• hendog

    • 2017 DraftKings FFWC Finalist

    Hello RotoGrinders community!

    (Edited to add this note) Grinders: feel free to use this thread to ask me anything about probability, statistics, modeling, programming, etc. Happy to give my thoughts on any relevant topic in the interest of getting some discussion going.

    I am pleased to announce a new one-of-a-kind service to the DFS community: personal statistical consulting and software development. It is no secret that some of the top DFS players either have advanced statistical or programming abilities of their own, or have hired data scientists and programmers. My goal is to make these services more accessible.

    Why am I doing this? First of all, I find the data-analysis issues surrounding DFS fascinating. I mostly play NFL so I’m looking for a year-round fix. I have been involved in many discussions of DFS strategy over the years; this is a community that is filled with people who have deep probabilistic intuition and complex ideas that they lack the technical skills to implement. I am also raising funds for a (non-fantasy related) sports analytics startup and trying to keep myself busy in the meantime. In general, I love sports analytics and I don’t want to get a real job!

    Some examples of services I can provide:

    • Automate your tedious daily workflow.
    • Implement a model based on your own intuitions and established statistical techniques, train it on historical data, and set it up to continue improving with more data.
    • Implement your own lineup construction strategy. Examples:
      • constraints you would like to use but no sites offer,
      • optimizing over multiple entries
      • optimizing only the unlocked positions in a late swap contest
    • Statistical analysis of your personal results. E.g.:
      • confidence interval around your ROI
      • simulations of different bankroll management choices
      • optimization of contest selection
    • Support for staking operations, for example developing a dashboard for your stakers to track results.
    • Backtesting your DFS strategy.
    • Upgrade from Excel to professional data analysis tools.
    • Personal instruction in statistics and/or statistical programming.

    These are just some ideas; with access to the right data, anything is possible.† You are my ideal client if you are already a RG premium subscriber, but you have a time-consuming process you would like to speed up, or big ideas that you don’t know how to achieve on your own. Just to be clear, I am not offering any particular data, content, or strategy advice. I am not selling my own methodology, rather I am offering professional assistance with your personal methodology.

    I will charge a simple hourly rate that is negotiable based on the project (although I could be open to other arrangements). PM me to discuss the possibilities; obviously, there is no charge for initial consultation. References are available on request.

    About me: I have been playing DFS since 2014 (almost entirely NFL). Since 2015 I have been mass entering lineups, I am not ashamed to admit. I have made every mistake you can think of: accidentally duplicating lineups, not replacing dummy lineups because of last-minute CSV issues, missing players because their name was spelled differently in my projections, etc. I have put a lot of thought not only into lineup construction and DFS strategy, but also these basic technical problems and how to avoid them. In 2016, I made a 5-figure DFS profit; in 2017 I attended the DK NFL live final and over the year profited low 6 figures.

    In 2017 I also completed a Ph.D. in Psychology. In particular I studied (roughly speaking) how the brain controls movement. My research developed new ways of analyzing human movement data (i.e. the kind of data they use for NFL Next Gen Stats). I am proficient in Python, R, SQL, and other languages. I have taught undergraduate- and graduate-level courses in statistics and statistical computing. Currently, I work part-time as a researcher in a sports medicine lab, where I help develop training methodologies in virtual and augmented reality and analyze motion-capture data recorded from athletes visiting our lab. I am also involved in the planning stages of a (non-fantasy related) sports analytics startup; we hope to offer advanced analysis of player-tracking data to soccer clubs around the world. Proceeds from the service I am advertising here will help us bootstrap this project.

    Caveat : I will not provide any assistance in circumventing legal restrictions of any jurisdiction or the terms and conditions of any website.

    (Although I did receive permission to make this post, I want to make it clear that this service is offered by myself personally and has no official relationship with RotoGrinders)

  • jimmyquinella

    • Blogger of the Month

    Sounds interesting, good luck with this project.

  • hendog

    • 2017 DraftKings FFWC Finalist

    Thanks Jimmy!

