What the interns have wrought, 2020 edition • Hacker News Board

Re: What the interns have wrought, 2020 edition

#191

Post by sweeneyrod » Tue, Aug 25, 2020, 11:42 PM UTC

nickbauman wrote: post #164

kevstev wrote: post #161

Earlier quoted context omitted.

Technically HFT is a subset of algorithmic trading. Algorithmic trading can be done on both the buy (IE hedge fund) and sell (Investment bank) side, but does not need to be high frequency. High Frequency also means low latency- sub millisecond, and I have never seen an algo trading system of any sort that deals in latencies that are not in the millisecond range, though this isn't really a requirement. Does this help?

Yes. So seller A wants to sell for $1.00 and buyer B is willing to buy at $1.06, an algorithm will calculate the presence of this potential (or actually have knowledge from other systems that this a certainty) and purchase seller A's stock with a margin of overhead of $1.03 and sell to buyer B for $1.06 netting $0.03. The algorithm is essentially a parasitic entity.

This isn't really what market makers like JS do. While they do make profit from the spread between prices they buy and sell at, the key difference is that they will buy from seller A and sell to buyer B at different times. For the duration between the two trades, they either own a positive or negative amount of whatever the good is, so they are at risk of losing money if the price changes. The spread pays them for taking on the risk and providing liquidity (allowing people to buy and sell any time even if there's no matching counterparty).

Re: What the interns have wrought, 2020 edition

#192

Post by bryan_w » Tue, Aug 25, 2020, 11:43 PM UTC

nickbauman wrote: post #164

kevstev wrote: post #161

Earlier quoted context omitted.

Technically HFT is a subset of algorithmic trading. Algorithmic trading can be done on both the buy (IE hedge fund) and sell (Investment bank) side, but does not need to be high frequency. High Frequency also means low latency- sub millisecond, and I have never seen an algo trading system of any sort that deals in latencies that are not in the millisecond range, though this isn't really a requirement. Does this help?

Yes. So seller A wants to sell for $1.00 and buyer B is willing to buy at $1.06, an algorithm will calculate the presence of this potential (or actually have knowledge from other systems that this a certainty) and purchase seller A's stock with a margin of overhead of $1.03 and sell to buyer B for $1.06 netting $0.03. The algorithm is essentially a parasitic entity.

Your example makes no sense. If both those orders were on the books then the exchange would have to immediately execute that trade at $1.00 (Assuming A is before B).

Re: What the interns have wrought, 2020 edition

#193

Post by therealdrag0 » Tue, Aug 25, 2020, 11:45 PM UTC

reaperducer wrote: post #138

gautamcgoel wrote: post #70

I know a guy who works at Jane Street; he was formerly one of the brightest undergrads at Caltech. He turned down an offer to do a PhD at Berkeley to take the Jane Street offer, IIRC. He is pretty idealistic and has an interesting philosophy about working in finance: he thinks moving money around isn't that societally valuable, but he donates 10-15% of his income to charity each year (I think he picks charities using…

he donates 10-15% of his income to charity each year For a lot of people, that's just normal life. When I was between jobs, I got a temp position installing a network and doing other IT stuff at a small midwestern Baptist church. During the transition to the new system, the secretary went on leave, so it fell to me to input the weekly donations into the accounting software. Some of the donations came in envelopes pre…

Especially at Jane Street level incomes... the marginal cost of 10% is not a lot compared to the 10% my mom gives from what she makes near minimum wage.

Re: What the interns have wrought, 2020 edition

#194

Post by sweeneyrod » Tue, Aug 25, 2020, 11:55 PM UTC

johnisgood wrote: post #7

polote wrote: post #3

Let me know if I'm wrong, but it seems like Jane Street is going contrary to the HN common belief that a lot of things should be "buy" instead of "build" and yet they are quite successful

Yep, and they use OCaml! I always feel like I have to point it out since people do not think much of it, have lots of misconceptions of it, and so on.

I think some popular misconceptions about OCaml are actually driven by its connection with Jane Street. Because they are the most prominent industrial user, people can think it is good for their niche but not useful outside it, or that it's very academic/difficult to learn/ideologically purist. But actually it's a fine general purpose programming language.

Re: What the interns have wrought, 2020 edition

#195

Post by Twixes » Wed, Aug 26, 2020, 12:17 AM UTC

Koshkin wrote: post #68

tcoff91 wrote: post #60

Earlier quoted context omitted.

You are making 3x the median family income for this country while getting to work from home and stay safe from COVID-19. The job market is imploding around us while we work cushy tech jobs. Abject misery abounds as people lose their loved ones and their livelihoods. It comes off as as quite tone-deaf when tech people complain about their $150k salaries. There are way too many entitled people in tech who have never wo…

People should not be complaining then about their $25k salaries, either. Because there are way too many people on this planet who do not have even that. (And, oh, by the way, one could be sent off to war only to die there after plenty of suffering.)

Cut the bullshit please, $150k is simply good money even in Silicon Valley and great pretty much everywhere else on Earth.

Re: What the interns have wrought, 2020 edition

#196

Post by DuskStar » Wed, Aug 26, 2020, 12:19 AM UTC

Aunche wrote: post #141

jeffbee wrote: post #90

Earlier quoted context omitted.

OK but it's still just false. Go read this obituary of Richard Gilder who recently passed away. He founded a firm that gave ordinary retail investors consistently high returns for decades. https://www.gilderlehrman.org/about/richard-gilder

I'm sure that some hedge funds know what they're doing, but it's impossible to tell which are the good ones. Investing in a hedge fund still is functionally equivalent to asking a monkey to throw darts at stocks.

