Hacker News new | past | comments | ask | show | jobs | submit login
Goldman Sachs will open-source some of its trading software (wsj.com)
189 points by newleaf on April 3, 2019 | hide | past | favorite | 96 comments



Of course, if this technology was able to price securities better than the market (the fundamental job of almost every market participant), it would be printing money and they would not release it.

This is not the secret sauce, but probably an implementation of a set of standard well known pricing and risk models. That's still useful, and can be expensive to develop, so thanks Goldman.


> the fundamental job of almost every market participant

Not true. You might make an argument that this is the effect of having them together in a market, but that's not their job:

- Market maker: hang around the market offering to trade with anyone (pref retail) at a spread. Doesn't care whether TSLA is gonna be able to make all those Model 3s.

- Pension fund: make sure they can pay the liabilities that are coming due. If that can be locked in, happy to pay a bit more than fair value to do so.

- Hedge fund: make absolute returns. Buy before it goes up, sell before it goes down. Whatever form of voodoo (or skill) fulfills this is fine. This doesn't have to mean finding the right price (could just mean you guess which way it's going), though of course often it is part of the objective.

- Broker: finds people on both sides of a trade. Doesn't care terribly much except to create excitement.

- Banks: lend money/securities and offer services to all of the above. Create research to make people trade. Securitise stuff so people can trade it. Often do a bit of everything.

Source: used to run hedge funds.


Hope you dont mind me asking a slightly unrelated and possibly dumb question.

Im a programmer, but on the side have made on average 150% profit in share dealing over the passed 5 years, but more importantly, a much higher return in other assets to 2 orders of magnitude higher.

My question is - would experience / gains like this - if i had proof etc - get me an interview in a fund as some type of well paid (6 figure atleast) analyst?


Yes, go on LinkedIn, there's a bunch of recruiters hawking jobs at funds. Generally they want a Sharpe ratio above 2.5, $10M profits a year.

And you need to know what machinery you'll need too. Capital requirements, counterparty agreements, access to stock lending, cost requirements, IT requirements, everything.

Some of the newer shops say you can keep your own IP. Haven't checked whether it's true, but most people I know are sceptical.


ok while we're playing "ask a hedge fund guy"

Say I set up a fund holding a low cost s&p500 index ETF, but at the end of each year sold naked puts with a ~1/25 risk of ruin to earn ~4% return. Therefore my fund consistently makes 4% over the market index, except for 1/25 years when it explodes and loses everything. Because the volatility is low, my sharpe ratio is good (until it explodes), correct?

Assuming it can stay in business >10-15 years won't I be a billionaire hedge fund manager by then and then change to a low risk strategy that only makes 1-2% more than market index with very low risk of ruin and just let my investors lose interest and quit the fund over the next decade while I continue to earn fees from them?


Investors will ask you what you're up to, and if you're just doing that they won't invest. They also keep an eye on whether you're doing what you say.

Anything that's both simple and mechanical is gonna have problems attracting investment. The guys you're talking to are gonna have problems justifying giving you 2/20 for buying a fund and selling options.

Or should. I've met a lot of investors who didn't ask the right questions.

Regarding the Sharpe, if they know what they're doing they're not just using the textbook version either. There's a paper by Andrew Lo about it, well worth a read, not terribly complex math.



Yes


But it would be pretty easy to obfuscate the whole thing behind a few layers of "We've got 50 PhDs working here, we run black box algorithms" etc etc, the strategy could be artificially made a couple of orders of magnitude more complex while still producing the exact same outcome.

Do hedge fund investors keep an eye on whether you do what you say? How did Bernie Madoff go for so long if that's the case?

Of course the mechanism is simple and mechanical, but that isn't how you'd market it to investors.

I'll go read the Lo paper...


The 50 phds thing is exactly what certain large firms are doing. The emperor's new hedge fund. I know people who've worked at major firms, and they've told me what they do. But you are going to have a hard time doing a startup with 50 phds.

Madoff was the cause of all the due diligence, though I'd say Europe was a bit different from the US at the time.


Strategy discussion aside, a 4% risk of ruin is not “low” by any means. A sophisticated investor would never accept.


If you are making $10m profits a year, what do you need a job for?


