|
Effort post incoming... This week didn't end so well... something from AP would have been a boost to my top lineup but still, multiple lineups in the top 1K out of 100,000 is fine by me. Finished just above even for the week. Still, a bit frustrating as I felt I had a lot of hits on value plays, but most of my lineups had some big misses as well. I feel like intuitively I should have done better; what I'm thinking is that I need a way to spread out my combinations of players more evenly. Creating lineups by hand, one at a time, I get multiple players in my head at once. I'm making lineups and I'm trying to increase my exposure to Benny Cunningham at the same time as I'm trying to get more Frank Gore, for example. Hit on one but it's canceled out. Even if I'm aware of these combinations and trying to ensure variety I think I end up with more "clumpiness" than is optimal. So, I've been thinking about developing a better strategy. At the same time I've been watching the developments of various DFS strategy sites over the past year and the proliferation of "lineup optimizer" tools. These all seem to be doing linear programming: given a set of linear constraints (for example, sum of salary must be less than or equal to X, sum of # of RBs must be exactly 2, etc). This type of problem can be solved easily and so the application of linear programming to DFS is cropping up as a web service available at many sites that offer fantasy-point projections. However, the true problem of setting lineups is not a linear optimization problem, it is a stochastic optimization problem. There is uncertainty in some of the elements of the problem: the salaries are known, but the fantasy points are not. Going off the mean or median expected fantasy points produces good lineups for cash games but offers no way to truly take advantage of the tails of the expected point distributions, which is where you need to hit to win a GPP. I have been doing some reading on stochastic linear optimization and I think I finally figured out a system for creating an optimal portfolio of GPP lineups. It is easy to implement and I am excited to try it out. The one piece missing, though, is projections in terms of probability distributions as opposed to medians or means (I need to generate Monte Carlo simulations of all fantasy points scores). I am firmly a believer in taking advantage of the work of others (not to say being a leech---standing on the shoulders of giants and all that), so I have little desire to make my own projection model when there are so many out there. Even though I have academic experience with mixed-effect regression I don't think my time would be well spent putting together my own model of fantasy point performances. So many people are doing it, that even if they aren't doing it right I think there's little edge there. The edge, I think, is in working out a better method for how to use projections; that is how to go from the projections to the lineups. If at all possible I'd like to get into this area. (and just to be clear, I am only thinking about implementing a stochastic optimization strategy for myself and collaborators, maybe posting lineups it generates, but I have no intention of making a web service in the foreseeable future). Basically, my approach will formalize the intuition behind a "spray and pray" GPP strategy. So, here are my options:
Why am I posting this? I'm wondering if anyone wants to collaborate. I'm not sure if anyone else is interested in going to this level, but if you already have a projection model that does reasonably well, we could work together to create probability distributions from it (if it doesn't generate them already). Alternatively I could use help to make a more realistic model of the expected points distributions using the numberfire SD as a predictory of variability (among others). I've got a lot going on in my life so I only have the weekends really to work on this. With the right help I could get this done quickly, I think. The stochastic optimization method I want to use is deceptively simple. The hard part is getting the projections in the form of distributions in the first place. If anyone is interested in helping, please PM me. Please don't PM me if you just want get onboard but don't plan on contributing. I promise I will come back and share the strategy once it is coded up and paired with a model. Also, I mentioned in the other thread that I have been working on tools to scrape lots of data sources, mostly for archival purposes (what I would need the data for is only just becoming clear to me). I don't want to make this code public but I have no problem sharing it with you all. PM me a GitHub username and I will give you read access to the repository. Currently I have the following modules. Each runs from the command line and saves to a CSV or HDF file. code:Anyways, those are my thoughts and plans after week 1. Looking forward to hear from some of you, even if it's just to mock me and my nerdiness (and my abysmal dollars per hour return of the time I've spent working on NFL DFS).
|
#
?
Sep 15, 2015 08:04
|
|
|
|
| # ? Apr 20, 2024 10:34 |
|
|
What's everyone looking at this week? I've only done my Last Grinder Standing lineup so far but: Foles Lynch Murray Landry Julio TWill Seferian-Jenkins Nugent Dolphins Wanted Eifert or Palmer but had to give somewhere.
|
|
#
?
Sep 16, 2015 18:44
|
|
|
Turned a $5 profit last week 
|
|
#
?
