What is HFT?
This is no doubt a naive question and stupid sounding question, but I'm dead serious. I've read a lot, and talked to industry people, but I still feel like I'm failing to comprehend the primary strategy in this game.
I'm not trying to break in or anything (not like I'd have a chance, but just wanted to clarify my intentions)--I'm not a STEM guy at all, and I've already made my money elsewhere, and frankly, finance/markets are boring as fuck imo. But it's always pissed me off to not know what these guys' sauce really is.
I've heard of a few different strategies being utilized. I will attempt to explain them in terms so simple that a grandmother, a 3 year-old, or even a banker could comprehend. Please tell me where I'm wrong:
STAT ARB (taking / making(?))
- Description: This probably barely even counts as HFT, if at all. You use historical information to determine the relationship between different securities. Say you look over the past couple years, and you see that there are a whole heckuva lot of instances where, when certain conditions are met or certain situations are occurring, stock A zigs when stock B zags. And then one day, you're sitting live in production, and these conditions are being met or these situations are occurring, and stock B zags, but stock A doesn't zig. You obviously then place a bet that stock A will zig. Rinse and repeat for thousands of products and correlations.
Questions:
- What cut of these guys actually make money? Most MM guys I know think stat arb guys are a joke. Like you only have to succumb to using stat arb once you've failed at MM. On the flip side, I've heard stat arb guys derisively talk jive at the MM guys, saying that stat arb is cooler because it requires ideas/intelligence/analysis/creativity, while MM is just about speed and "who has the bigger ray gun". But I dunno, though, the guy who said that to me sounded like he secretly had some major dick envy for that ray gun.
- How many people do this successfully? How many fail?
ARB (making)
Description: So simple I straight up can't believe people can do it. GOOG is a nice liquid stock that trades at multiple ECNs--let's use BATS and ARCA in this example. GOOG is 100.01/100.02 in BATS, so you blindly make a similar but slightly shittier market in ARCA. Some dumbass pours his stupid money into ARCA and fills your order. You now have a position, which you immediately close by taking the opposite side in BATS. Boom done. You made like a ten bucks or some shit. Rinse and repeat for every arb that exists.
Questions:
- Is there more to it than that? There has to be more to it than that?
- How many people do this successfully? In our "finance" "classes" at the dreadful undergraduate institution at which I "studied", we learned that it's impossible to make money this way, because markets are efficient. The crucial element that they failed to acknowledge or explain is that, taking a step back, there are people who do make money this exact way, which is what makes the markets efficient. How many of those people are there?
NO-ARB MM (making)
Description: No fucking idea. You sit there in a market, making both sides, capturing the spread. Fine. But if you can't offset to another market... then
Questions:
- What's the strategy? How do you decide where to price the market?
- What do you do with a position? Like you're making a market for some doods, and they buy 10 times in a row, and now you're sitting there with a short position, and the price is steadily creeping up in your face piece. How do you offset this? Just price your bids more aggressively and back off your offers? Even if that works, once you close that shit, you will have lost way more money due to the price change than you made capturing the spread. Wtf?
- Even worse, what do you do when you get run over? You're an MM, so you have a nice big stack on both sides, right? What happens some guy sweeps your 10 offers and now you have this massive position at shit price? Wtf?
REBATE MM (making)
Description: Not a clue. You trade to scratch and then the exchange pays you if you provide enough liquidity.
Question:
- What's the strategy? Arb or no arb?
There is absolutely no money to be made in any kind of trading, let alone quantitative.
All the banks have (had) quant / stat arb / other kind of prop desks, purely to seem prestigious to college freshmen. The returns they posted are of course due to shady accounting practices and are not real profits.
The fact that you do not know 'the secret sauce' does not mean that the traders and funds are secretive as spilling the information might ruin their pnl - this is purely because there is no sauce - all of the trading is a joke.
Markets are very static and boring, I agree.
P.S Markets ARE efficient, only fools would argue
Nice. But I think you misinterpreted what I'm asking for. I went out of my way to not say "secret sauce", because that to me means what you described: the proprietary variations or tweaks on a strategy that a desk uses to truly gain edge.
I'm just looking for a explanation of the basic strategies, conceptually.
Renaissance Technologies, Citadel, SAC, Tudor, and Millennium?
As justin88 mentioned, I think all those are more classical "quant trading", rather than HFT, with the possible exception of Citadel.
I'm wondering more about Getco, Jump, Tower, Teza, et al. I should've cleaned up my original post and not mentioned stat arb at all, but I am still curious of its use.
