4 ms·

I have no problem with HFT, but only because it is turning machines into things that are doing what humans did before, only at a speed that we can't keep up wit

I have no problem with HFT, but only because it is turning machines into things that are doing what humans did before, only at a speed that we can't keep up with as humans with our inherently physical interfaces. How is HFT different than traditional arbitrage except for the fact that it is faster and potentially at a scale that normal traders would never be able to keep up with? How is buying something one second and selling the next inherently different than buying one day and selling the next? Speed? Does computers being faster make it any more unfair than when one trader has access to millions or billions of dollars when I only have access to hundreds or thousands? I don't think so. Now, one could argue that HFT disrupts the way that Wall Street has worked for decades, but really, it is shifting the profit from traders, bankers, and fund managers to HFT firms. Sure, Wall Street probably doesn't like it, but does anyone like it when someone disrupts their market and takes their profit? You could also argue that HFT doesn't add value because HFT firms inherently don't do anything other than trade millions of stocks only for the sake of trading them, but how is this worse than day trading or running any other fund other than the turnover rate? Perhaps I'm blind to the real issues, but I just don't see a huge difference between HFT and traditional Wall St. trading, except maybe that it magnifies the problems inherent in the system due to speed and scale.

wmf 14y ago
The difference is that a crash can happen faster than any human can stop it. It's fine if you trust your high-frequency circuit breakers.
- unabridged 14y ago
  What exactly is a flash crash and why should individuals care? Isn't it just bank's computers temporarily offering stock at a severe discount and then the price going back to normal. It doesn't hurt anyone but the people with poorly programmed algorithms. No trades should ever be broken just because someone can't control their trading bots.
  - mistermann 14y ago
    Are you familiar with stop loss orders?
  - runako 14y ago
    >> poorly programmed algorithms You're describing the stop-loss orders that non-professionals use to protect themselves from losing money when the fundamentals of their companies deteriorate suddenly. A stop loss is a very simple algorithm that is widely available in retail brokerage platforms. (Example stop-loss order: "Sell when the last trade is 3% below my purchase price.") For instance, you own 100 shares of Pepsi and plan to hold for a while (years, not days). Tomorrow while you are in a meeting, a large-scale accounting fraud at Pepsi is announced, sending the stock down some large %age before the circuit breakers halt the stock. Prudently, you had a stop-loss order limiting your losses to (say) 3%. In this case, the stock probably won't recover when the stock starts trading again. If the stock had dropped and suddenly recovered due to a non-news flash crash, you would have just lost 3% of your investment for no fundamental reason. The flash crash in essence caused you sell at a localized bottom.
    - nradov 14y ago
      Except that in the real world your stop loss order for PEP probably wouldn't work under that scenario. The price is likely to instantly gap down further than -3%. Or trading may be halted immediately before your order is executed at all. When everyone rushes for the exit at once no market is going to be orderly.
      - mistermann 14y ago
        > The price is likely to instantly gap down further than -3% Right, but in a crash due to news, it is likely to go down and stay down, or at least take some time to recover so you can manually make a decision. In a flash crash it can go down 50% and recover within minutes. They are different scenarios, and they behave differently.
russtrpkovski 14y ago
If you have a 401K or own shares in a mutual share, you should care about it. The HFT algos siphon off billions every year from the big, slow "dumb money" large institutional investors (e.g. pension funds, index funds, insurance companies). For example, if an insurance company needs to liquidate its portfolio to pay a claim, the HFT will sense the insurance company placing an order in the market and try to skim 5-10 cents per share by manipulating the share price during the trade. Its a zero-sum game. The insurance company lost and the algo won. Market markets exist to provide liquidity (match buyers and sellers), not to front-run and bid up orders from customers.
- nradov 14y ago
  The big insurance companies are not that stupid any more when it comes to trading.
russtrpkovski 14y ago
If you are interested, check out the book below. Good read: http://www.amazon.com/Dark-Pools-High-Speed-Traders-Financial/dp/0307887170/ref=sr_1_1?ie=UTF8&qid=1340761584&sr=8-1&keywords=dark+pools http://www.amazon.com/Dark-Pools-High-Speed-Traders-Financia...
prodigal_erik 14y ago
It's at least conceivable that the correct valuation of a company changes from one day to the next due to human-scale events, and that speculators are providing a useful pricing signal. Changes from one millisecond to the next are amplifying pure noise produced by flaws in the mechanics of trades reaching the market and clearing.
