I know what it generally is. I searched using google, but most of the articles and links really only explain what it is and explains it in a superficial way. I would like to get more insight on its implications in the market place and delve deeper into its pros and cons and its relationship to normal market participants: traders, day traders, long and short investors, speculators and the such. I do know that it adds liquidity to the market, and I suppose that’s a pro. I do know that it involves complex software and very little actual human decision making, and I suppose that’s a con. But beyond that if you could please share your insights and opinions into this HFT. The CFA curriculum (at least Level 1) doesn’t seem to delve a great deal into it… or at all, if I’m not mistaken.
This is not something the CFA curriculum deals with. I also think you are giving it more importance than what it realy is. It’s program trading on steroids, and it’s what active day traders do with and without programming assistance.
Dreary Wrote: ------------------------------------------------------- > This is not something the CFA curriculum deals > with. I also think you are giving it more > importance than what it realy is. It’s program > trading on steroids, and it’s what active day > traders do with and without programming > assistance. Maybe I’m confused here, I’m not talking about day trading. HFT deals with statistical software which makes thousands of trades every second and profits from fractions of cents. This isn’t something day traders do. Day traders use the same technical analysis, but they certainly don’t make millisecond trades, if I’m not mistaken. HFT is also mainly used by large banks’ proprietary trading desks isn’t it? It’s not something individual day traders would have access to. And furthermore, I guess the moral question to ask is if HFT is so profitable, then why would hedgefunds, mutual funds, etc. not choose this over buy n’ hold strategies? We’ve learned through rigorous studies that technical analysis shouldn’t yield any significant gains over the long run… i.e. that markets are weak form efficient. How is it that HFT is able to reap benefits from this? Or, if that question is misleading, is HFT using any technical analysis at all even? And if HFT is not beneficial over the long run, what are its cons? What are its downfalls as a tool? I always figured that Goldman Sachs’ prop trading desk involved active day traders actively making decisions on a daily basis from technicals. But if there’s anything to this HFT thing then that means there are a lot less day traders around and a lot less decision making on the part of humans; and much more automation on the part of computers. I guess what I’m saying is all the articles on google seem to be presenting HFT as if it were a free lunch… if individual day traders did get access to this software, they’d just sit back and keep the decision making to a minimum and let the software run its algorithms. This seems too good to be true.
tj2001 Wrote: ------------------------------------------------------- > Dreary Wrote: > -------------------------------------------------- > ----- > > This is not something the CFA curriculum deals > > with. I also think you are giving it more > > importance than what it realy is. It’s program > > trading on steroids, and it’s what active day > > traders do with and without programming > > assistance. > > > Maybe I’m confused here, I’m not talking about day > trading. HFT deals with statistical software > which makes thousands of trades every second and > profits from fractions of cents. This isn’t > something day traders do. Day traders use the > same technical analysis, but they certainly don’t > make millisecond trades, if I’m not mistaken. HFT > is also mainly used by large banks’ proprietary > trading desks isn’t it? It’s not something > individual day traders would have access to. > > And furthermore, I guess the moral question to ask > is if HFT is so profitable, then why would > hedgefunds, mutual funds, etc. not choose this > over buy n’ hold strategies? > > We’ve learned through rigorous studies that > technical analysis shouldn’t yield any significant > gains over the long run… i.e. that markets are > weak form efficient. How is it that HFT is able > to reap benefits from this? Or, if that question > is misleading, is HFT using any technical analysis > at all even? And if HFT is not beneficial over > the long run, what are its cons? What are its > downfalls as a tool? > > I always figured that Goldman Sachs’ prop trading > desk involved active day traders actively making > decisions on a daily basis from technicals. But > if there’s anything to this HFT thing then that > means there are a lot less day traders around and > a lot less decision making on the part of humans; > and much more automation on the part of computers. > > > I guess what I’m saying is all the articles on > google seem to be presenting HFT as if it were a > free lunch… if individual day traders did get > access to this software, they’d just sit back and > keep the decision making to a minimum and let the > software run its algorithms. This seems too > good to be true. Someone has to create the algorithms and program the software thats where the ‘human touch’ of the quants and programers come into play. Level Three very briefly touches on HFT but since the CFA is almost entirely focusted on investing based on the fundamental characteristics of a security you can see why they do not dedicate much of the curriculum to covering these types of activities.
