Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

FinRL: The first open-source project for financial reinforcement learning

github.com

Post reply on HN
Print view View original

41–50 of 58 posts

Re: FinRL: The first open-source project for financial reinforcement learning

#41

Post by nickles »

charcircuit wrote: post #37
nickles wrote: post #35

Earlier quoted context omitted.

Equities settle T+2. Without a margin account, Reg T requires that cash be tied up until a trade is settled. It’s not viable to day trade like that.

Doesn't the exchange match people with another person who is buying / selling it? Every single cryptocurrency or video game exchange I have used has settled pretty much instantly. How hard is decrementing the shares of one user and incrementing the shares of another.

> Doesn't the exchange match people with another person who is buying / selling it? Every single cryptocurrency or video game exchange I have used has settled pretty much instantly.

There are multiple exchanges where stocks are traded, and order routing is typically transparent when using retail brokerages. The exchanges only provide the matching engine, while brokers handle custody. The process of transferring the ownership between brokerages is the clearing step which is handled by the DTCC.

Cryptocurrency exchanges provide both the matching engine and custody, so transferring ownership can be done instantly by updating records in the exchange’s database. The clearing step occurs when transferring coins to another wallet or exchange, when the transaction actually hits the blockchain.

> How hard is decrementing the shares of one user and incrementing the shares of another.

It’s surprisingly difficult. Clearinghouses (centralized) and blockchains (decentralized) exist to solve this very problem.

Re: FinRL: The first open-source project for financial reinforcement learning

#42

Post by xxxtentachyon »

charcircuit wrote: post #37
nickles wrote: post #35

Earlier quoted context omitted.

Equities settle T+2. Without a margin account, Reg T requires that cash be tied up until a trade is settled. It’s not viable to day trade like that.

Doesn't the exchange match people with another person who is buying / selling it? Every single cryptocurrency or video game exchange I have used has settled pretty much instantly. How hard is decrementing the shares of one user and incrementing the shares of another.

You can settle as quickly as you want if you’re engaged directly with your counterparty. But yes, T+2 is wildly outdated and will likely move to T+1 in the next few years. Longer term, anything less than same-day settlement both wastes capital (of BDs who need to post collateral while their customers’ trades wait to settle) and risks contagion (since each additional day of settlement gives a party more time to go bankrupt) IMO

Re: FinRL: The first open-source project for financial reinforcement learning

#43

Post by l337 »

artemonster wrote: post #39
harveynick wrote: post #31

Earlier quoted context omitted.

https://www.amazon.co.uk/Quantitative-Equity-Portfolio-Manag... Caveat: it’s not exactly “easy reading”, and you might want to have the “three blue one brown” YouTube channel on standby.

Watched every video of them, I am okay with math. Thank you very much! Btw, since quant industry is full of secrets - is there any source on some actual money figures - how much firms make, what are the sums that are required to be shuffled, what are the frequencies of trades, profitability figures? Anything to get a slight glimpse into inner workings of this. Unfortunately googling turns out nothing or scams

Hedge funds are a very diverse group: from small shops managing tens of millions to the large funds managing tens of billions.

The very good ones probably make around 20% a year (meaning that they earn around $40M per year per $1B that’s under management).

HFT companies are a different story. The biggest ones make around $1B/year (e.g Virtu, Flow Traders which are public), with a few of them even bigger.

There are many tier two companies making tens - few hundred million dollars per year.

So none of them are really big (compared, to say, a big bank), but they do make a lot of money per employee, and the salaries reflect that.

Re: FinRL: The first open-source project for financial reinforcement learning

#44

Post by charcircuit »

v0idzer0 wrote: post #40
charcircuit wrote: post #37

Earlier quoted context omitted.

Doesn't the exchange match people with another person who is buying / selling it? Every single cryptocurrency or video game exchange I have used has settled pretty much instantly. How hard is decrementing the shares of one user and incrementing the shares of another.

> How hard is decrementing the shares of one user and incrementing the shares of another. whose database are you incrementing/decrementing in? Real life stock trades happen across multiple exchanges with multiple intermediaries (brokers, market makers, end users). Without a blockchain, this is not trivial to coordinate safely

>whose database are you incrementing/decrementing in

Your own. An an exchange you own 10 total shares and in your database you store who those 10 shares belong to.

Re: FinRL: The first open-source project for financial reinforcement learning

#45

Post by charcircuit »

xxxtentachyon wrote: post #42
charcircuit wrote: post #37

Earlier quoted context omitted.

Doesn't the exchange match people with another person who is buying / selling it? Every single cryptocurrency or video game exchange I have used has settled pretty much instantly. How hard is decrementing the shares of one user and incrementing the shares of another.