If [0] those returns are accurate, can be proven, and your strategy can be scaled up to multi-million dollar positions, you should take it to a hedge fund/prop shop/etc. and be able to make 7 or 8 figures. Your bonus should be dependent on performance; you may be able to negotiate a % of the profit. Those companies don't care about your background if at the end of the day you make them money.

[0] Granted, that's a huge "if".


I forgot to add the "worst" bit.

It was mostly a gut feeling, after lots and lots of research.

My question basically is, if I show I have been really good in the past - without a specific model - is there anyway I would be taken seriously.

Thanks


and how do you prove some method scale up without either already being on the million dollar scale or disclosing your secret sauce?


The reason financial transactions don't scale is volume - your $100 trade can move a penny stock to $.02 from $.0002 a share if there is no volume.

Some markets can absorb an enormous amount of volume, such as the FOREX market.

Now, when you start making good trades in a "big" market, intelligent players can mimic / play off of them, so it is hard to prove in that regard.

At a minimum though, you could prove that it works on assets that are not volume-dependent.


Short of spending a million dollars right then and there, you can use historical market data. Or datasets provided by the person you're trying to prove it to. Keep in mind that after a certain volume, your model would start influencing the market itself.


You mean +50 or +150% ?

Either way, it's insignificant and you're actually losing money. The stock of all the big companies went up tremendously in the past 5 years. Anybody who invested in large US equities made just as much if not more.


You missed the key sentence after it, I have made what one consider a very good increase in other assets which I would discuss with the fund


I think you're missing the gist of his statement and also overlooking the word "almost"?

That being said, most of your definitions are still dependent on competitively pricing securities. A market maker who can't calculate reasonable theos won't be a market maker for long.


> most of your definitions are still dependent on competitively pricing securities

No, the MM doesn't care if TSLA is overpriced. He just sees where everyone is and makes a market roughly there.

> You can't provide quotes if you don't have something to quote around.

But you don't have to quote around the actual value of the item. That's the point.


I work in a MM group so take this as you will, but the days of profitably making markets by just fitting everything to the screens are long gone.


I think it depends a whole lot on what exactly you are market making and how you're trying to profit. Of course many desks take views as well, but mixing in prop positions doesn't change the essence of it: the MM is trying to make money off being available for other participants.

You can of course also learn some things about where to market is going in the course of this business, and many desks are able to piggy back on some flow information for their advantage.


The only real practitioner in this thread has one of the only downvoted comments. Incredible.


One little correction.

Hedge Fund: Convince customers to invest their money in the fund. Extract as much of it as possible through fees. Doesn't care much where the market goes, charge a fee either way, a bit more if it goes up.


Thanks for being so concise. Getting license to sell estates (life insurance), hope to make $$$, worm my way into FINRA to take series 7, then try operate under reg crowdfunding to raise $1 million capital for new biz. What you do now?


It's akin to Facebook open sourcing their social network framework, rather than the entire personal information of all their users. Of course it's useful to some people, but it's not what makes them money.


It's probably not a download it and run it piece of software, there are probably lots of configurations and likely requires the user to provide their own models and trading rules.


Investment banks don't even have the secret sauce, hedge funds like Renaissance Technologies and D. E. Shaw & Co. are the ones that have it


It's a bit disingenuous to say that BB's don't have a stake in HFT.

At the time of Aleynikov's case, Goldman was routinely at the top of the NYSE rankings with regard to programmatic trading volume.


> Goldman was routinely at the top of the NYSE rankings with regard to programmatic trading volume.

I took 'secret sauce' in GP's comment to mean 'profitable strategies'. Don't confuse volume with profitability. It's conceivable that a BB would deliberately lose money in some activities to have clients give them other, more profitable flow/business.


That's true. It's "LeFevre's corollary", if you will :)

N.B, Lapdance not gauranteed


Investment banks have proprietary trading desks that are effectively hedge funds.


GS used to be nicknamed “the worlds largest hedge fund”. They have plenty of world class proprietary tech, but the OP is right: this is not likely to be secret sauce at all.


That's Harvard's endowment. Harvard University itself exists at this point merely as a tax dodge for the $37 billion AUM hedge fund.


I don’t think this is accurate. Regulators release the Volcker Rule to shut down prop trading at banks.


The Volcker rule contains exemptions for certain prop trading activities. For example, trading in US government bonds is exempt.