Also, in general if there's a consistent and significantly market-beating (after fees) hedge fund that's available to the public, it's a scam. Most real strategies can't scale infinitely (see Medallion), and so have a maximum amount of useful capital. If the owners can't fill this amount of capital on private markets after a few years, then it probably failed a few sniff tests.

Re: What the interns have wrought, 2020 edition

#197

Post by logicslave » Wed, Aug 26, 2020, 3:00 AM UTC

maest wrote: post #182

logicslave wrote: post #140

Earlier quoted context omitted.

You can tell someone worked on the old wallstreet when they jump to explain that engineers are not a part of the core finance business. I've seen this trope a million times, and it used to be true. In 2020, the hedge funds making the most money are the ones treating their technology people like they treat their investment people. They interact, receive similar (though not 2M+) pay, and the programmer often has a part…

You're basing your assumptions on very little information. In particular, my position was with a buy-side fund set up in the last 5-10 years. It was definitely not "old wallstreet".

I worked at a well known hft prop trading firm, and then at an asset manager. The asset manager was staffed by mostly ex two sigma software engineers, a few from tower research. You basically cant even get an interview there unless you already worked with someone working there. A regular engineer there, a few years of experience made around 650k. I know the industry extremely well, and I know your type. I've watched so many buy side firms fall to the wayside because "good tech doesnt help investing, I will hire the humanities degree from princeton instead". Meanwhile, run through the top returning firms in the last decade, they all also happen to be the firms that pay their engineers the most.

Re: What the interns have wrought, 2020 edition

#198

Post by johnisgood » Wed, Aug 26, 2020, 3:23 AM UTC

non-entity wrote: post #104

nesarkvechnep wrote: post #101

Earlier quoted context omitted.

Just to make it clear, OCaml is not purely functional.

Oh sorry, not sure why I thought it was.

Yeah, it most likely is not. I mix imperative, OOP, and functional all the time. Parts of my codebase that has to do with algorithms are written using imperative style, the "public API" is OOP (class, object, methods) which makes the use of my library a breeze, and the rest is functional.

To give you an example of how you would use some crypto-related library that implemented its API (?) in OOP:

  let x = new Foo.bar in
  x#absorb a;
  x#squeeze b;
  x#reset

Or take a look at: https://github.com/xavierleroy/cryptokit/blob/master/src/cry...

This (this entire file) is a great example, too!

Re: What the interns have wrought, 2020 edition

#199

Post by johnisgood » Wed, Aug 26, 2020, 5:18 AM UTC

sweeneyrod wrote: post #194

johnisgood wrote: post #7

Earlier quoted context omitted.

Yep, and they use OCaml! I always feel like I have to point it out since people do not think much of it, have lots of misconceptions of it, and so on.

I think some popular misconceptions about OCaml are actually driven by its connection with Jane Street. Because they are the most prominent industrial user, people can think it is good for their niche but not useful outside it, or that it's very academic/difficult to learn/ideologically purist. But actually it's a fine general purpose programming language.

I am not sure why one would assume it is niche (someone said OCaml is "super-niche") just because Jane Street is a prominent user of it. In any case, yeah, it is an excellent general purpose programming language. I replaced a scripting language with it. I now use Go and OCaml instead of say, Lua, because I had to reimplement the wheel when I used Lua. Perl and other scripting languages were too slow for my typical use-cases.

Re: What the interns have wrought, 2020 edition

#200

Post by atq2119 » Wed, Aug 26, 2020, 5:24 AM UTC

t1lthesky wrote: post #184

atq2119 wrote: post #172

Earlier quoted context omitted.

HFT specifically is a form of largely zero-sums arms races. Is it really useful to society if there are entities that issue orders with microsecond (or less) response times rather than on the order of seconds or minutes? Yet a significant amount of brainpower and resources is essentially wasted on this problem. The story about lower spreads is also rather dubious. I can believe that going to sub-second HFT reduces sp…

I think at this point the benefits of HFT to markets (lower spreads, more liquidity, faster incorporation of information into prices) is pretty undisputed. I'll let you do your own research, but just to address your point about spreads: When you participate in the market, the spread is the "price" you have to pay to transact. When this goes down, it benefits all participants in the market, and especially the ones tha…

You may be confusing HFT with algorithmic trading in general.

You also disregard my whole point about just how high the HF in HFT needs to be. One can easily imagine a market that operates in rounds of blind auctions, one auction per second or one per minute or something along those lines. This would take out a lot of the arms race, and it's implausible that spreads would be much higher in such a market in a way that would hurt other investors: after all, you'd still expect competition between participants in a way that drives their profits down.

> A big part of the reason why there's been an almost universal reduction in fund management fees, saving retirement savers an enormous amount of money over the last 10 years or so is this reduction in spreads.

That makes zero sense. A significant loss due to higher spreads wouldn't show up in fund management fees, it would just show up as lower returns of the fund before management fees.

> If anything, HFT is more productive per unit of labor, relative to other sectors of society - that's why the compensation is so high!

That's at least doubtful. I would argue that compensation in HFT is high because it sits adjacent to large streams of money. In practice, a lot of compensation is ultimately about siphoning small fractions away from the streams of money you're near to, and the size of that stream makes more of a difference than almost anything else, but that's really only a form and function of power -- it doesn't correlate with how productive you are to society. (I suppose if you just define productivity as compensation per hour worked, as economists often do, then what you say is strictly speaking true, but it's also kind of circular and therefore meaningless.)