3 ms·

I do HFT and think about this a lot. I am essentially a middleman that facilitates trading across: 1. Time - I quote buy and sell prices continuously in Micros

I do HFT and think about this a lot. I am essentially a middleman that facilitates trading across: 1. Time - I quote buy and sell prices continuously in Microsoft stock trying to earn the spread between them. You want to buy stock and my offer to sell is the cheapest in the market. We trade. Later, someone else comes along to sell and I have the highest bid, so I buy the stock back from them. In between I'm exposed to price fluctuations, so sometimes I make money and sometimes I lose, but I make a profit on average by charging a spread between my buy/sell prices and predicting small price movements. The two of you could have met had you waited, but I let you lock in a sure thing by bearing risks you didn't want. Think of this like insurance. Odds are you spend more on car insurance than you're expected to receive in payments, otherwise the insurers would go out of business. But you probably prefer spending $1000 a year vs. spending $0 most years and $30000 once in your life. 2. Place - You're in the US and want to buy Nokia stock. Someone in Finland wants to sell for a cheaper price. You don't have access to the European exchanges and she doesn't have access to US exchanges, but I do. I trade against both your orders and earn a couple cents per share in profit. I helped you meet halfway across the world and also keep prices efficient to reflect global supply and demand. 3. Product - You want to invest in an index ETF that holds 40 different stocks. I quote an offer to sell based on where they're trading in the market plus a small profit margin. You buy my offer. I turn around and buy the stocks to hedge my risk. I'm better at trading than you are and have very low costs. Even considering my profit margin, it's cheaper and less risky for you to buy the ETF from me than 40 individual stocks. I keep the ETF and stock prices aligned and also helped people who wanted to sell stocks get their orders filled. Win win win. I think the work I do clearly benefits the market and natural investors. All of these trades are competitive too, so I can only earn razor thin margins per trade, or someone else will undercut me. I do sometimes question how the market picks winners and losers, typically based on speed. Is someone whose system is hundreds of nanoseconds faster more deserving of arbitrage profits? Do high technology costs naturally lead to consolidation and reduced competition? https://faculty.chicagobooth.edu/eric.budish/research/HFT-FrequentBatchAuctions.pdf https://faculty.chicagobooth.edu/eric.budish/research/HFT-Fr... this paper has some interesting thoughts on the topic. FWIW I do think continuous price-time priority trading is flawed in that it creates an arms race in speed, but it's the least bad option. Other mechanisms like batch auctions are even worse. As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF? PS: I do think it's a fun job. You get to play with cool technology and it's like a game that gets harder every day against very skilled opponents. Even the arms race isn't all bad. HFT firms have effectively bankrolled modern innovations in high speed ethernet, FPGAs, low latency kernel bypass networking, etc. that have knock-on benefits for other applications.