Here's nice flowchart[0] to determine what counts as proprietary trading. There is a surprising amount that is still allowed.

https://www.davispolk.com/files/DavisPolk_Final_Volcker_Rule...


There still exist principal investing groups, as well as market making groups that can contribute to the bottom line.


Right. They have tons of trading activities that are non-proprietary. I’m sure they can always hide some positions that are truly proprietary but largely they are not supposed to put on positions for their own gain.


To a regular software developer who uses (and does extremely minimal contributions) open source software regularly, this seems more for marketing / branding (even from an HR perspective) than for anything else (like a contribution without returns in mind).

We're reading about how GS "will" open source some software on WSJ, it would be better to just see a blog announcement with a link to the repository.


What about that dude who they put in jail for open sourcing software he developed in a personal side project while under contract with them?

Didn't they like completely throw the book at him by charging him under that completely insane law the Computer Fraud Act?


That account is quite different from what actually happened. See my comment on this: https://news.ycombinator.com/item?id=9044805.

In particular, the claim that he was "open sourcing software he developed in a personal side project" turned out to be a self-serving fabrication. From the Second Circuit's opinion:

> Aleynikov’s last day at Goldman was June 5, 2009. At approximately 5:20 p.m., just before his going-away party, Aleynikov encrypted and uploaded to a server in Germany more than 500,000 lines of source code for Goldman’s HFT system, including code for a substantial part of the infrastructure, and some of the algorithms and market data connectivity programs.

> Aleynikov also transferred some open source software licensed for use by the public that was mixed in with Goldman's proprietary code. However, a substantially greater number of the uploaded files contained proprietary code than had open source software.

It's the high-tech version of a man accused of murdering his wife giving the excuse "I swear, I thought I was shooting at a burglar that had broken into our bedroom!"

Note also that while Aleynikov's conviction was vacated by the Second Circuit, it was because of a loophole. The Second Circuit decided that stolen source code did not count as a "stolen good" under the Economic Espionage Act. (Congress corrected that loophole the same year.)


You have a citation? The only case I know about is a programmer who ftp'd a whole bunch of internal code to a site in Germany before his last day of work.



Yeah that one. Was internal code (well derived from code off stackflow) but completely harmless. The response - GS getting the FBI to go after him was wildly disproportionate.


You sound like someone who definitely has a lot of the surrounding context and some of the intimate details of this story. Can you tell us more?


Michael Lewis wrote an article about Sergey Aleynikov for Vanity Fair which led him later to write, "Flash Boys" -- it's online and worth your time [0]. Some interesting bits include:

* Advice and information that could keep a programmer out of jail.

* There's the description of Aleynikov's jailhouse enlightenment -- there is no other word for it -- e.g. what his lawyer says:

“Every time I would come to visit him in jail, I would leave energized by him,” she says. “He radiated so much energy and positive emotions that it was like therapy for me, to visit him. His eyes opened to how the world really is. And he started talking to people. For the first time! He would say: People in jail have the best stories. He could have considered himself a tragedy. And he didn’t.”

* There's the discussion of how no one involved (except Aleynikov) actually understood anything about the case and how Aleynikov's attempts to help clarify things were used against him.

* And, then, there's the piece de resistance, where Michael Lewis convenes a jury of cynical programmers -- i.e. some people who actually have a clue -- to meet Aleynikov and judge his actions (spoiler: their cynicism about the case is replaced with incredulity when they talk to him and realize he didn't care about Goldman Sach's "secret sauce" trading algorithms, etc.).

(Also, Lewis discusses his article in a Q&A in Vanity Fair [1] where he gets to talk about his own reaction to it all).

Again, well worth your time.

[0] https://www.vanityfair.com/news/2013/09/michael-lewis-goldma...

[1] https://www.vanityfair.com/news/2013/08/michael-lewis-on-gol...


I dunno about this one essay, but let's all remember that Flash Boys is a very bad book that gets almost everything wrong.

Read the extensive rebuttal for all the details:

https://www.amazon.com/Flash-Boys-Insiders-Perspective-High-...


Michael Lewis writes about him in Flash Boys, his book from a few years ago.


Anyone who thinks GS is taking any action without the intent of profiting from said action is hopelessly naive.