I’m not sure where to look for HFT stuff. It’s not in the CFA curriculum, unless they added some stuff this year. CFA definitely doesn’t teach you how to do it, but they might have some stuff to make you aware that those guys are out there. You may want to check out the Wilmott guys and also set your search to algorithmic trading. With one exception, I don’t think HFT really makes a huge difference to people with low frequency trading strategies (like swing or position trading) and longer term investors. It really shouldn’t matter much if a stock is 5 cents more or less expensive because of HFT when you buy it if you are a long-term holder, and even with the HFT people around, there is still plenty of variability in prices over the day. The one exception is that HFT may make markets more vulnerable to flash crashes, which definitely affect long-term trading and investing. The jury is still out on whether last year’s flash crash is HFT caused or not. I personally think it’s most likely that it is. The danger is that two or more HFT programs may start to get caught in a positive feedback loop, and since they control an enormous proportion of the immediate liquidity, they can overwhelm the market and send it flying all over the place. Circuit breakers can help, but by the time a circuit breaker kicks in, the damage may already be done. Now HFT can be combined profitably with LFT because what the price does in the next 10 seconds or 30 minutes or even the next 500 milliseconds is probably completely uncorrelated with what it’s likely to do over the next month or year or so. So having a HFT and a LFT strategy gives you two uncorrelated portfolios, which can do wonders for your Sharpe ratio. The real trick, of course, is to find a HFT strategy which has a positive expectancy after transaction costs, and I’m not sure how they do that. Part of the challenge is that funds obviously don’t want to reveal their HFT algorithms, because that is a trade secret. Some of the algorithms go through the book of available trades by offering 100 shares of something at some price, then go up in small increments until they find the highest price that gets a response, then they dump a ton of shares at that higher price, then they go and buy 100 shares at progressively lower prices and buy a whole bunch at the lowest price. If you investigated some of these techniques, they would get very close to the borderline with market manipuation… selling 100 at a low price so you can then buy 100,000 shares at that lower price, then buying 100 shares at a high price and so you can dump 100,000 at that higher price a moment later. You may only make 0.01 cents per share trade, but if you can do it with 1,000,000 shares and do it 10 times per second, that still comes out to $1000 per second. Not bad for a day’s work. Other HFT strategies are just technical analysis techniques done at one second or 10 second intervals. I don’t know how well they work, but someone believes in them, it seems.
Math error, that’s $10 per second (because I said 1/100 of a cent, not dollar, per stock per trade). But that’s still $600/minute/stock, and the trading day is 6.5h long or 390 minutes. That would be $234,000 per stock per day. Definitely not bad for a day’s work.
Hey bchadwick, thanks for your help! My question now is, if it is so profitable on the part of the hedgefund say, then why aren’t there more hedgefunds utilizing HFT and its strategies over LFT strats or simply value investing? Furthermore, what incentive is there for hedgefunds and/or PMs to hire value investors or people who use LFT and other strategies not involving these types of complex lightning fast algorithms over HFT?
HFT strategies are almost certainly uncorrelated with LFT and traditional value investing strategies, because what happens in the next 10 seconds has almost nothing to do with the long term value of the company. Now remember your portfolio theory. If you have two assets that are uncorrelated with each other, you want BOTH in your portfolio, you get more return per level of risk by doing this. So there is still room for LFT and value investing strategies, even if HFT makes a lot of money. Secondly, what is it that is going to make HFT profitable. Really, it’s a faster algorithm, faster computers, and closer location to the exchanges. There’s only a certain amount of space available for awesome HFT firms. And the firm that is profitable this quarter, might not be profitable next quarter, because their competitor has figured out how to do exactly what they do, but 20 milliseconds faster. So as an investor in a HFT fund, you can’t really just put your money in one fund and let it grow. That firm may lose out in 6 months because a competitor is now making the trades faster, or bribed the exchange to give them a better colocation deal, or something. So HFT isn’t guaranteed… there are risks that the fund suddenly stops working. Both diversification and the additional risk that HFT will be an all-or-nothing deal means that there is room for variety in investing styles. There’s also regulatory risk. It may be that regulators suddenly decide that HFT puts ordinary traders at an unfair disadvantage and then lots of HFT firms go out of business.
