search

matthewdwood82 / ETR-Slack-Data-Science-Book-Club

Continuation of ETR Slack Book Club, focused on data science topics. General comms via slack/discord; book discussion via Github Issues & Google Meet.
1 stars 0 forks source link

DFS applications/projects? #5

Open jdsdog10 opened 1 year ago

jdsdog10 commented 1 year ago

Thread for possible applications for our texts

jdsdog10 commented 1 year ago

Curious if you see any applications from BSM so far? My current interests don't include much projection building. More so simulating a pool of entries for a dfs contest (given player ownership projections) and also simuation of sporting events (given various marketshare assumptions)

jdsdog10 commented 1 year ago

In my soccer sims, I have projected stats for each scored stat_line. More or less I treat that projection as my p/lambda for a binomial/poisson distribution and simulate. Might be worth investigating if there are any gains from treating my parameters as non-fixed and making use of a conjugate prior... Edit: On second thought, the poisson-beta conjugate prior is dependant upon observed data, which is not really something I want to do since abilities of opposing teams can fluctuate alot (Lebron and the Lakers will score alot more playing against an NCAA team than they would the mavericks for example)

matthewdwood82 commented 1 year ago

Thinking a little bit our discussion on Week 3 XFL Kirklin/Burnett late swap decision, I think there's a nice application that might not require building a projection per se but might use an off-the-shelf one to calculate some proposed outcome distribution and use it to get a rough wag at the likelihood that your player(s) remaining are in a LU that achieves some score X, and then make the same estimate for other combinations of player(s) at that salary who would also reach that score with their respective likelihood

Combine with some naive assessment of the # of LUs that do have that combination (using off-the-shelf product ownership * field size), and you have an estimate of your win equity from one or more sets of players that would fill out a LU in late swap.