GS has a track record of working well with the OSS community. They open-sourced and maintained GS-collections (now Eclipse collections), a highly-performant Java collections library that they had developed internally. They benefitted from community use and contributions, but no more than any other company that open-sources an internal project.

GS benefiting from open-source doesn't imply malicious intent.


But them making a decision to benefit themselves is not necessarily mutually exclusive to a decision that benefits other developers and the OSS community. Your point is valid, but just want to clarify that it isn't necessarily all negative.


It's likely not a huge profit decision, and the path to profit is likely very indirect. They did it probably because

- it increases its brand awareness/respect for technical customers, potential job candidates, and business partners

- it can teach finance and CS students about basic algorithms used in the industry. In some industries, it is healthier for a company's technical edge to not be so far ahead of competitors.

- the open-source """community""" (whatever that is) can find bugs and extend the software for the benefit of GS. I highly doubt this is a motive, but it's a possibility/daydream.


that can be said of any corporation. or most humans


Or any life form, really.


Am I the only one who thinks they want to push the market into applying strategies that they can then trade against? This is classic Wall Street, persuading the masses to take the wrong side of bets.


That was my first thought too. And even if that’s not what GS intends to do, it will still be the end result, because their non-public automated trading will quickly become adversarial to this.


This is quite a change from when they ruined a man’s life for using Github.


It was actually subversion and that was part of the case against him, using the dictionary definition of “subversion” to prove malicious intent.


"We deliberately release some code as open source" is very different from "everything we write is open source and a random engineer is free to independently do whatever they want with it".


Funny story, my neighbor is a partner in the law firm that represented Aleynikov.

https://arstechnica.com/tech-policy/2009/07/goldmans-secret-...


Did the guy managed to get attorney fees from them?


Oh, the firm got paid, but you have to think they feel for the guy a bit.


I mean did he get it from of Goldman, not the client.


Oh no, they've been dragging the court case out for years. His employment at another firm was suspended when the story broke and he's been in and out of court and jail ever since (convicted, appeal, overturned, new charges, etc) from both NYS and Federal courts.

Last I heard, he was given time served, but the founding partner of the firm - who was his primary legal counsel in the case - was trying to get that conviction stricken from his record as well.


Context: https://blog.garrytan.com/goldman-sachs-sent-a-brilliant-com...

And this is why people are saying to exercise caution around this stuff!


This makes me have second thoughts about getting the Apple Card, not sure I want to support this company after that.


Not really. If someone try to do something what they consider stealing, they will still pursue the case with all legal might.


Hey, serious question but off topic from the thread:

You used the word "what" where we would normally use the word "that".

I've heard this done during speech numerous times from everyone ranging from fluent native speakers to learning nonnative speakers. Was there a specific reason you did it here?


Sorry. I am non-native speaker so very likely I am wrongly using 'what' in place of 'that'. After you mentioned I am feeling 'that' would be correct usage.


It's a British dialect thing, but not sure if that's the case with the poster here. See: https://english.stackexchange.com/questions/162619/usage-of-...


Emphasis on a British dialect. There are many and most (by geographically area) don't use what in that context. Except for emphasis e.g. The way Jeremy Clarkson says "an ..." where he shouldn't.


https://www.businessinsider.com/goldman-sachs-is-making-its-...

"Aug. 12, 2015, 12:33 Goldman Sachs is going the way of Google and Facebook. The investment bank is giving away some of its trading technology to clients through open-source software, according to The Wall Street Journal...."

Am I missing something?


How much do we wanna bet this is just a recruiting gimmick?


As a general rule I assume that most “open source” software released by companies is just a recruiting tool. Some projects manage to form a community beyond the mothership, so there certainly exist exceptions, but others are to be approached with extreme caution.


Oh, you mean like them offering a VP position (without telling you maybe half of the people there are VPs)?


This isn't unusual/deceptive in the industry: https://en.wikipedia.org/wiki/Vice_president

> In business, "vice president" refers to hierarchical position that ranges from extremely senior positions directly reporting to C-level executives (in non-financial companies), to junior non-management positions with four to 10 years of experience (in financial companies).

...

> In brokerage firms, investment banks and other financial companies, "vice president" is a seniority rank rather than denoting an actual managerial position within the company. It is a relatively junior position, usually does not denote managerial responsibilities and companies have a large number of vice presidents, perhaps as an inexpensive way for a company to recognize employees, or perhaps because of delayering when an employee can't be moved higher in the organization but still deserves recognition. In most cases, the title merely implies that someone is in a medium-seniority individual contributor role.