Sep 16, 2015 20:48
|
|
|
JoshTheStampede posted:What's everyone looking at this week? I've only done my Last Grinder Standing lineup so far but: What's Hyde at on FD? On DK he is 5100 and am taking him all day long.
|
|
#
?
Sep 16, 2015 21:10
|
|
|
jabro posted:What's Hyde at on FD? On DK he is 5100 and am taking him all day long. $7000
|
|
#
?
Sep 16, 2015 21:12
|
|
|
Yeah at 7k he's not a must take but I am wavering between him and Lynch.
|
|
#
?
Sep 16, 2015 23:19
|
|
|
Ryan Mallett is starting for the Texans this week. He looked decent last week in 6 minutes of work: 8/13, 98 yards and a TD. Granted it was at the end of the game and not necessarily garbage time but close enough that the Chiefs were playing a pretty simple defense. That said, the Texans play the Panthers next week whose D is not nearly as good as the Chiefs. At $5100 on DK, I'm considering him for some of my GPP lineups. Am I being a crazy homer?
|
|
#
?
Sep 17, 2015 00:16
|
|
|
Sub Par posted:Ryan Mallett is starting for the Texans this week. He looked decent last week in 6 minutes of work: 8/13, 98 yards and a TD. Granted it was at the end of the game and not necessarily garbage time but close enough that the Chiefs were playing a pretty simple defense. That said, the Texans play the Panthers next week whose D is not nearly as good as the Chiefs. I'm going to try Mallett in one of my leagues but I'm a little worried, the Panthers D is much better than people give them credit for. Carson Palmer is either 6k or under I think he's the far better play against a swiss cheese Chicago secondary. Alfred Blue also isn't running the ball well so I think Mallett struggles until Foster returns.
|
|
#
?
Sep 17, 2015 02:05
|
|
|
Good points. In the end I think I'll stack him with Hopkins in one lineup and roll the dice.
|
|
#
?
Sep 17, 2015 02:07
|
|
|
Sub Par posted:Good points. In the end I think I'll stack him with Hopkins in one lineup and roll the dice. You aren't alone, I did the same. It's a dollar so it can't hurt! Good luck!
|
|
#
?
Sep 17, 2015 02:14
|
|
|
On the way to implementing my Master PlanTM, I made some correlation plots similar to the ones I made last year. Correlations range from -1 to +1 and describe the connection between two variables. A positive correlation means that they tend to swing in the same direction (for example, temperature is positively correlated with air conditioner usage). A negative correlation means that they tend to swing in opposite directions (for example, car ownership rates are negatively correlated with public transit usage). In general, consensus is to avoid negative correlations and seek positive correlations in GPPs, and the reverse in cash games. Putting positive correlations in your lineup will result in a higher variance for that lineup: more likely to hit big but also more likely to miss completely. Instead of splitting it up, I did it in kind of confusing way to get all the info on the same plot. Along the diagonal, the correlations are between players at the same position on opposing teams. For example, the top left square (on the first graph I'll post) shows a weak positive correlation (.073) between opposing QBs. Above the diagonal, the correlations are between players on the same team. For example, the second square in the top row shows a moderately weak positive correlation (.14) between the QB and RB1 on the same team. Below the diagonal, the correlations are between players on opposing team. For example, the first square in the second row shows a very weak positive correlation (.017) between a QB and the opposing RB1. I should add some kind of indicator for these different regions of the chart but I don't have any good ideas. Fantasy-point correlations Note: all correlations are multiplied by 100 to reduce the number of zeros. So the range is from -100 to 100.  Points to note: the correlations are modest overall. Obviously the QB is highly positively correlated with his receivers. Perhaps surprisingly the QB is positively correlated with the RBs on the team as well. The kicker is positively correlated with the big scorers on the offense and even more positively correlated with the defense. The defense is positively correlated with RB1 but has little relationship with anyone else (besides the kicker). Probably the most surprising to me is the positive correlation between RB1 and WR1 on the same team. On opposing teams, perhaps unsurprisingly, the biggest correlations are negative ones between the defense and the other team. There is some positive correlation among the passing games on opposing teams. But, in my opinion, these aren't the correlations to look at. After all, the WR1 and the WR2 always have the same QB throwing to them. They always play the same defenses, and experience the same game script. And those are factors that you are already taking into account if you are doing it right. So, taking these correlations at face value will result in an overestimation of the tendency of these scores to swing in one direction. What we want to know is not "how are fantasy points correlated" but "how are deviations from what I otherwise expect to happen correlated?" To get an index of general expectations, I used numberfire's 2014 weekly projections. Correlations in projected fantasy points  This graph shows, in a sense, the correlations in our expectations of player performances. My example stands out: WR1 and WR2 on the same team are much more highly correlated here than in the previous chart. Correlations in projection errors To put it together, here are the correlations in the projections errors. How much are the swings (relative to what we already expected to happen in our basic analysis) correlated?  