sudden cutoffs, where two players might be indistinguishable from one another, but one barely makes it into the higher tier while the other becomes the start of the lower tier. Yeah, instead of having a tier system you could have an extra column called like "individual tier" or something where for each player you go down until they're significantly different. These results might not always be intuitive, like the RB5 might have a very high variance so his tier goes down to 9 or 10, but the RB6 is a low variance so his tier only goes down to 8. I don't think that's a bad result, and it actually might help people consider the variance more.
|
# ? Aug 17, 2015 21:15 |
|
Beer4TheBeerGod posted:Here's the raw data I'm working with: https://docs.google.com/spreadsheets/d/1Jz08tdIaz7WU9-9OfO8zpGPneSPdfGmNMPe8Z43f-rw/edit?usp=sharing Based on like 20 seconds on wikipedia, a gaussian mixture model sounds like a fancy way of saying a weighted average of normal (Gaussian) distributions. I haven't worked with one myself, but they probably use it to give more weight to the "better" experts when coming up with their averages?
|
# ? Aug 17, 2015 21:19 |
|
Beer4TheBeerGod posted:You haven't won a championship because FF is arbitrary and capricious. Man I'd love to agree with you but last draft I passed on Luck to select Foles and passed on DeMarco Murray to select Andre Ellington. I am trash.
|
# ? Aug 17, 2015 21:24 |
|
kiimo posted:The last two posts are making it clear why I haven't won a fantasy championship in over ten years. I have the hugest mad boner respect for statistics but the fanciest poo poo I ever did was a regression analysis and I probably couldn't do it again if you held a gun to my head. Seriously, you guys are doing work in improving Beersheets, and that's awesome.
|
# ? Aug 17, 2015 21:25 |
|
kiimo posted:Man I'd love to agree with you but last draft I passed on Luck to select Foles and passed on DeMarco Murray to select Andre Ellington. I am trash. On the flip side I've led my league in scoring like four out of the five years and placed second in four out of five years. Championships are snakepoison.
|
# ? Aug 17, 2015 21:32 |
|
Beer4TheBeerGod posted:One of the guys on Reddit uses something called a Gaussian mixture model. Either of you two know anything about what that could be? Oh, yeah that's a fine idea. GMMs are fitted parametric models that try to maximize the probability of the data given the model (in contrast, KDE is nonparametric). Each "guassian" is just a mean and a variance (ie a distribution) that might be behind the data, and the "mixture" is a weighed blending of these distributions. In this case, he's making the assumption that our projections may be drawn from a set of overlapping tiers to which players belong (and working backwards, that any given player will have a certain probability of being in any given tier). It's not what I would have chosen but it could probably do the job. FWIW I think it's possible to make a hybrid KDE-GMM, but I wouldn't even know where to start! Re: my simple moving pooled estimates, I'll adapt the formula to your format later today.
|
# ? Aug 17, 2015 21:42 |
|
Forever_Peace posted:Oh, yeah that's a fine idea. GMMs are fitted parametric models that try to maximize the probability of the data given the model (in contrast, KDE is nonparametric). Each "guassian" is just a mean and a variance (ie a distribution) that might be behind the data, and the "mixture" is a weighed blending of these distributions. In this case, he's making the assumption that our projections may be drawn from a set of overlapping tiers to which players belong (and working backwards, that any given player will have a certain probability of being in any given tier). It's not what I would have chosen but it could probably do the job. Here's what he's been doing: http://www.borischen.co/2015/08/2015-fantasy-football-draft-kit-standard.html
|
# ? Aug 17, 2015 21:48 |
|
Forever_Peace posted:Oh, yeah that's a fine idea. GMMs are fitted parametric models that try to maximize the probability of the data given the model (in contrast, KDE is nonparametric). Each "guassian" is just a mean and a variance (ie a distribution) that might be behind the data, and the "mixture" is a weighed blending of these distributions. In this case, he's making the assumption that our projections may be drawn from a set of overlapping tiers to which players belong (and working backwards, that any given player will have a certain probability of being in any given tier). It's not what I would have chosen but it could probably do the job. Yeah, I don't think a GMM will give super different results from your suggestion anyway.
