STRATEGY FORUM

Comments

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Inspired by the recent discussions about GPP structures, I created a webapp that attempts to simulate the long-term consequences of playing different contests for your bankroll. Check it out here:

    https://gppsim.herokuapp.com/

    It’s pretty minimal for now, and may not hold up to a lot of traffic, but please give it a shot. Let me know what you think. I haven’t yet put instructions into the app yet, so here they are:

    1. Choose your contest (click a row of the table to select it)
    2. Choose your Skill Percentile. This is the percentage of the field that you are better than (so, the default value of 80% means that your average lineup score is better than 80% of other players’ lineup scores). This just effects the averages.
    3. Choose your Starting Bankroll (in $). This doesn’t effect your chances of any specific contest but will affect how long you stay in the black.
    4. Click Simulate Contests.
    5. Wait. Be patient. If it takes longer that means the sims are not going bust.

    The app will run 20 simulations, ending each either at 3000 entries or when the bankroll gets below the contest entry fee.

    Currently there is no multi-entry effects simulated. I will probably not add those in as this is not accurate enough that this it would make a difference.

    I can add other contests. Post a direct link to any DK contest and I will add them when I get a chance.

    Caveat: This is for informational purposes only. I am not responsible for any decisions you make as a result of the information provided. All simulations are probabilistic and any outcome is possible.

  • thehazyone

    RG Contributor

    • Blogger of the Month

    This is pretty sweet – thanks!

  • jimmyquinella

    • Blogger of the Month

    Very interesting, stick with it!

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Thanks for checking it out.

    I just updated it to a new version that should be faster and more reliable (the last version was only running sims for one user at a time—oops!), and will show sims as they finish instead of waiting for all 20 to complete.

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    I’m seeing some errors in my log, and I am not sure what is causing them. So if anyone is trying to use it and having trouble please let me know what is happening.

  • edro990

    I just clicked the link, but only getting a blank page. Tried in Safari and Chrome, same result. Interested in giving it a test drive though…….

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Thanks for the report, I’m working on it.

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Alright, it should be operational. I’m not a web developer so dealing with concurrent users is new to me. Got it simming now in 5 different tabs without issue so we should be good.

    Note that the figure toolbar allows you to pan and zoom and even save the plot. Here’s a +70% ROI starting with 200 buy-ins that goes bust 19 out of 20 times:

    Edit: image doesn’t fit. Right-click and select open in new tab to see the whole thing.

  • anilprao88

    • 18

      RG Overall Ranking

    • Ranked #16

      RG Tiered Ranking

    • 2017 DraftKings FFWC Finalist

    • 2018 DraftKings FBWC Finalist

    This is really cool. Thanks for sharing. My main suggestion would be to allow for more entries and a larger bankroll

  • vorg7

    I’m pretty sure this is broken. I tried hundreds of simulations with skill set to 1 and 99 and a massive bankroll and there was no discernable difference. Unless the point is that it’s all luck. xD

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    @anilprao88 said...

    This is really cool. Thanks for sharing. My main suggestion would be to allow for more entries and a larger bankroll

    Done. Larger bankroll at least. I’m reluctant to expand the length of the simulations because I will have to pay for more memory and CPU time. I will consider a slider to adjust this (leaving the default small) if there is more demand, though.

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    @vorg7 said...

    I’m pretty sure this is broken. I tried hundreds of simulations with skill set to 1 and 99 and a massive bankroll and there was no discernable difference. Unless the point is that it’s all luck. xD

    Well, there is a hidden parameter which determines the relative importance of skill. I started with a value that I have calculated for my own purposes from data I collected over the years. I thought people would be too frustrated with the results so I doubled the effect of skill before publishing. I just increased it another 50%, so now skill is 3x the impact I believe is accurate.

    Yes, it is mostly luck, but skill should make a difference. I have confirmed in my tests. It might be more noticeable now, especially with the increased max on the bankroll slider.

  • vorg7

    Alright, I tried it out yesterday and got nearly identical average rois over 100 trials on each ($.25 pga) with a 500$ bankroll. 100 trials is still not a ton so maybe I got an incredible outlier in my results?

  • vorg7

    Now it seems to be working though! Thanks for posting this, actually a really cool idea.

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Yes, average ROI will be pretty constant unless you can get the bust rate up.

    Also I have personally been more concerned that the top-heavy contests have too much variance but it could equally be the case that the ridiculously flat contests do not reward skill enough to beat the rake. This is not something I have explored.

  • anilprao88

    • 18

      RG Overall Ranking

    • Ranked #16

      RG Tiered Ranking

    • 2017 DraftKings FFWC Finalist

    • 2018 DraftKings FBWC Finalist

    How are you modelling skill? Are you just using skill to shift the center of the distribution, or does it alter the shape of the distribution?

  • hendog

    • 930

      RG Overall Ranking

    • 2017 DraftKings FFWC Finalist

    Skill mainly shifts the center of the distribution, but the scale also shifts slightly proportionally to the center. This effect is pretty small though. I thought about adding noise to the scale parameter to model different play styles, but with only one entry per distribution this is overkill—as long as the overall distribution of points is accurate it doesn’t really matter if some of your opponent in the sim are the ones with more noise vs. others.

    If I ever implement multi-entry it might make sense to do this, especially for the gppsim user. More important would be a parameter of correlation among lineups.

    Anyways, maybe you assumed as much but the skill parameter of the user is taken from the skill distribution PPF function whereas for the opponents the skill parameter is randomly selected from the skill distribution. All the distribution shapes are estimated from actual data, not particular to the sport of the selected contest. Sports may have some differences in their distributions that are not modeled.

  • X Unread Thread
  • X Thread with New Replies*
  • *Jumps to your first unread reply

RotoGrinders.com is the home of the daily fantasy sports community. Our content, rankings, member blogs, promotions and forum discussion all cater to the players that like to create a new fantasy team every day of the week.

If you or someone you know has a gambling problem, crisis counseling and referral services can be accessed by calling 1-800-GAMBLER (1-800-426-2537) (IL). Gambling problem? Call 1-800-GAMBLER (NJ/WV/PA/MI), 1-800-9-WITH-IT (IN), 1-800-522-4700 (CO), 1-800-BETS OFF (IA), 1-888-532-3500 (VA) or call/text TN REDLINE 1-800-889-9789 (TN).