Here, some of the correlations from the first plot are toned down; for example, the same-team RB1-WR1 correlation goes away. This correlation in the first plot was not due to unpredictable swings, but rather the factors that we are already taking into account when we rank the players every week. I think these are the correlations you should keep in mind making your DFS lineups. Conclusions Aside from the obvious ones (QB to his targets, kicker to rest of team, and defense to opposing offense), correlations should not be a big factor in your decisions. They are pretty small overall. Notes I assigned each player a position on the depth chart by ranking their fantasy-point projections (i.e., not the actual performances). I am saving rotoworld's depth charts for this season so this can be more accurate going forward. Also, I only have data from 2014 so the sample sizes are small. I could have incorporated 2015 week 1 but  . I have that data but it's in a slightly different format, but eventually I'll be incorporating 2015 data (if people are interested I could update this every week, but the changes won't be that interesting). Without actually calculating a significance threshold I can say with some confidence that all or nearly all of these correlations are not significant. Past fantasy-point totals are easy enough to get but I can't analyze the swings meaningfully without projections as well. And I am not comfortable assigning players to their position on the depth chart based on results alone. The fantasydata.com historical package has historical depth charts, but it costs $799 so I'll have to take down a GPP first  .
|
|
#
?
Sep 17, 2015 04:21
|
|
|
In case anybody wants it, I used NLFDB to build an archive of DraftKings fantasy points by player, by week, for every regular season game going back to 2009. It doesn't include DST but if there's demand for it I can do that. You can download it here from DropBox. I'm actually having quite a bit of fun messing with NFLDB and using it for some of my own purposes, so if anyone has datasets they'd like to see, let me know and I'll see if I can produce them. Edit: in case you're curious Jamaal Charles has scored the most DraftKings points in a single game since 2009 with a whopping 62.5 in this game. Sub Par fucked around with this message at 16:46 on Sep 17, 2015 |
|
#
?
Sep 17, 2015 16:42
|
|
|
So I'm building my lineups for the week and noticing I'm super heavy on Lynch and to a lesser extent Hyde. I feel good about both but how bad an idea is it to have so much exposure to one player? Should I diversify more or is it safe to lean on Lynch this week?
|
|
#
?
Sep 17, 2015 17:35
|
|
|
Sub Par posted:In case anybody wants it, I used NLFDB to build an archive of DraftKings fantasy points by player, by week, for every regular season game going back to 2009. It doesn't include DST but if there's demand for it I can do that. You can download it here from DropBox. It would be cool to see which players were closest to their projected points, and which were furthest away; or perhaps a list of "reliable" players who hit or exceeded their projected points at least 85% of the time or something
|
|
#
?
Sep 17, 2015 20:25
|
|
|
winvirus posted:It would be cool to see which players were closest to their projected points, and which were furthest away; or perhaps a list of "reliable" players who hit or exceeded their projected points at least 85% of the time or something I don't know that there are accessible data sets of past week-by-week projections from the various projection sources. We'd need that data in order to do those comparisons.
|
|
#
?
Sep 17, 2015 20:53
|
|
|
CompeAnansi posted:I don't know that there are accessible data sets of past week-by-week projections from the various projection sources. We'd need that data in order to do those comparisons. Doesn't numberfire have them? At least 2014's (see: a few posts up)
|
|
#
?
Sep 17, 2015 21:07
|
|
|
I saved numberfire 2014 weekly projections but I would love to have more. Feel free to use them yourself: https://www.dropbox.com/s/kimu8bw9i0hsdcv/numberfire-2014.csv This year I'm saving everything I can think of. Edit: I just checked numbefire and they do have weekly projections for every player. I'm wondering if what I scraped last year was not the weekly projection made just before that week but rather the weekly projection made at the beginning of the season for all future weeks. Hmm.. I also have CSVs from many but not all weeks, from the FantasyPros cheatsheet. It would take some work to put that all together. SurgicalOntologist fucked around with this message at 21:11 on Sep 17, 2015 |
|
#
?