Markets are efficient my ass
Sounds like something I would post.
Hey man - I don't know anything about HFT, but if you want, I'll PM you a book that may answer your questions.
Option MM: Simple spread MM but you can offset your options with either other options with different strike prices or by hedging the delta & Gamma. Only thing that has to be forecasted is the volatility.
And I believe but am not sure but I believe that there are programs that trade on the latest news or social media.
Thanks for your post.
I did forget to mention option MM explicitly but I think this would fall under the category of arb, correct? Like take one option, Kospi for example, you would sit there and make a market in every single contract "blindly", pricing it solely by looking at all the other contracts' markets, and if you get filled you would try to take the prices you were using in those other markets in order to neutralize your delta+gamma. And if you're not fast enough, someone else takes those orders first, and you have to figure out how to wisely achieve those hedges, right? So it's basically the same as my GOOG example across ECNs, except you're creating a delta/gamma neutral position rather than a flat position with fungible products.
I also forgot entirely about news/media driven strategies, but I think those are very simple, right?
At least the economic release ones seem simple. Look historically at every non-farm number versus its estimates and how it affected the market. Create a model to determine the optimal bet size based on how far the number deviates from estimates. At 7:30am, use black magicke to somehow be the first fucking guy to learn what the number is, and bet the appropriate size if the market hasn't moved yet. I guess exit strategies are a little more complicated here. Any ideas?
As for the news/social media, that's a little more complicated. Certainly some natural language processing or other machine learning involved, but in the end it probably comes down to a similar model from the basic non-farm example, right? With the difficult part being translating the presser or twitter post or whatever into the "number" and also coming up with a way to quantitatively represent the current market sentiment as the "estimates".
But I feel like the news/number based strategies are not the primary profit drivers here. Like numbers come out every once in a while, and you can make some money, or not. That's no serious firm's #1 strategy. I want to know conceptually what, say, Teza, is looking at, when they're making equity markets all day.
HFT trading, is a subset of algo trading that uses speed as an advantage. I'll explain in further detail. HFT firms use technology to execute trades in microseconds, making only .0001 in profit per trade. HFT firms have servers that are co-located with the exchange servers which allow them to execute trades faster than any other firms.
So basically they can front-run your order, and they can do something called quote stuffing which creates the appearance of fake liquidity. You think there are real bids and offers but when you go to execute on them, they disappear faster than a blink of an eye.
There are other things hft's do, but mainly they use their speed advantage, to make theoretical risk less trades.
Right, I knew all that already. I'm primarily interested in your last sentence. How exactly does speed lead to those "theoretically riskless trades"? Like what are the trades, exactly? Especially as it pertains to no-arb MM and rebate MM. (Also, I will be the first to admit I'm a naive buffoon, but I'm pretty sure riskless isn't the right word. They're +EV certainly, but I don't think anyone says a single trade is theoretically risk free. Even in straight arb, you have to account for the risk of movement, right?)
Also, I forgot about that other stuff in my original post, so thank you for bringing it up. The "shady shit", I'll call it: - quote stuffing - flash orders (don't exist anymore?) - gaming the market by making trades solely to temporarily move the market
What else is there? And are these the primary strategies anywhere? I always figured they were one-off gimmicks that shops maybe used to make a lot of money in a short period of time until the market got wise, and then they moved on to another gimmick. And, correct me if I'm wrong, but I feel like Getco and co. frown upon these dicey strategies, and are more interesting in market transparency and efficiency? Or is that a load of utter shit?
You're summarizing quant trading, not HFT. Also, you don't seem to have a strong understanding of transaction costs or quantitative risk management, both of which are important parts of any type of quant trading (including HFT).
Start with the rebate strategy. That's as simple as it gets. Certain trading venues have "inverted" commission structures; i.e. the exchange pays you when you execute a liquidity-providing order, c.f. Bats:
http://cdn.batstrading.com/resources/regulation/rule_book/BATS-Exchange…
If you get in and out at the same price, you make money, namely two rebates. Sounds simple? It's not. Risk must be managed across your entire book of inventory, since you're going to employ the strategy across many instruments. For instance, you probably don't want to just provide long/short liquidity along any risk factor like Beta or a sector/industry like transports/airlines. Also, you need to calculate and balance everything as fast as possible, or you get picked off by technologically superior market participants.