You can settle as quickly as you want if you’re engaged directly with your counterparty. But yes, T+2 is wildly outdated and will likely move to T+1 in the next few years. Longer term, anything less than same-day settlement both wastes capital (of BDs who need to post collateral while their customers’ trades wait to settle) and risks contagion (since each additional day of settlement gives a party more time to go ban…

Don't you mean the opposite? Shorter is better since there is less risk?

Re: FinRL: The first open-source project for financial reinforcement learning

#46

Post by xxxtentachyon »

charcircuit wrote: post #45
xxxtentachyon wrote: post #42

Earlier quoted context omitted.

You can settle as quickly as you want if you’re engaged directly with your counterparty. But yes, T+2 is wildly outdated and will likely move to T+1 in the next few years. Longer term, anything less than same-day settlement both wastes capital (of BDs who need to post collateral while their customers’ trades wait to settle) and risks contagion (since each additional day of settlement gives a party more time to go ban…

Don't you mean the opposite? Shorter is better since there is less risk?

Yes, my phrasing was poor. I meant to say “any reform that does not go as far as same-day would be bad”

Re: FinRL: The first open-source project for financial reinforcement learning

#47

Post by Galanwe »

nightski wrote: post #18
mumblemumble wrote: post #11

Quick hot take: when the project has a Venmo QR code for soliciting donations, you can safely assume they have not actually figured out how to make lots of money on the securities markets. Which is not me looking to trash it. I'd just encourage people to be aware of the context. This is an academic project which has apparently only been tested on hindcast data. Real life trading may present challenges that were well…

I'd agree with the skepticism. That said, it takes a minimum of 25k to get started using something like this in the real world due to the PDT rule (honestly more than that so you don't dip below 25k). It's possible they just aren't ready to throw that kind of capital at it yet. Also even if they aren't using real money that does not mean they are limited to historical data. You can run algos on live data even without…

What's presented here is not a full blown strategy, it's just a framework for building signals (+ some of example signals) from which a strategy could be built, or to complement an existing strategy.

The signals presented are too weak to beat their cost and OF to generate money if traded, you would need at least 50 others blended together.

Its like listening to a bass player alone and thinking "I should put that guy in a live concert, he will earn millions", while the reality is you would need a guitarist, a drummer, a singer, etc.

Re: FinRL: The first open-source project for financial reinforcement learning

#48

Post by Galanwe »

t_mann wrote: post #15

I really can't help but wonder - say I want to create a trading strategy that I actually want to use. Is there really any point in learning an open source framework, where I can assume that the most profitable strategies it can produce are already being used by someone? I think I'd rather start by learning the underlying frameworks like TensorFlow, that this one seems to wrap, and produce something that has at least…

> Is there really any point in learning an open source framework, where I can assume that the most profitable strategies it can produce are already being used by someone?

In practice, yes, because the methodology itself is only as good as the data you feed it.

There's only so much you can do with e.g. ML. The framework gives you that methodology. You can fiddle with it, fine tune it, etc, but that's often just marginal gain. The real juice is in the features that you prepared, and the data you have available for these features.

Re: FinRL: The first open-source project for financial reinforcement learning

#49

Post by Galanwe »

bravura wrote: post #30
melony wrote: post #28

Earlier quoted context omitted.

Get an internship at a reputable prop shop or hedge fund (that specialize in quant finance). There are plenty in Switzerland.

Aren't they going to impose a strong non-compete agreement on you, so that once you've actually learned something you are handcuffed to them?

As mentioned in comments, it's pretty rare to give a strong non compete to interns.

But interns aside, the whole industry is built around these competes, so mechanism are in place to circumvent them.

Some HF will happily buyback the non compete, some will offer a welcome bonus to compensate, some will be ready to wait, some will make an offside deal with the departing company, etc.

Re: FinRL: The first open-source project for financial reinforcement learning

#50

Post by Galanwe »

torbTurret wrote: post #32
Galanwe wrote: post #2

As someone with 10+ years of exp in the industry, I had no clue what this was actually doing after reading the github page. Had to dive through the examples and videos. The alphas showcased seem weak: 1.5 max gross SR in sample, unresidualized, unconstrained and standalone. That would need to be blended with a hundred un correlated others to have a decent portfolio.

Quants: So much fancy vocabulary just to consistently underperform your favorite mutual index fund :)

Do you really believe the world is that simple?

It poses you no problem to live in a world where your outsider opinion on a topic you probably just read about on the internet, is right versus a whole industry of very smart people managing hundreds of billions?

At no point do you wonder that if thousands of smart, successful people and companies do something, then maybe that something is not as absurd as some internet blog posts may suggest?

Post reply on HN