- sseveran 14y ago
  What is the correct time interval?
iterationx 14y ago
Think about the optimization. >>According to Information Week Magazine¹: “A one (1) millisecond advantage in trading applications can be worth $100 million a year to a major brokerage firm”. Currently electronic trading makes between 60% and 70% of daily volume of NYSE¹. Tabb Group, a research firm, estimated that High-frequency traders generated about $21 billion in 2008.² From a site promoting microwave communication towers for low latency packet transfer. http://www.aviatnetworks.com/solutions/low-latency-microwave/ http://www.aviatnetworks.com/solutions/low-latency-microwave...
steve8918 14y ago
The real issue is that in order to get an edge, HFT traders are engaging in electronic warfare in order to slow down their opponents. One of the worst things they are doing is "quote stuffing", which means submitting bids and cancelling them tens of thousands of times per second. They stuff the exchanges with quotes and eventually cause a huge latency with respect to the quotes that are shown by the exchange, such as NYSE, NASDAQ, etc. However, they aren't affected because they can see the prices because they have a raw data feed. This gives them an edge relative to the other traders that don't have their own raw feed straight from the exchanges, and they use that to basically shoot fish in a barrel. This was one of the main causes for the flash crash in 2010. The excessive quotes The thing is that they could stop this by implementing a minimum time for a quote to exist, something innocuous like 100 ms, but they don't because this would be a detriment to the HFT shops. The other thing is that HFT shops behave like market makers, and they claim to provide liquidity, however, they have no obligation to provide liquidity, like real market makers. So just like during the flash crash, they turned off their machines, and the markets went haywire. So they want their cake and eat it too. If they are going to trade like market makers, they need to have the obligation to continue to provide liquidity.
- javert 14y ago
  If I recall correctly, wasn't there, like, an extremely fast recovery from the flash crash? Why does the flash crash even matter? (I'm not being snarky at all, BTW.)
  - steve8918 14y ago
    A lot of people lost a lot of money during the flash crash, through margin calls, stop loss hits, etc. It also called into question the stability of our markets. If people lose confidence in our stock markets, then people stop trading on them. If we have another flash crash, you can bet that the already-low volumes on the markets will be met with even lower volumes, which could damage our economic recovery. Most peoples' 401ks are tied to the stock market, and the last thing we need is a generation of un-retireable people because they lost a lot of their retirement on the stock market.
Dn_Ab 14y ago
The problem of HFT is two. First, HFT is very difficult. It requires the talents of extremely intelligent individuals. These individuals are a scarce resource. Now, being as they are near the generation and handling of money, HFT firms are best placed to outbid all others for these people. So R&D and other such lose out on top minds. Additionally, if much of the money of high net people is being invested in funds without finding its way into risky high tech research, this too can be a drag on the economy in the long term. http://sirc.rbi.org.in/downloads/4Cecchetti.pdf http://sirc.rbi.org.in/downloads/4Cecchetti.pdf The second is a problem of misaligned priorities. It's one thing to suck in all the best minds and it's another to have them all expending a massive amount of effort trying to shave down microseconds so their order goes in first. Chris Stucchio here http://www.chrisstucchio.com/blog/2012/hft_whats_broken.html http://www.chrisstucchio.com/blog/2012/hft_whats_broken.html argues very well that the so called sub-penny rule (abs(changeinprice) >= 0.01) is antiquated and from a time when humans reigned. This rule is a source of friction for the bots and is very likely a key player in the current misallocation of talent. It's removal would make HFTing cheap enough that not so many people would be needed. Some would leave to cure cancer or mine asteroids. The rest could then work on more sophisticated pricing techniques - which is a net boon, maybe those techniques could make prices more stable or be reused for optimal donor matching or something.
msellout 14y ago
To increase speed, you must decrease computation and memory access. The less computation and memory access possible, the smaller the variety of strategies possible. In regular timescales, the strategy space is much larger than the number of machines / agents competing. The ratio of the number of agents to the number of possible strategies approaches zero. As more machines compete in nanosecond time scales, the ratio of the number of machines to the number of strategies in this restricted space becomes much larger. As this ratio increases, the system dynamics undergo a phase shift. The probability of dramatic movements becomes much greater. When too many agents are crowded onto the same strategy, they all try to take the same side of a trade -- liquidity disappears. This was the cause of the flash crash of 2010, and will be the cause of many more.