Right. Have you read what I've written? Rebate MM is one of the strategies I'm completely clueless on, specifically because of the risk you mentioned. Trading at a scratch doesn't sound simple at all. For starters, can you describe the exact mechanics, from start to finish, of a simple successful scratch trade? Like a classic rebate trade strategy, where everything goes right. Is it "arbing" the same product across different ECNs but aiming for a scratch instead of a profit? If so, then I understand. If it's trading for a scratch in a single market, then I'm a little lost.
Everything goes right: 1. You post a limit bid on XYZ @ 10.00. 2. Time passes. 3. Your bid is hit, so you bot XYZ @ 10.00 collecting 1 rebate from the exchange. 4. You post a limit offer on XYZ @ 10.00. 5. Time passes. 6. Your offer is lifted, so you sold XYZ @ 10.00 collecting 1 rebate from the exchange. 7. You have no positions and +2 rebates in your account.
Ok.
And determining the alpha is the "secret sauce"? Completely proprietary and unique to firms? Is this the predictive "weighted midpoint" alluded to in the Teza-Citadel court docs?
I understand that different people will use different things to project alpha, but what are some of the common components? (Again, just to keep it simple, as it would apply to this very simple strategy of scratching for rebates.) As a complete dumbass, I can only think of a few things: book weight, order activity, trading activity, volatility...?
I'm not familiar with the Teza-Citadel docs.
The rebate strategy above has no alpha; if anything, as you pointed it out it probably has slightly negative alpha.
Ok, let's take a step back. In the example you described, where your buy at 10 gets filled, and then you place an offer at 10 and wait for it to get filled, I'm trying to determine when and where you would make that initial bet.
Like say you had no position at all and no orders. You're completely out of the market. And you want to enter and put on this rebate trade. Where do you bid and offer? Certainly you don't just sit a TOB, right? Isn't there a predictive component? Because, like you said, to simply sit at TOB to initiate this trade, would certainly have negative alpha, not fully rectified by the rebate, right?
(This is all aside from the risk stuff. I'm just talking, one market, one order, how you can justify buying at 10 and trying to sell at 10, when it's 50-50 or worse that you'll be able to sell at 10.)
HFT is what computer science geeks used to do on Wall Street (for like 4 years before all of the profits went away) to make money.
if you're still clueless i'll drop a detailed post when i have a minute (current HFTer).
dupe
I'd like to take you up on this offer, homie.
Ok. I had typed a longer version of this a while back and then i rebooted without thinking and raged hard.
STAT ARB (taking AND making), you can put quotes in and wait to get hit, or you can take prices that you like.
essentially right, but you're wrong on the HFT thing. Depending on what timeframe you want to trade on, you can do RDSA vs. RDSB and hold the position for months, or you can do an ishares vs. powershares etf of the same index and day trade it. It becomes very much part of the speed game when there is a pure arb involved.(anything with a liquid future, or a basket of exact composition. Even if you arent trading the same thing the pure arb guys are, the prices will move too quickly for you and you'll get a bad result.
What cut of these guys actually make money? Most MM guys I know think stat arb guys are a joke. Like you only have to succumb to using stat arb once you've failed at MM. On the flip side, I've heard stat arb guys derisively talk jive at the MM guys, saying that stat arb is cooler because it requires ideas/intelligence/analysis/creativity, while MM is just about speed and "who has the bigger ray gun".
Market making involves next to no risk, Stat arb does. You can back test to kingdom come, and so most that do stat arb strats live will make money, but every so often a black swan can wipe out your profits if you are exposed. They are completely different games imo. It's true you wont make crazy money from a few trades, but you can make good money if you find a strat that you can scale up in volume and across many instruments. Automation is king here.
ARB (making)
Description: So simple I straight up can't believe people can do it. GOOG is a nice liquid stock that trades at multiple ECNs--let's use BATS and ARCA in this example. GOOG is 100.01/100.02 in BATS, so you blindly make a similar but slightly shittier market in ARCA. Some dumbass pours his stupid money into ARCA and fills your order. You now have a position, which you immediately close by taking the opposite side in BATS. Boom done. You made like a ten bucks or some shit. Rinse and repeat for every arb that exists.
Pretty much there, with a few less adjectives.
Questions: - Is there more to it than that? There has to be more to it than that?
Yes, because while you do that, im doing it too, and if my pc is faster, you've bought in ARCA and i've taken the good price in BATS, so you now own google shares. Enjoy the delta gamble.