twic 9y ago
> As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF? My idea for this is that when you rest an order, you should be able to attach a set of hedge orders to it that will initially be inactive, but will fire when the owning order is filled. This would let everyone get hedges done at better-than-HFT speed, and make things like speed bumps or batches a lot more tolerable for market makers. You'd want some kind of pro-rata, or multiple orders at fill thresholds, so you could show big size and respond sensibly to partial fills. You might want alternative hedges to cope with price moves in the hedge target. You might want some mechanism whereby the exchange will pull the owning quote if the hedge disappears. There's all sorts of fun you could have. For this to work really well, you'd want the hedges to be processed synchronously with the fill of the owning order, so there's no HFTable window of vulnerability between fill and hedge. That couldn't be done across different matching engines in the typical architecture current exchanges use. For futures, though, it might be workable - i think all the different expiries of a given underlying, and their spreads, are already handled synchronously, so that the exchange can do implied matching. That might not include some of the more exotic constructs which aren't in the implied chain, though.
- mrchicity 9y ago
  This sounds similar to OneChronos, though I don't know if they've launched anything: https://www.onechronos.com/ https://www.onechronos.com/ You're right that there's precedent for this: -Implied engines as you mentioned where traders can atomically express a view on the term structure without legging risk. For the benefit of non-finance readers: Say you believe the fair price difference between Dec and March futures is 2 index points, you could quote in the spread between the contracts (long one, short the other) buying 1.75 selling 2.25 without being fastest to respond to every shift in the index itself. Your order in the spread will create implied orders in the underlying contracts themselves and you can only be filled if both your long and short trade. -IEX's D-Peg effectively runs a high frequency pricing model inside the matching engine to predict when an adverse price tick is likely. -I think D.E. Shaw has a patent on a domain-specific language for complex orders where they can be priced off any other instrument(s), ratios between instruments, features of the order book, etc. The real problem with batch auctions is coordination. So long as you have other exchanges trading correlated assets in continuous time, there will still be a race to adjust orders in the batch auction market after a pricing signal occurs. Even with a random delay or crossing time, it's still beneficial in statistical expectation to be faster. You'd need to trade most similar products worldwide on a single platform. Since exchanges are a network effect business, this would likely lead to monopoly profits for the exchange operator. I'm not sure "no high-speed arbitrage/monopoly exchange" is net better than "some high speed arbitrage/competing exchanges." I suppose you could have the government run the exchange as a utility, but that would stifle innovation. A lot of things we now take for granted started in experimental ATS and ECN startup markets.

twic 9y ago

> As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF? My idea for this is that when you rest an order, you should be able to attach a set of hedge orders to it that will initially be inactive, but will fire when the owning order is filled. This would let everyone get hedges done at better-than-HFT speed, and make things like speed bumps or batches a lot more tolerable for market makers. You'd want some kind of pro-rata, or multiple orders at fill thresholds, so you could show big size and respond sensibly to partial fills. You might want alternative hedges to cope with price moves in the hedge target. You might want some mechanism whereby the exchange will pull the owning quote if the hedge disappears. There's all sorts of fun you could have. For this to work really well, you'd want the hedges to be processed synchronously with the fill of the owning order, so there's no HFTable window of vulnerability between fill and hedge. That couldn't be done across different matching engines in the typical architecture current exchanges use. For futures, though, it might be workable - i think all the different expiries of a given underlying, and their spreads, are already handled synchronously, so that the exchange can do implied matching. That might not include some of the more exotic constructs which aren't in the implied chain, though.

mrchicity 9y ago
This sounds similar to OneChronos, though I don't know if they've launched anything: https://www.onechronos.com/ https://www.onechronos.com/ You're right that there's precedent for this: -Implied engines as you mentioned where traders can atomically express a view on the term structure without legging risk. For the benefit of non-finance readers: Say you believe the fair price difference between Dec and March futures is 2 index points, you could quote in the spread between the contracts (long one, short the other) buying 1.75 selling 2.25 without being fastest to respond to every shift in the index itself. Your order in the spread will create implied orders in the underlying contracts themselves and you can only be filled if both your long and short trade. -IEX's D-Peg effectively runs a high frequency pricing model inside the matching engine to predict when an adverse price tick is likely. -I think D.E. Shaw has a patent on a domain-specific language for complex orders where they can be priced off any other instrument(s), ratios between instruments, features of the order book, etc. The real problem with batch auctions is coordination. So long as you have other exchanges trading correlated assets in continuous time, there will still be a race to adjust orders in the batch auction market after a pricing signal occurs. Even with a random delay or crossing time, it's still beneficial in statistical expectation to be faster. You'd need to trade most similar products worldwide on a single platform. Since exchanges are a network effect business, this would likely lead to monopoly profits for the exchange operator. I'm not sure "no high-speed arbitrage/monopoly exchange" is net better than "some high speed arbitrage/competing exchanges." I suppose you could have the government run the exchange as a utility, but that would stifle innovation. A lot of things we now take for granted started in experimental ATS and ECN startup markets.