Thanks again, I think it’s much more clear to me now.
milliseconds? in HFT, every microsecond counts.
bchadwick Wrote: ------------------------------------------------------- > I’m not sure where to look for HFT stuff. It’s > not in the CFA curriculum, unless they added some > stuff this year. CFA definitely doesn’t teach you > how to do it, but they might have some stuff to > make you aware that those guys are out there. You > may want to check out the Wilmott guys and also > set your search to algorithmic trading. > > With one exception, I don’t think HFT really makes > a huge difference to people with low frequency > trading strategies (like swing or position > trading) and longer term investors. It really > shouldn’t matter much if a stock is 5 cents more > or less expensive because of HFT when you buy it > if you are a long-term holder, and even with the > HFT people around, there is still plenty of > variability in prices over the day. > > The one exception is that HFT may make markets > more vulnerable to flash crashes, which definitely > affect long-term trading and investing. The jury > is still out on whether last year’s flash crash is > HFT caused or not. I personally think it’s most > likely that it is. The danger is that two or more > HFT programs may start to get caught in a positive > feedback loop, and since they control an enormous > proportion of the immediate liquidity, they can > overwhelm the market and send it flying all over > the place. Circuit breakers can help, but by the > time a circuit breaker kicks in, the damage may > already be done. > > Now HFT can be combined profitably with LFT > because what the price does in the next 10 seconds > or 30 minutes or even the next 500 milliseconds is > probably completely uncorrelated with what it’s > likely to do over the next month or year or so. > So having a HFT and a LFT strategy gives you two > uncorrelated portfolios, which can do wonders for > your Sharpe ratio. The real trick, of course, is > to find a HFT strategy which has a positive > expectancy after transaction costs, and I’m not > sure how they do that. > > Part of the challenge is that funds obviously > don’t want to reveal their HFT algorithms, because > that is a trade secret. Some of the algorithms go > through the book of available trades by offering > 100 shares of something at some price, then go up > in small increments until they find the highest > price that gets a response, then they dump a ton > of shares at that higher price, then they go and > buy 100 shares at progressively lower prices and > buy a whole bunch at the lowest price. > > If you investigated some of these techniques, they > would get very close to the borderline with market > manipuation… selling 100 at a low price so you > can then buy 100,000 shares at that lower price, > then buying 100 shares at a high price and so you > can dump 100,000 at that higher price a moment > later. You may only make 0.01 cents per share > trade, but if you can do it with 1,000,000 shares > and do it 10 times per second, that still comes > out to $1000 per second. Not bad for a day’s > work. > > Other HFT strategies are just technical analysis > techniques done at one second or 10 second > intervals. I don’t know how well they work, but > someone believes in them, it seems. In the performance evaluation and monitoring study section they mention that HFT exists and it has lowered trading costs and improved liquidity since the 1970’s but its not in a LOS as far as I know.
> > In the performance evaluation and monitoring study > section they mention that HFT exists and it has > lowered trading costs and improved liquidity since > the 1970’s but its not in a LOS as far as I know. Correct me if I’m wrong here, but I wonder what the EMH has to say about HFT. There is data to show that markets are weakform, so in theory, LFT which uses technical analysis shouldn’t be able to yield superior returns. But HFT doesn’t use technicals… on the otherhand, as bchadwick had mentioned, HFT faces significant techological and regulatory risks along with potentially high transaction costs.