|
# ? Aug 17, 2015 21:55 |
|
Beer4TheBeerGod posted:Here's what he's been doing: http://www.borischen.co/2015/08/2015-fantasy-football-draft-kit-standard.html Somebody posted this dudes work in the IRC last year. I remember looking at it and distinctly thinking "this is a bunch of smoke and mirrors to convince people that Forsett is overvalued so that he can get easy trades in his league" Forsett was a top 10 RB at the time and Taliaferro was out of the picture
|
# ? Aug 17, 2015 22:14 |
|
I won my league last year because I won the OBJ lottery and also having Antonio Brown who I picked up for $32. Sometimes it's luck and being in the right spot at the right time, also because I'm the best.
|
# ? Aug 17, 2015 22:23 |
|
TNG posted:I won my league last year because
|
# ? Aug 17, 2015 22:25 |
|
I have the option of keeping Eddie Lacy or CJ Anderson and I keep second guessing myself. Lacy definitely has the track record, but I'm worried about him staying 100% healthy all season. Anderson has only started half a season, but looked like a monster. Anderson is definitely helped by the league being .5 ppr, but what if last season was a fluke? I haven't really kept up with the Broncos either, is he in danger of ending up in some nightmare committee backfield?
|
# ? Aug 17, 2015 22:29 |
|
I think Anderson just got hurt and Lacy was a no brainer before that. Broncos still have Ball and Hillman around whereas the Packers have...starks? haha, no contest.
|
# ? Aug 17, 2015 22:30 |
|
GUISSEPPE PIZZAPIE posted:I think Anderson just got hurt and Lacy was a no brainer before that. Broncos still have Ball and Hillman around whereas the Packers have...starks? haha, no contest. Anderson is fine. I'd take Lacy over him for sure, but I also have a sneaking suspicion Anderson is going to have a huge year.
|
# ? Aug 17, 2015 22:32 |
|
Beer4TheBeerGod posted:Here's what he's been doing: http://www.borischen.co/2015/08/2015-fantasy-football-draft-kit-standard.html This is very cool. I will say that using rankings instead of projections is adding some pretty extreme heteroscedasticity (in other words, the standard deviation of any player depends on the ranking of that player because of "wall" effects), but I have no idea how that impacts Gaussian mixture models so I'm not even sure that's much of a problem. In any case, I'll get the quick-and-dirty moving pools done and then explore KDE a bit. Thanks for the data sample.
|
# ? Aug 17, 2015 22:36 |
|
The Broncos have a ton of RB depth. Anderson is far from a sure thing. He has looked good, and while most people think he will have a good year, he's still a fairly risky pick. There's also a shitload of revisionism about Montee Ball. Because he crashed and burned so badly last year, magically everybody everywhere claims they called it, despite his consensus 2nd round draft position. Anderson is a better candidate this year than Ball was last year, but this time last year most fans weren't convinced the Broncos had someone who could take over and perform if Ball went down.
|
# ? Aug 17, 2015 22:46 |
|
Traffic Reporter posted:I have the option of keeping Eddie Lacy or CJ Anderson and I keep second guessing myself. Lacy definitely has the track record, but I'm worried about him staying 100% healthy all season. Why? Honest question. He's missed one game his first two years. Lacy is a legit candidate for the first pick, is way more established and has virtually no competition, and he'll be playing with Rodgers for the next few years at least. No idea why you're second guessing keeping him, and especially if CJ Anderson is the other guy you're considering.
|
# ? Aug 17, 2015 22:58 |
|
Agreed, Lacy is a no-brainer in your situation.
|
# ? Aug 17, 2015 22:58 |
|
BRAKE FOR MOOSE posted:The Broncos have a ton of RB depth. Anderson is far from a sure thing. He has looked good, and while most people think he will have a good year, he's still a fairly risky pick. Just because he was projected by "experts" doesn't mean this board didn't overwhelmingly hate him. (Mostly they just thought he was overrated. I don't think they expected him to flame out so spectacularly, just not match his ADP.)
|
# ? Aug 17, 2015 23:17 |
|
RVProfootballer posted:No idea why you're second guessing keeping him, and especially if CJ Anderson is the other guy you're considering. Just overthinking things. Thanks guys
|
# ? Aug 17, 2015 23:19 |
|
MacheteZombie posted:Agreed, Lacy is a no-brainer in your situation. He'll be a no-brainer in all situations if his concussions keep up
|
# ? Aug 17, 2015 23:21 |
|
After removing keepers, this is my overall list at the top and it is horrifying. Who do you guys think should move up the list? I have additional picks after trading away players from last team's poo poo house team who I re-named Nick Foles and the Assholes. Now I have Sammy Watkins as my keeper (!) and I think I'm going to be hosed again. Jamaal Charles CJ Anderson Demaryius Thomas Rob Gronkowski Dez Bryant Jeremy Hill Odell Beckham, Jr Jordy Nelson Julio Jones Calvin Johnson AJ Green Alshon Jeffery T.Y. Hilton Mark Ingram Mike Evans Emmanuel Sanders Alfred Morris Lamar Miller Kelvin Benjamin DeAndre Hopkins Jimmy Graham Carlos Hyde Justin Forsett Russell Wilson Melvin Gordon Peyton Manning Jordan Matthews (14 teams, normal scoring, snake draft)
|
# ? Aug 17, 2015 23:23 |
|
kiimo posted:After removing keepers, this is my overall list at the top and it is horrifying. Who do you guys think should move up the list? I have additional picks after trading away players from last team's poo poo house team who I re-named Nick Foles and the Assholes. Now I have Sammy Watkins as my keeper (!) and I think I'm going to be hosed again. is this the list of guys who will be available? If so, who do you have on your roster?
|
# ? Aug 17, 2015 23:29 |
|
Yes and I have Sammy Watkins. That's it. You probably don't need to quote that post though.
|
# ? Aug 17, 2015 23:29 |
|
guys we think Andre Ellington can bear this load unencumbered - Bruce Arians, liar
|
# ? Aug 17, 2015 23:30 |
|
Chris Johnson just stole carries from Andre Ellington. That guy will never be the Jamaal Charles clone I imagined him to be.
|
# ? Aug 17, 2015 23:31 |
|
kiimo posted:Yes and I have Sammy Watkins. That's it. You probably don't need to quote that post though. I don't understand. That is a list of available players? Why are guys like Hill, OBJ, and Mike Evans not being kept?
