This is my current "hey nice to connect on Linked-In can I help you" page. Here's the (almost) free stuff:
- A book on prediction networks. Audio for chapter one here.
- Free open-source software ... see here
- Papers, articles, blogs etc are listed on my home page
If you run a business I will briefly list the small set of specific things I can legitimately help you with:
or market data provider, or someone who needs to optimize execution, then I can enhance your already high quality real-time data feeds (mids and short horizon predictions) leveraging a community of 6,000 quants. There's nothing mysterious about this. I just harvest the best univariate martingality attackers. See mid-one contest for an example of this type of crowd-sourcing. Please don't expect more than about 50.7% directional accuracy over short horizons. Please don't expect to profit from this if you are crossing the bid-offer spread. I don't like to over-promise.
I can help you use the world's largest stock-pile of algorithms for detecting small departures from martingality. You will instantly become the envy of all your model review friends, and establish yourself as a paragon of humble statistical virtue in your firm, just because you managed to get over your own cleverness whereas others could not.
There's a free tool at https://pi.crunchdao.com/ where you can quickly test your hunches. It will expand soon.
You should. It is a lot easier than it was when I wrote my book. We can talk.
... through some portfolio optimization that is entirely novel and unpublished, I can provide the algorithm via API and Python tools for evaluating it easily. How sneaky is it? Very. It is a cunning cunning plan my Lord, using a cunning cunning algorithm I wrote years ago and recently repurposed. It is so very cunning it is even more devlishly cunning than unifying the two major schools of portfolio theory in a mathematical insight (I should know I'm the one who did that, seepdf). I've tested this new, cunning cunning algorithm against thousands of time-series and on opinionated and unopinionated portfolios. I'm in talks to create a new index. But I can probably enhance your opinionated portfolios as a final step by making extremely defensive use of dependency information.
I'm quoting 20bps because that's the actual long term result, probably, not an exaggerated claim. If that doesn't sound like a lot, you are not a likely customer for this and that's cool.
Or Mid+One in particular, see discord and look for the midone channels.
You don't need to. If you join the Google Meet listed here on any Tuesday night between 8:30pm-10:00pm EST there's an 80% chance I'll be there. I would rather commit to this than schedule deterministic interruptions to my work I usually forget.
I don't have any structured guidance and can only suggest some bespoke opportunities to learn:
- See lecture one, lecture two and lecture three put together by the Broad Institute at MIT and Harvard. That's your biology crash course. Then you'll be ready to take H&E images and predict transciptomics data. In other words, predict the result of an expensive test procedure from a cheap one.
-
Also new: a $120,000 contest to detect small departures from martingality. You'll love the spash page where you can play space-invaders.
-
If you are interested in generative methods there's also this long-running eleven-dimensional game I wrote to challenge you. Its a way to get familiar with Copulas and that sort of thing but you might very rapidly get to or beyond the SOTA for risk management by participating here. There is currently no prizemoney but I strongly suspect that will change soon.
See CrunchDAO.com for more challenges
I don't know but here's a reading list that's pretty narrow. It's all I've got.
Not really. Well, maybe just this:
Never, ever, ever kid yourself into thinking that you can determine that something mathematical or technical-but-not-arbitrary won't be useful or beneficial to try to understand.
Brief comment here on the topic.
Currently not but if you are good enough, you won't need a job. Create an awesome algorithm on CrunchDAO and hand it down to your grandchildren.
Oh boy, terrific. Best way would be to:
- Read my book.
- Sign up at CrunchDAO.
- Go through the process of entering a contest to help advance the science of causality, say, or help detect cancer.
- Ask questions in the discord and gain an understanding of what might help data scientists build models.
- If you have some cool tool that is free forever that a community of quantitative people could use to assist them, by all means recommend it.
Much appreciated!
You can help me by...
- Not asking me if I've considered using AI in my business.
- Not asking me if I need consultants.
- Not replying to the link I sent you with "what is this link?". It's a link. You discover what it goes to when you click on it. (You cannot imagine how often I've got this question).
- Not forcing me to very graciously decline your kind offer to discuss my org chart in great detail with a complete stranger.
- Not reaching out to me with a pitch that begins with a declarative statement like "Firms are doing blah blah blah" or "Companies are finding that ...". This will make me want to strangle you.
- Not offering a limited time trial of anything at all, ever.
For example:
- Constructive feedback on my various writings are invariably greeted warmly (see my home page).
- Ah forget that. I really just want effusive unconditionally positive 5-star reviews for my book.
- Introductions to active early to mid-stage investors are appreciated
- I'm sure you're dying to write more unit tests for midone or timemachines or humpday.
- Nah? Well go win lots of money here instead.
- Introductions to financial market participants whose interest might be piqued by our unique setup are appreciated.
- Introductions to extremely good developers (solana/web3; kafka etc/streaming analytics; fx/ir/market making algo)
- I'm sure you can think of something else.