A couple of things for you to research: Market neutral Multifactor models Statistical arbitrage Program trading half life Long story short, many hedge funds that undertake HFT do it as part of their trading strategy. It is not a strategy in and of itself. Case in point statistical arbitrage runs multiple factor regression on lets say stocks. Groups the universe of stocks into those with similar factors e.g. growth vs value (as basic as you can go). Looks at pricing differentials between the stocks and trades when they go out of sync e.g. two standard deviations either way by a pair of stocks in that universe and then immediately enter trades that are market / beta etc neutral to exploit the expected mean reversion. They do this process across a bunch of factors (lets say 30 for argument’s sake). This gets monitored constantly and the trading program is constantly modified and reviewed for signs of degradation aka half life. Part of the degradation is because other hedge fund shops B, C, D & E are fast on their tail and looking at the market and updating their systems with the new information and this begins to eat into their profits and eventually make the program unprofitable for the new environment. It’s quite circular and symbiotic. Continuous trading warfare if you will. To compete in this world, you need a team with hardcore science, maths & programming background. It’s institutional and designed to be able to cope with several billion AUM. While in general it works, there are periods of great market stress where factors rotate, unravel and so on that make it give back large dollops of performance - particularly from an exogenous event e.g. non-market related. A war or terrorist attack or a suggestion of either might do it. In fact a massive problem for these strategies in 08 was the ban on short selling around the world that meant you could not implement half the strategy and it just did not make sense to run the programs. Like any other investment strategy known to man, these periods of underperformance mean that it makes sense to diversify the strategy with other well known attempts to capture alpha over time. So these quant heads tend naturally to look at other arb strategies in the same space e.g. fixed income. They sure love their research and you normally find that they have a small army of pointy head guys with a ton of Phds. Shame for science & manking really, but there you go.
http://www.iijournals.com/doi/abs/10.3905/jot.2010.5.4.050 Here’s a link to an article on HFT (subscription required) and references to a bunch of other free articles elsewhere. I’m sure this will give you more than enough to get your teeth into.
HFT is kind of a BS catch-all term - there’s a few main areas here: 1) HF prop - stat arb, rebate trading, trading market microstructure effects, etc. 2) Order execution - SS algorithms 3) HF market making - options, futures, etc. Levels of sophistication can vary widely and are not directly correlated with the AUM of the strategy. Basically this can vary from using free tools at the retail level to fully customized IT infrastructures - but anyone with some knowledge and a few hundred bucks can do “HFT”. In effect, for #1, HFT allows you to reach relatively higher sharpe ratios (which, as bchad points out will be appropriately capacity constrained.). For #2/3, it allows you a way to reduce costs/increase reliability. There are probably groups out there doing/using all three.
Muddahudda Wrote: ------------------------------------------------------- > It’s institutional and designed to be able to cope with several billion AUM. I’ll admit I do not know the full intricacies of this realm - though from what I hear, capacity is almost directly tied with the sharpe of the strategy. Meaning you can have 2 HFT strategies with drastically different sharpes (i.e. .5 vs 5), and their AUM will be scaled accordingly. I guess my point is that HFT is not in and of itself designed for large AUM strategies and is accessible at all levels of skill and AUM (presumably with varying levels of success).
I also don’t know the full details, but the relationship of Sharpe ratio and capacity is more complex. At the large scale, the sharpe ratio is tied to the capacity of the strategy. As you apply more AUM, the number of juicy opportunities declines and eventually runs to zero, pushing down your Sharpe ratio (possibly even to negative values if you try too hard). It can be tricky to try to figure out how much you can do before your SR erodes, but sometimes you can try to make back of envelope estimates. At the small scale, where you assume that additional capital deployment doesn’t affect your strategy’s sharpe ratio, you want to apply more capital to the higher sharpe ratio strategy. This is because as you take on more risk, you want to get the maximum expected bang for the buck-at-risk. The only reason you’d apply stuff to the lower sharpe ratio strategy at all is to the extent that it is uncorrelated with the higher sharpe ratio strategy. What’s interesting is that the micro and the macro view lead to opposite conclusions about whether additional AUM should be applied to larger or smaller sharpe ratio strategies. Presumably, this means that every time you have more AUM, you apply the additional disproportionately to the higher Sharpe ratio strategies, but progressively less over time.
bchadwick Wrote: ------------------------------------------------------- > I also don’t know the full details, but the relationship of Sharpe ratio and capacity is more complex. Agreed. Don’t mind me, changed my post. Oops.
I wasn’t slamming you, LPoulin. I thought it was a good post… made me think a bit… and just add some detail.
bchadwick Wrote: ------------------------------------------------------- > I wasn’t slamming you, LPoulin. I thought it was a good post… made me think a bit… and just add some detail. No, no, I didn’t see it that way at all. I just started to go off on a tangent about other basics that need to be satisfied before capacity becomes an issue wrt to sharpe ratios (i.e. funding constraints, basic reasonableness test, etc.) - then kind of realized what’s the point (so I posted another post to explain why I didn’t post, ha).