About the fastest 3 do for the most part. However in stuff thats less profitable/easy, slower guys have a chance because the bigger guys dont waste their time chasing a few hundred a day, when their IT bill is $100m. However, because of all the electronic trading now going on, the price moves and the spreads being tighter, trading costs mean you cant make enough money to cover your costs off of 1c profits, this game is a race to the bottom, do not expect to make a living off of it in the future. The risks that come with taking small money can be pretty high as well.
NO-ARB MM (making)
Description: No fucking idea. You sit there in a market, making both sides, capturing the spread. Fine. But if you can't offset to another market... then
Mean reversion. if 80% of the time what goes up comes down, you're golden. And you're the first person to think of it. :)
Questions: - What's the strategy? How do you decide where to price the market? If i told you that i'd lose my job. - What do you do with a position? Like you're making a market for some doods, and they buy 10 times in a row, and now you're sitting there with a short position, and the price is steadily creeping up in your face piece. How do you offset this? Just price your bids more aggressively and back off your offers? Even if that works, once you close that shit, you will have lost way more money due to the price change than you made capturing the spread. Wtf? Manage your risk, and dont only quote in 1 stock. if you're playing statistics, quote smaller in loads of instruments.
if you're not happy with the price you got, dont offer it.
REBATE MM (making)
Description: Not a clue. You trade to scratch and then the exchange pays you if you provide enough liquidity.
Question: - What's the strategy? Arb or no arb?
exchanges pay 0,2 bps of the value of the trade if you provided liquidity, they take 0,3 bps if you take it (numbers arbitrary). you can do all the above strategies with much lower offsets if you do them passively (provide liquidity), or just take. I've not seen a pure rebate strategy, but i have seen people move strategies or take better prices because of it being on offer.
You would probably arb or market make this.
HFT: Trading Faster Than Reality (Originally Posted: 07/06/2013)
The quick plunge appears to have been triggered or exacerbated by computer algorithms that automatically trade based on monitoring and analyzing social media streams. Indeed, a growing number of firms and startups claim they can tease emotion, meaning and context out of social media chatter and identify financially relevant information. Augify, SNTMNT, and Lucky Sort (acquired four months ago by Twitter) are a few examples. Although we don’t know how any particular product handled the fake AP tweet, the incident illustrates a potential pitfall of automated trading on news feeds in real time.
But the computer algorithms that trade on social media streams aren't hardly the nimblest of the bunch, as you all know, high-frequency trading algorithms act on data by the millisecond.
The actual focus of high-frequency trading algorithms is on rapid response to detailed events in the internal behavior of financial trading networks rather than on anything that is happening in the economy or world affairs.
The catch is that the public ticker is always just a little bit out of date. There’s an unavoidable lag, or latency. It takes time for the exchanges to communicate updated price information and continually merge and publish these revisions. The delay may be fairly short — measured in milliseconds. But any disparity in time opens the door for somebody with faster connections and computers to figure it out before it becomes public.
Trading on price disparities across markets is called arbitrage, and if the latency edge of a high-frequency trader is large enough, the trader can in some cases obtain a risk-free profit.
But, a study done by doctoral students at the University of Maryland may have identified high-frequency trading harms the market. (I take no stance on their alleged findings, I'm simply passing on the results of their work.)
This is because high-frequency trading aggressively clears trades across exchanges, whereas the market sometimes performs better by letting orders accumulate for a while to better judge which ones should trade.
In fact, the presence of high-frequency trading in their model takes profits away from regular investors and, moreover, reduces the overall profit of the system
What do you think of HFT? Do you trade on latency?
Thoughts on the study and it's results?
The market crashed when the squawk box(based in london) read the news. My firm immediately pulled our mass quotes ahead of the crash and every other market maker did too, all liquidity was gone, so any market sell order was filled much much much lower. Anyone claiming they trade off twitter feeds algorithmically is a liar or now broke, I almost guarantee this.
Liquidity is the fundamental reason we have markets. As long as HFT has no adverse effect on the amount of liquidity in the market I don't see how it "harms the market". These doctorate students obviously have their heads up their ass not to realize that in the first place.
i dont think HFT effects supply and demand
I don't know of any trading firms who use algos parse the news to trade it faster than other traders. It seems unsustainable and unnecessarily risky, when one of the main points of HFT is (low risk + low reward) times N, where N is a very large number.
Gr. 8 Kid Discusses High-Frequency Trading (Originally Posted: 01/20/2011)
Popped up on CNBC -- don't know about anyone else but in grade 8 I wasn't doing this type of research
http://www.cnbc.com/id/41177060
http://www.youtube.com/embed/9ddoha-34gg
the kid's name is william arnuk...obviously the son of sal arnuk, who along with joseph saluzzi has been lying about high-frequency trading on various tv programs for years now. it's pretty obvious from the nonsense they spout that their firm just can't keep up with the high-frequency traders and they're just trying to sway public opinion to get the government to ban their competitors. see the stuff written by manoj narang to understand why these two are liars.
