Hacker News Reader

Comment by swanson | Hacker News Reader

Funny enough, the automation/optimization/scripting/algorithm aspects of DFS are a big part of what makes it interesting to me. I have domain knowledge about sports, so DFS is a better testbed for me to learn and explore these areas than, say, automated stock trading.

It's a great place to practice linear algebra, machine learning, forecasting and projection models, web scrapping, data cleaning/normalization, and API wrangling. And all without needing a brokerage account, minimum account balance, or transaction fees!

View on HN

I find Quantopian is a pretty good testbed given you can get their stock data for free in their contests.

No need for a brokerage account, minimum, or any sort of fee.

https://www.quantopian.com/open

I never thought about it this way, and while you don't have to worry about brokerage accounts and minimum fees, you have to pay for accurate stats. Using a place like ProFootballFocus or StatsInc is going to cost you a pretty penny if you want to scrape an API with up-to-the-second accurate information.

You don't need "up to the minute" information though. Data can be days old and still be viable as long as you're using twitter's streaming API to get injury updates from one of each 32 teams beat writers as they occur. If I'm setting an NFL lineup on Friday I know the stats haven't had an opportunity to change in 6 days.

Monday night football! So not quite 6.

Injury reports are serious wrinkles. Often time players will be scratched within a few hours before the game. If you are optimizing, you need to account for this because it sends the backups value much higher than his salary. For example, last night Rajon Rondo was scratched. His backup Darren Collison, was in over 80% of the lineups in a FanDuel 50/50 tournament. I didn't have him, and I lost $2.

It also depends on the data you want. Most data sources don't have all of the things you could want to use in order to create meaningful stats.

ESPN Boxscores -- or FanDuel/DraftKings themselves -- have this data available.

Or there are projects like https://github.com/BurntSushi/nfldb

I got into this for the very same reason. If you are in it for these academic aspect, would you be interested in teaming up?

dope. let's connect.

mwfd541 at gmail

I literally created a ASK HN post about teaming up with someone earlier today but got zero responses: https://news.ycombinator.com/item?id=10852495

Mind if I get involved too? I can shoot you an email.

You could also apply these same skills within NLP (Natural Language Processing), replacing "Sports" with "Linguistics" and provide a lot more positive net impact on the world

True, perhaps... but whenever I'm learning something for fun, I apply it to something without any applicable human benefit.

I dabbled in machine learning algorithms... but just so I could play board games with my computer. As a student, I think it matters less what you're applying your emerging knowledge to.

Reply on news.ycombinator.com