    Grinders: feel free to use this thread to ask me anything about probability, statistics, modeling, programming, etc. Happy to give my thoughts on any relevant topic in the interest of getting some discussion going.

  • TnRiddles

    • Blogger of the Month

    Building my own model is something I would like to do, But , I have no idea where to start

  • hendog

    • 2017 DraftKings FFWC Finalist

    Well, I would say there are two approaches to take.

    First, is to attempt to learn statistics. In my opinion this approach will take you further, but has a steeper learning curve before you actually have a worthwhile model. If you are comfortable with math, or took statistics in high school or college and remember anything, this may be a reasonable choice. Or if you have any programming skills, this method will allow you to skip Excel and move directly on to statistical computing (I would recommend Python or R) which is much more powerful, repeatable, and automatable. The type of model that would make a good starting point is called “regression” and there are tons of resources online—although not all of the same quality. But expect to have to spend some time understanding the abstract concepts before you can apply it to DFS.

    The second, and probably more realistic approach for most people, is an ad hoc approach. In my experience people who build their models this way eventually end up re-inventing regression. On the one hand, it means they wasted some time solving problems that have already been solved, but on the other hand it makes it much easier to learn the statistics once you realize it’s just a more direct way of doing what you already were doing.

    Anyways, the key here is to start small with an extremely basic model and then build it up from there. For example, you might predict that every player’s fantasy points will be the same as the points they scored in the most recent slate. Obviously, this model will not be very accurate and you shouldn’t actually build lineups with it, but it will provide a starting point. Set up an Excel or Google Sheet that implements this basic model. The equation you use will be very simple, just = another cell, or maybe a function like VLOOKUP depending on how your data is formatted.

    From there, ask yourself “how, specifically, is my model inaccurate? How can it be improved?” In this case, the next step would be take into account multiple slates, for example averaging each player’s last five fantasy scores (the specifics of how far to look back depend on the sport of course, and can be estimated with statistics, but for now you can work off your own opinions). This will require learning some new spreadsheet functions and organizing your sheet to incorporate more data.

    After that, repeat the same question: what’s missing? A reasonable next step is to take into account the strength of the opposing team (e.g. pitcher in MLB). If batters historically tend to get 1.3x their average fantasy points against a certain pitcher, then multiply the model output of those batters by 1.3. Another step would be to split your model into components, for example (in NBA) modeling minutes played separately from points per minute, then multiplying them together. For each component, start with a basic model that is just a historical average and building up from there.

    If you make it this far, congratulations, you’re not too far off from the methodology from many public projections. You have a basic starting point from which you can start including your own personal ideas on what factors might bump the predictions up or down.

    Of course, it is possible to go far deeper than this and there are plenty of advanced statistical methods out there. But don’t be daunted by that; if your problem is just knowing where to start, then start simple, get comfortable with how it works, and then add improvements piece by piece.

  • hendog

    • 2017 DraftKings FFWC Finalist

    I just wrapped a couple projects so I thought I’d bump this thread to see if there’s anyone else out there who could use my services… or just chat about modeling/scientific computing.

  • topspinner

    I wanted to put a plug in for Henry, I used to run a fairly complex spread sheet manually every morning that took about a half hour all told. Henry automated it and is giving me more up to date info and takes less than a minute to run now, for a reasonable price. Very smart guy.

  • Benf15harp

  • rannas23

    hendog can you pm me with what products your selling for a newbie like me. I don’t use any software…sometimes export to excel to filter and sort thru data a little big from other sites. Just wondering what kind of products you had and price. Thanks for your time and post here.

  • hendog

    • 2017 DraftKings FFWC Finalist

    Thanks for the kind words, topspinner. rannas23, I’m sending a PM.

  • dreamweaver1984

    im interested as well

  • dolo20

    Sent you a pm hendog

  • mark_james1985

    Hey I am interested. You are exactly the type of guy I have been looking for.

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

Subforum Index 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), 1-800-9-WITH-IT (IN), 1-800-522-4700 (CO) or 1-800-BETS OFF (IA). 21+. NJ/PA/WV/IN/IA/CO/IL only.