I work in the same industry, and my boss is a vice president (and had been a vice president for years despite becoming a manager only last year); his boss is a vice president; his boss is a managing director; his boss is a managing director; her boss is the CTO; his bosses own the company. I'm at roughly a level 4 in Google terms, and at my next promotion I can apply for VP and at the one after that I must apply for VP, regardless of whether I stay on the IC track or move to the manager track. So that should get you an idea of what "VP" means.

All of this was new to me when joining, though, and it would be nice to have it more publicly known.


It may be more like saying they support open source in the same way companies rushed to say they are green, for their image. The plethora of VPs is partly a sales thing, so that every client gets to meet with a VP. Just like how every sales person is a director or similarly inflated title at other companies. Hopefully one knows this before taking such a job!

Edit: After RTFA, it seems to be 2.5 things: - Crowdsourcing ideas while getting everyone to conform to their platforms - Image improvement to appear innovative - Speculating: monetization of old code that doesn't actually work anymore


just like I'm a sr engineer but still basically do a tech's work. Inflated titles at companies is a joke


A friend of mine interviewed for Goldman last year. All the senior software engineering positions were Vice Presidents.

He told me that according to his friend who works there and got him the interview, none of them are comfortable advertising that. They all choose to put "Sr. Software Engineer" on LinkedIn instead of their actual titles because nobody outside finance would understand why they're a VP.


this is just how titles in finance work: analyst < associate < VP < executive VP < managing director* < C-level executive

The lowest level managing directors typically have several additional managing directors between them and the C level executives. In other words, you can be in charge of a 15 person unit and be called a managing director or you can be in charge of an entire division of 500 people and make as much as some fortune 500 CEOs and be called a managing director. Generally, your clients are fairly sophisticated and know exactly where you are in the totem pole.

In addition to just being a rung on the ladder, the VP title often also means that you are a corporate officer, who can enter into agreements on behalf of the company.


> The lowest level managing directors typically have several additional managing directors between them and the C level executives.

Often called SMD (Senior) or EMD (Executive)


Anyone who's seen or read American Psycho should be aware of that.


That's just how title inflation works in finance.


This is an honest question, and forgive me if it's a bit naive, but if they release trading software to the masses under the assumption that the vast majority of users will make few changes, can they rely on the herd to buy enough stock in their software's predictions at a slower rate to get an uptick?


Take it with a grain of salt since I did not see the actual source and the article is paywalled. They are probably just releasing the tooling, not actual strategies.


And even if they do provide strategies, they would likely be obsolete/unprofitable ones as examples of how to use the system.


Goldman Sachs does nothing that fails to profit Goldman Sachs.

This might be just PR, in which case the code could be useful for some. However, they may be doing this is because they found something that they can exploit if other people are using this code. It might be nothing particularly bad for a given user, but if a big block of investors begin using code Goldman Sachs knows intimately, the market may suddenly start doing stuff that just happens to fall to Goldman Sach's advantage.


No company does anything which doesn’t benefit them whether via tech development, attracting programming talent, reputation etc... That is why they are companies not humans. Don’t confuse the two or pile on GS just cause they are GS.


I spoke with some developers from Goldman several years ago. They couldn't even access Github website from work. Open source software and libraries were not regarded well either. Things may be changing, who knows. My impression was that Goldman is one of the most restrictive places to work.


On Github, they only posted LICENSE, https://github.com/gg4real/gs-quant


Please pretty please open source plottool


Probably a poison pill that Goldman has engineered countertrading schemes against.

Not unlike the magic numbers the NSA suggested for various encryption schemes.


Maybe pull request trolling could be a thing? Pull requests that print “why hasn’t anyone involved in the GFC gone to jail?”


it will probably make money, but this is the same as sp500 et al being public list of components. they know that showing their "secret" will influence the irrational market to folow suit, hence showing sp500 components will make more people buy into those and make sp500 even more profitable.

this can be an attempt at the same for the new generation. if your algo is making money and it can benefit from a mass adoption of the same strategy (or more likely they have a secret tweaked one that takes that more into account) why wouldn't they offer it in the open?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: