Aariv Iyengar

Super excited to introduce my #BigDataBowl submission for this year! I adapted the MLB statistic of Outfielder Jump to NFL safety play, using it to identify safeties who can be taken advantage of as "guessers": kaggle.com/competitions/n…

Admittedly, the usage difference isn't massive, but a quick paired permutation test shows that the difference is statistically significant. And all of this is dependent on the change actually "fixing" the tanking problem. But it is an interesting possible outcome for all this

And more importantly, how does this effect the playoffs? Could these new rules lead to more tired legs in the playoffs, with potentially worse basketball and more injuries resulting from that? Will be really curious how this plays out next year

One effect of the new lottery rules I've been thinking about: Part of the reason NBA March is such a slog is stars tend to take their foot off the gas to save energy against tanking teams. If bad teams are going to start trying to win, will stars have to stop resting in March?

A little late but here's STEPH's bracket for this year. Not really any surprises here, excited to see if it can stay perfect on finals winner picks

Full code is on my github: github.com/aariviyengar/S…

The "STEPH rating" score for the leaderboard is generated by running the model on every team in the range you pick playing a home series against the 2014 Atlanta Hawks. A higher win probability means a higher STEPH rating (the rating is just the win probability multiplied by 100)

STEPH 3.0 is officially out!!! This might be my favorite update I've made to the app. On top of making improvements to the model, I added a leaderboard tab! aariv-iyengar.shinyapps.io/STEPH_compare_…

Super excited to be presenting at this in a couple weeks!

@903124S classifying the role on a certain play it evens out enough over the entire dataset to get a good approximation of who is consistently guessing and who isn’t. Thanks for reading and for the feedback!

@903124S obviously there’s no way to make it a 100% pure dataset while still having a usable sample size unfortunately. That’s also kind of why I moved it more towards being a playstyle/role describer rather than an ability metric, I think that despite those concerns with properly

Super excited to introduce my #BigDataBowl submission for this year! I adapted the MLB statistic of Outfielder Jump to NFL safety play, using it to identify safeties who can be taken advantage of as "guessers": kaggle.com/competitions/n…

@Kar_tech2 Thanks so much!

@AarivAnalytics Really great work @AarivAnalytics! Excited to see what you do next!

@3murim Yea lol just finished my first sem, thanks so much for reading!

@AarivAnalytics tuff, u fr a ‘29??

@tangotiger Thanks so much!

Terrific adaptation and expansion!

I went into more detail on the Kaggle post, so check that out if this looks interesting! Full code: github.com/aariviyengar/H…

To back this up, I created another stat that I called Correct Guess Rate, looking at how often a safety moved in the correct direction within the first 5 frames. And sure enough, this was directly correlated with Jump.