Sep 17, 2015 21:07
|
|
|
CompeAnansi posted:I don't know that there are accessible data sets of past week-by-week projections from the various projection sources. We'd need that data in order to do those comparisons. Yeah unfortunately, getting a reliable dataset of pre-action projections going back a few years is difficult. There are a lot of people looking for something like that on reddit, and as best I can tell the people who have it aren't sharing. Edit: I can get some projections from FFToday, but I will need to write some code to scrape the data. I'll work on that tonight. Sub Par fucked around with this message at 21:29 on Sep 17, 2015 |
|
#
?
Sep 17, 2015 21:22
|
|
|
What's the best way to find GPP with overlay? I'm only on DraftKings so far but will probably diversify as the season (and my DFS knowledge) progresses
|
|
#
?
Sep 17, 2015 23:43
|
|
|
Tried out draft kings this Sunday. Since the games were already under way I did just Sunday/Monday night games. I spent $9 in various newbie games and only won $2, but I wish I had played that lineup more. Going for some full week games this week. But is there any added advantage/disadvantage to more limited games like the night games or just early games etc?
|
|
#
?
Sep 18, 2015 02:45
|
|
|
Fingers McGee posted:Tried out draft kings this Sunday. Since the games were already under way I did just Sunday/Monday night games. I spent $9 in various newbie games and only won $2, but I wish I had played that lineup more. Going for some full week games this week. But is there any added advantage/disadvantage to more limited games like the night games or just early games etc? Having less games to choose leads to scores being lot closer together since everyone will be picking from a smaller pool of players. Therefore there's a higher chance that one of your players having a bad game will cause you to miss out on cashing. Conversely if you're able to pick out whichever 3rd string rb can vulture a td then you' can move up higher so it's a tradeoff. Comes down to preference really but playing with a smaller slate is higher variance and something you might want to avoid as a newer player.
|
|
#
?
Sep 18, 2015 04:10
|
|
|
Teeter posted:What's the best way to find GPP with overlay? I'm only on DraftKings so far but will probably diversify as the season (and my DFS knowledge) progresses This here seems to be pretty handy. Haven't used it myself though.
|
|
#
?
Sep 18, 2015 14:54
|
|
|
I took a shot and started Mike Evans in a GPP that locked last night. He's only 0.2% owned. Looks like he's going to be good to go on Sunday. Go to work please, Jameis.
|
|
#
?
Sep 18, 2015 15:16
|
|
|
Gobias Ind. posted:I took a shot and started Mike Evans in a GPP that locked last night. He's only 0.2% owned. Looks like he's going to be good to go on Sunday. Hah, I did exactly the same in a couple of $1 fanduel GPPs. Even went with a Jameis/Evans stack on one. Go Bucs.
|
|
#
?
Sep 18, 2015 17:36
|
|
|
So i am going real light on the TE spot, so i'm picking between Crockett Gillmore or Richard Rodgers, i'm leaning Rodgers since the Seahawks are vulnerable against TE's and even more so without Kam. thoughts?
|
|
#
?
Sep 18, 2015 19:01
|
|
|
VietCampo posted:So i am going real light on the TE spot, so i'm picking between Crockett Gillmore or Richard Rodgers, i'm leaning Rodgers since the Seahawks are vulnerable against TE's and even more so without Kam. I would go Gillmore if you are punting TE. Rodgers is better as a player but the Packers have like nine other weapons to throw to and the Ravens don't really have anyone. And Week 1 they showed they will go to Gillmore with the game on the line, he just didn't come down with it against double coverage. Edit: Ugh I need to stop reading so much stuff before like Saturday. I'm talking myself in circles about Hyde this week. On the one hand, great talent, still cheap especially on DK, playing against a bad Steelers run D. On the other, short week traveling east in a game where he's likely to be behind and SF may abandon the run. JoshTheStampede fucked around with this message at 21:20 on Sep 18, 2015 |
|
#
?
Sep 18, 2015 19:03
|
|
|
Speaking of overlay, DraftKings has a $100k gtd NFL Freeroll this week. Don't forget to toss a lineup into there - first place is $10k.
|
|
#
?
Sep 18, 2015 21:24
|
|
|
Just did, and joined a few other smaller entry fee contests. I'm not used to this stuff yet so I don't want to kill my bankroll in a few weeks. Plus I'm sort of waiting for hockey to start as I feel like I might have more of an advantage there than football
|
|
#
?
Sep 18, 2015 23:02
|
|
|
Duro posted:Just did, and joined a few other smaller entry fee contests. I'm not used to this stuff yet so I don't want to kill my bankroll in a few weeks. Plus I'm sort of waiting for hockey to start as I feel like I might have more of an advantage there than football I feel you on that but with NBA. I feel like my grip on the NFL is decent but I'm really interested for NBA action to see how that compares. My only real worry is that the player pool will likely be a lot smaller and refined, whereas I feel like they're catering perfectly to fish like me during NFL season.
|
|
#
?
Sep 19, 2015 07:10
|
|
|
I love NFL but I don't think there's enough games in a season to even out variance. Even if you play cash games. If you're a degenerate like me who prefers tournaments then you will never beat variance. I mean, maybe in 20 years I'll have played enough NFL slates to know if my strategies work, but probably that still wouldn't be enough. For that reason alone I'm tempted to play other sports. But I'm scared off of NBA because of the need to pay such close attention to gametime decisions. And I'm scared off of MLB because of the ridiculous day-to-day variance, even if over the course of a super long season it would balance out. Actually, that gives me an idea. Do any DFS sites do multi-day MLB slates? Like you get 5 games of each team? That could be pretty awesome actually.
|
|
#
?
Sep 19, 2015 07:38
|
|
|
TheKevman posted:I feel you on that but with NBA. I feel like my grip on the NFL is decent but I'm really interested for NBA action to see how that compares. The key to nba is finding the value plays because you have to at least one star on your lineup
|
|
#
?
Sep 19, 2015 14:13
|
|
|
I only started playing this year, but I've definitely had the most success playing NBA.
|
|
#
?
Sep 19, 2015 14:24
|
|
|
TheKevman posted:I feel you on that but with NBA. I feel like my grip on the NFL is decent but I'm really interested for NBA action to see how that compares. They've started this massive ad blitz, I assume they'll try to keep all the new NFL fish like you and me into hockey and basketball seasons. The newbies will carry over.
|
|
#
?
Sep 20, 2015 02:33
|
|
|
My 3 dollar DraftKings line up for the week: Sam Bradford Carlos Hyde Marshawn Lynch DeAndre Hopkins Jordan Matthews Donte Moncrief Tyler Eifert Ameer Abdullah Titans
|
|
#
?
Sep 20, 2015 02:47
|
|
|
Well I won $8 on my free $3 MLB game. The rest of my CFB games today were all 0s  . Down $2.75. Hopefully Utah and UCLA can put some numbers on the board for me in the later game 
|
|
#
?
Sep 20, 2015 03:04
|
|
|
old dog child posted:My 3 dollar DraftKings line up for the week: Hilton might actually play so if he does be ready to switch out Moncreif for someone.
|
|
#
?
Sep 20, 2015 03:36
|
|
|
JoshTheStampede posted:Hilton might actually play so if he does be ready to switch out Moncreif for someone. I'd keep Moncrief. Revis is likely all over Hilton and Cromartie should not be playing so it will probably be a backup on Moncrief or he will play the slot. Andre Johnson is hot garbage right now and can't get separation and Luck will be throwing the ball a whole lot so Moncrief has some serious value especially how cheap he is.
|
|
#
?
Sep 20, 2015 03:50
|
|
|
e: what he said ^^^JoshTheStampede posted:Hilton might actually play so if he does be ready to switch out Moncreif for someone. If Hilton plays, he's drawing Revis so I'm willing to take the chance of a big game from Moncrief. Cromatrie won't be playing so whoever is covering Moncrief and AJ80 is going to be much worse. I'm only swapping out Moncrief if Hilton sits, but at this point starting seems more likely.
|
|
#
?
Sep 20, 2015 04:53
|
|
|
fantasydraft.com is a new dfs site and they had some big overlays last week. It looks like that will continue. If you want, use my referral link https://www.fantasydraft.com/play/rmexico
|
|
#
?
Sep 20, 2015 08:10
|
|
|
|
| # ? Apr 20, 2024 10:34 |
|
|
GlobalHero06 posted:fantasydraft.com is a new dfs site and they had some big overlays last week. It looks like that will continue. I signed up under you. Now I have one! https://www.fantasydraft.com/play/jabro
|
|
#
?
Sep 20, 2015 08:46
|
|