wmf 14y ago

The difference is that a crash can happen faster than any human can stop it. It's fine if you trust your high-frequency circuit breakers.

unabridged 14y ago
What exactly is a flash crash and why should individuals care? Isn't it just bank's computers temporarily offering stock at a severe discount and then the price going back to normal. It doesn't hurt anyone but the people with poorly programmed algorithms. No trades should ever be broken just because someone can't control their trading bots.
- mistermann 14y ago
  Are you familiar with stop loss orders?
- runako 14y ago
  >> poorly programmed algorithms You're describing the stop-loss orders that non-professionals use to protect themselves from losing money when the fundamentals of their companies deteriorate suddenly. A stop loss is a very simple algorithm that is widely available in retail brokerage platforms. (Example stop-loss order: "Sell when the last trade is 3% below my purchase price.") For instance, you own 100 shares of Pepsi and plan to hold for a while (years, not days). Tomorrow while you are in a meeting, a large-scale accounting fraud at Pepsi is announced, sending the stock down some large %age before the circuit breakers halt the stock. Prudently, you had a stop-loss order limiting your losses to (say) 3%. In this case, the stock probably won't recover when the stock starts trading again. If the stock had dropped and suddenly recovered due to a non-news flash crash, you would have just lost 3% of your investment for no fundamental reason. The flash crash in essence caused you sell at a localized bottom.
  - nradov 14y ago
    Except that in the real world your stop loss order for PEP probably wouldn't work under that scenario. The price is likely to instantly gap down further than -3%. Or trading may be halted immediately before your order is executed at all. When everyone rushes for the exit at once no market is going to be orderly.
    - mistermann 14y ago
      > The price is likely to instantly gap down further than -3% Right, but in a crash due to news, it is likely to go down and stay down, or at least take some time to recover so you can manually make a decision. In a flash crash it can go down 50% and recover within minutes. They are different scenarios, and they behave differently.

russtrpkovski 14y ago

If you have a 401K or own shares in a mutual share, you should care about it. The HFT algos siphon off billions every year from the big, slow "dumb money" large institutional investors (e.g. pension funds, index funds, insurance companies). For example, if an insurance company needs to liquidate its portfolio to pay a claim, the HFT will sense the insurance company placing an order in the market and try to skim 5-10 cents per share by manipulating the share price during the trade. Its a zero-sum game. The insurance company lost and the algo won. Market markets exist to provide liquidity (match buyers and sellers), not to front-run and bid up orders from customers.

nradov 14y ago
The big insurance companies are not that stupid any more when it comes to trading.

russtrpkovski 14y ago

If you are interested, check out the book below. Good read: http://www.amazon.com/Dark-Pools-High-Speed-Traders-Financial/dp/0307887170/ref=sr_1_1?ie=UTF8&qid=1340761584&sr=8-1&keywords=dark+pools http://www.amazon.com/Dark-Pools-High-Speed-Traders-Financia...

prodigal_erik 14y ago

It's at least conceivable that the correct valuation of a company changes from one day to the next due to human-scale events, and that speculators are providing a useful pricing signal. Changes from one millisecond to the next are amplifying pure noise produced by flaws in the mechanics of trades reaching the market and clearing.

sseveran 14y ago
What is the correct time interval?

iterationx 14y ago

Think about the optimization. >>According to Information Week Magazine¹: “A one (1) millisecond advantage in trading applications can be worth $100 million a year to a major brokerage firm”. Currently electronic trading makes between 60% and 70% of daily volume of NYSE¹. Tabb Group, a research firm, estimated that High-frequency traders generated about $21 billion in 2008.² From a site promoting microwave communication towers for low latency packet transfer. http://www.aviatnetworks.com/solutions/low-latency-microwave/ http://www.aviatnetworks.com/solutions/low-latency-microwave...

steve8918 14y ago

The real issue is that in order to get an edge, HFT traders are engaging in electronic warfare in order to slow down their opponents. One of the worst things they are doing is "quote stuffing", which means submitting bids and cancelling them tens of thousands of times per second. They stuff the exchanges with quotes and eventually cause a huge latency with respect to the quotes that are shown by the exchange, such as NYSE, NASDAQ, etc. However, they aren't affected because they can see the prices because they have a raw data feed. This gives them an edge relative to the other traders that don't have their own raw feed straight from the exchanges, and they use that to basically shoot fish in a barrel. This was one of the main causes for the flash crash in 2010. The excessive quotes The thing is that they could stop this by implementing a minimum time for a quote to exist, something innocuous like 100 ms, but they don't because this would be a detriment to the HFT shops. The other thing is that HFT shops behave like market makers, and they claim to provide liquidity, however, they have no obligation to provide liquidity, like real market makers. So just like during the flash crash, they turned off their machines, and the markets went haywire. So they want their cake and eat it too. If they are going to trade like market makers, they need to have the obligation to continue to provide liquidity.

javert 14y ago
If I recall correctly, wasn't there, like, an extremely fast recovery from the flash crash? Why does the flash crash even matter? (I'm not being snarky at all, BTW.)
- steve8918 14y ago
  A lot of people lost a lot of money during the flash crash, through margin calls, stop loss hits, etc. It also called into question the stability of our markets. If people lose confidence in our stock markets, then people stop trading on them. If we have another flash crash, you can bet that the already-low volumes on the markets will be met with even lower volumes, which could damage our economic recovery. Most peoples' 401ks are tied to the stock market, and the last thing we need is a generation of un-retireable people because they lost a lot of their retirement on the stock market.

Dn_Ab 14y ago

The problem of HFT is two. First, HFT is very difficult. It requires the talents of extremely intelligent individuals. These individuals are a scarce resource. Now, being as they are near the generation and handling of money, HFT firms are best placed to outbid all others for these people. So R&D and other such lose out on top minds. Additionally, if much of the money of high net people is being invested in funds without finding its way into risky high tech research, this too can be a drag on the economy in the long term. http://sirc.rbi.org.in/downloads/4Cecchetti.pdf http://sirc.rbi.org.in/downloads/4Cecchetti.pdf The second is a problem of misaligned priorities. It's one thing to suck in all the best minds and it's another to have them all expending a massive amount of effort trying to shave down microseconds so their order goes in first. Chris Stucchio here http://www.chrisstucchio.com/blog/2012/hft_whats_broken.html http://www.chrisstucchio.com/blog/2012/hft_whats_broken.html argues very well that the so called sub-penny rule (abs(changeinprice) >= 0.01) is antiquated and from a time when humans reigned. This rule is a source of friction for the bots and is very likely a key player in the current misallocation of talent. It's removal would make HFTing cheap enough that not so many people would be needed. Some would leave to cure cancer or mine asteroids. The rest could then work on more sophisticated pricing techniques - which is a net boon, maybe those techniques could make prices more stable or be reused for optimal donor matching or something.

msellout 14y ago

To increase speed, you must decrease computation and memory access. The less computation and memory access possible, the smaller the variety of strategies possible. In regular timescales, the strategy space is much larger than the number of machines / agents competing. The ratio of the number of agents to the number of possible strategies approaches zero. As more machines compete in nanosecond time scales, the ratio of the number of machines to the number of strategies in this restricted space becomes much larger. As this ratio increases, the system dynamics undergo a phase shift. The probability of dramatic movements becomes much greater. When too many agents are crowded onto the same strategy, they all try to take the same side of a trade -- liquidity disappears. This was the cause of the flash crash of 2010, and will be the cause of many more.