|
# ? Aug 17, 2015 23:41 |
|
Traffic Reporter posted:I haven't really kept up with the Broncos either, is he in danger of ending up in some nightmare committee backfield? GUISSEPPE PIZZAPIE posted:I think Anderson just got hurt and Lacy was a no brainer before that. Broncos still have Ball and Hillman around whereas the Packers have...starks? haha, no contest. BRAKE FOR MOOSE posted:The Broncos have a ton of RB depth. Anderson is far from a sure thing. He has looked good, and while most people think he will have a good year, he's still a fairly risky pick. Y'all I solved your problem a page ago Benne posted:Gary Kubiak doesn't do timeshares hth
|
# ? Aug 17, 2015 23:45 |
|
Beer4TheBeerGod posted:Here's the raw data I'm working with: https://docs.google.com/spreadsheets/d/1Jz08tdIaz7WU9-9OfO8zpGPneSPdfGmNMPe8Z43f-rw/edit?usp=sharing Alrighty, here's the updated excel sheet. Columns I added are highlighted. You should be able to copy-paste any new data array into here and have it work, but you will need to sort the data (i.e. the unlighted bits) by "Average" (column S). I currently have it set to include players within 2/3 of a standard deviation of the mean projection for that tier before moving on. This is the point at which a random projection moves from being more likely to have that player inside the tier to being more likely to have that player outside the tier, either above it or below it. You can change the size of the tiers as you see fit by changing the two instances of "2/3" in the formula in cell A3 (don't forget to change both). Again, here's what I'm doing: - Calculate the pooled variance of the current tier (this weights the variances by the sample sizes to give more emphasis to the players you have more projections for), and turn it into a pooled standard deviation. - Calculate the mean of the current tier. - Add the current player to the existing tier if it is within the cutoff from the mean (ie 2/3 SDs) and if the top player of the tier has not moved outside of the same distance from the mean from the lowered average. It's a quick and dirty approximation but it should help fix the current emphasis on the exact values for the top player in each tier. Forever_Peace fucked around with this message at 00:00 on Aug 18, 2015 |
# ? Aug 17, 2015 23:57 |
|
I hope the Chris Johnson signing brings Ellington's ADP down a little so I can easily get him in every league.
|
# ? Aug 18, 2015 00:02 |
|
I survived my first draft, you guys
|
# ? Aug 18, 2015 00:03 |
|
jabro posted:You should ask them where they placed in their leagues! Hey now, that doesn't seem particularly fair for our league because I had inherited a bad team last year after the draft had already occurred, and in by-far the most competitive division! One year later I now have a passable squad of vets, added 3 top-12 rookies in my first rookie draft (plus three more nice fliers), picked up an extra rookie 1st for next year, and accumulated a third of the first-round FA picks the year the startup contracts expire. I'm holding my own. In my competitive redraft league I went to the championship (and lost, but still).
|
# ? Aug 18, 2015 00:10 |
|
coronaball posted:I don't understand. That is a list of available players? Why are guys like Hill, OBJ, and Mike Evans not being kept? We get one keeper. If a guy like OBJ is on a team with Lacy, he isn't going to be kept. We also have a 3-year contract rule. Charles has been kept 3 years so he goes back in the pool.
|
# ? Aug 18, 2015 00:13 |
|
Fight Club Sandwich posted:He'll be a no-brainer in all situations if his concussions keep up
|
# ? Aug 18, 2015 00:39 |
|
I draft 7th in my league and need some opinions on strategy. Here is the most likely scenario I see for picks 1-6 (it could go any order): 0.5 PPR 2 RB, 2WR, 1 Flex, TE 1. Bell 2. Manning/Rodgers (Owner likes QB's in the first and loves Manning for some reason) 3. Peterson 4. Brown 5. Charles 6. Lacy If any of those 6 drop to me at 7, I would grab them. Here are the guys I am targeting given the above scenario 1. Dez Bryant 2. Marshawn Lynch 3. CJ Anderson I am having a tough time deciding the best play here and would appreciate any thoughts/opinions on the safest bet. I also projected what would be available in Round 2 so I could judge whether an RB or WR would be better in the first. At pick 2.6, I think I should have one or more o the following available; Jeremy Hill Jordy Nelson AJ Green Justin Forsett Randall Cobb Rob Gronkowski I know that I am overthinking this but would appreciate some thoughts on the best options for the first two rounds. MrSargent fucked around with this message at 01:35 on Aug 18, 2015 |
# ? Aug 18, 2015 00:40 |
|
PPR or standard? Also, how many/kinds of Flex?
|
# ? Aug 18, 2015 01:19 |
|
|
# ? Jun 6, 2024 06:20 |
|
Thwomp posted:PPR or standard? Also, how many/kinds of Flex? God I'm terrible, sorry. 0.5 PPR 2 RB, 2WR, 1 Flex, TE
|
# ? Aug 18, 2015 01:35 |