Live data from Hacker News

Lessons Learned from Two Years as a Data Scientist

dawndrain.github.io

11–20 of 89 posts

Re: Lessons Learned from Two Years as a Data Scientist

#11
post #2

Nice pictures. (Didn't read the words... but I liked the pictures.)

I was distracted (in a good way) by the pictures

but I did read most of it. I thought it was a fairly honest and good blog

I did like the diagram of the different company styles

I'll will read it 100% when I have enough time

Re: Lessons Learned from Two Years as a Data Scientist

#13

I was hoping for more DS related stuff. It almost sounds like you're learning to be a SWE! The investing section is curious. > On brilliant advice from the man who arguably went from mere millions to decabillions faster than anyone in modern history, I've put a large chunk of my money in leveraged index funds and etfs Who are you referencing here and did you do any DD other than taking his advice? I'm wondering what…

> did you do any DD other than taking his advice?

I did some backtesting simulations that made leveraged investing look pretty awesome. The effective borrow rate for funds like spxl is crazy low, way better than if I were to borrow myself. (Also, fwiw I was pretty conservative and am overall only around 2x-leveraged.)

The internet is very opposed to leveraged investing imo, but I think most of the concerns are pretty dumb. There was this one blog post where this guy ran ten simulations of his own, most of which showed the leveraged portfolio doing comparably to the baseline, but one a couple showed it doing worse and one saw the leveraged portfolio 100x'ing or something... and he concluded that it wasn't worth it??

People will also appeal to volatility drag as a superficially sophisticated knockdown (in short, imagine all four two-step paths in which the market goes up or down by 10% at each step. Then the baseline market averages out to (.81 + .99 + .99 + 1.21)/4 = 1, and a 3x leveraged portfolio averages out to (.49 + .91 + .91 + 1.69)/4 = 1. Volatility drag is those two middle worlds where the leveraged portfolio does badly despite the market as a whole basically ending up where it started.

Re: Lessons Learned from Two Years as a Data Scientist

#14

I was hoping for more DS related stuff. It almost sounds like you're learning to be a SWE! The investing section is curious. > On brilliant advice from the man who arguably went from mere millions to decabillions faster than anyone in modern history, I've put a large chunk of my money in leveraged index funds and etfs Who are you referencing here and did you do any DD other than taking his advice? I'm wondering what…

> I was hoping for more DS related stuff. It almost sounds like you're learning to be a SWE!

It kind of felt that way too :). Some more data sciency things were learning how transformers work, hacking pytorch, using visualization tools like tensorboard and wandb, web scraping, better using parallelism, tuning hyperparameters (mostly the learning rate tbh), better fluency with the command line than I assume most swe's need, getting very comfortable inspecting data, making experiments more reproducible, reading lots of papers, writing papers, and trying (somewhat half-heartedly) to get published.

Re: Lessons Learned from Two Years as a Data Scientist

#16
About Java vs Python, modern Java would be:

    import java.util.ArrayList;
    var cars = new ArrayList();
Python would be:

    cars: list[string] = []
The big difference seems to be that ArrayList is not a "default" data structure in Java, but it is in Python. While I like the Python example better, I'm not offended by the modern Java.

Re: Lessons Learned from Two Years as a Data Scientist

#17
post #6

I was hoping for more DS related stuff. It almost sounds like you're learning to be a SWE! The investing section is curious. > On brilliant advice from the man who arguably went from mere millions to decabillions faster than anyone in modern history, I've put a large chunk of my money in leveraged index funds and etfs Who are you referencing here and did you do any DD other than taking his advice? I'm wondering what…

The leverage means they go up faster, but they also go down faster. In regular investing you feel good when you get a modest return, and bad when you experience a modest loss. With leveraged funds you feel like a genius when the market goes up and like a complete moron when you get wiped out.

I understand how leverage works, just not in the context of an index fund or ETF.

From your description it sounds like the leverage is baked into it, so it's kind of like "safe" leverage in that you can't lose more than you have.

Re: Lessons Learned from Two Years as a Data Scientist

#18
post #12

I love the tip about using the python debugger "pdb". This reminds me of the similar Ansible debug feature (e.g. "debugger: on_failed") which let's you jump into an in-flight playbook.

I remember using OCaml a bit and it has a time-travelling debugger. You launch the program with the debugger, it crashes and then you can inspect the program as you wish. I was really impressed by this.

Re: Lessons Learned from Two Years as a Data Scientist

#20
post #7

I was hoping for more DS related stuff. It almost sounds like you're learning to be a SWE! The investing section is curious. > On brilliant advice from the man who arguably went from mere millions to decabillions faster than anyone in modern history, I've put a large chunk of my money in leveraged index funds and etfs Who are you referencing here and did you do any DD other than taking his advice? I'm wondering what…

Almost definitely Warren Buffett, who went so far as to make a very public bet that index funds performed better than hedge funds (and thus most active investing). You can see details on the deal and outcome here https://www.investopedia.com/articles/investing/030916/buffe... Also, he's got tens of billions. Edit: Or maybe not on reflection about the "faster than anyone" bit. I dunno.

Sam Bankman-Fried, unless I miss my guess.
Post reply on HN