I don't really know anything about HFT other than this video and something I saw on 60 minutes that had a similar take on the situation. The 60 minutes thing was just saying that if you have better computers and faster connections, and by this I mean massive super computers, then you can get your trades in milliseconds faster and thus have an unfair advantage.
Now like I said, I don't really know anything thing about this but if that is true I agree that it is wrong and should be illegal. So I'd like to get your take on it seeing as though you seem adamant against this kids positions.
so daddy just made his kids presentation?
Man, screw that. I thought this guy was some child genius.
Who were the people that he surveyed? His mom, dad, sister, and 8th grade classmates?
jews!
Annon... way for some low ranked monkey to make a comment about Saluzzi. The guy's smarter than you, and I'd wager he's right over you calling him a liar and citing a player in the HFT market. Let me refer you to Spread Networks, an HFT Pipeline with a 13.33 MS round trip data flow between New York and Chicago. Please explain to me how a retail investor can compete with that? Please explain to me how any of us can compete against the machine? From start to finish, there is no way an individual can compete with that. The costs are prohibitive and how much do you think you would need to shell out to even get into the HFT game to begin with? And really, some of their practices boarder on illegal.
How much do you know about the flash crash? The evidence is damming enough that HFTs were to blame (if you read the CFTC Report or can explain the various mini-crashes that have occurred over the last few months) and should be better regulated. Did you consider the NANEX report? What about the negatives for HFT Trading and serious potential for a forced CQS Delay? Hell, even Art Cashin has questioned HFT trading, and Cashin, a 40 year Veteran and the Director of Floor Operations for UBS in Stamford, has been in this business for far longer than guys like Narang, Amuk and Saluzzi have. Make and argument or present one instead of just saying they are wrong without even bringing up an argument and saying to us, here go look him up. His arguments aren't compelling at all, by the way.
ke18sb, it's not just faster computers and connections. It's about the software and the type of computer used. Even down to the OS on each machine. Look at RenTech, as they've been in this game for a long time. Citadel - Large firm with an HFT Fund. Even Goldman flipped out over their HFT Algos getting stolen. Each one of these firms builds their shit from the ground up and makes it their own. This is the last bastion of wild west trading, and when it's only computers trading at the extremely tight Millisecond level, if something goes wrong it goes systematic. Hell, look at the Feb 3rd Oil Shock Trade. An HFT Firm with an oil algo turned it on at 2:26:28 on Feb. 3rd, and was turned off a few seconds later after it went berserk and caused the price of oil to go up by $1/barrel during the last 4 minutes of trading. That resulted in havok for the next day or two of oil trading as everyone reacted to this.
Quia non laboriosam quis repellat earum animi laborum. Sapiente ullam alias dolor sed debitis adipisci necessitatibus. Reprehenderit explicabo et quo labore quis blanditiis qui est. Quam unde quae quo iure magni dolorum ipsa. Quo magnam quidem tenetur error autem. Accusantium cum aliquid sint qui excepturi cum est.
Modi velit sunt quia et adipisci dignissimos nulla. Excepturi quae quod laboriosam dolores quia atque accusantium.
Qui recusandae vero voluptate vel. Alias animi sit quos dicta a. Sit saepe error totam consequatur quos. Ducimus ut voluptas magni.
Doloremque ut qui a aut at quam cumque. Ullam corrupti architecto et nam et dolor doloribus repellat. Sed quisquam sed beatae ducimus quis.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Ipsa et rerum consequuntur sapiente aut sunt enim voluptate. Eveniet suscipit provident consequatur doloremque est. Aut odit aspernatur assumenda quo. Et natus voluptas iure ea minima quas. Ut enim eveniet et aut vitae laudantium sapiente.
Ullam dicta neque velit voluptas labore inventore. Recusandae sit quam et aut. Est adipisci labore modi aut rerum.
Aut ratione et placeat ad temporibus. Velit vel omnis ut pariatur soluta consequatur dolores deserunt. Libero placeat est sed saepe eaque sunt nostrum. Aspernatur odio ullam aspernatur amet reprehenderit.
Nesciunt aliquam tempore illo eos nihil exercitationem et in. Similique minus nulla est doloremque et nihil qui.