5 ms·

SEC, CFTC Report Says Algorithm Trade Set Off May 6 'Flash Crash'

17 points by jakarta 16y ago

jakarta 16y ago
PDF of the actual report: http://online.wsj.com/public/resources/documents/WSJ-flashreport10012010.pdf http://online.wsj.com/public/resources/documents/WSJ-flashre...
fr0sty 16y ago
The title of this post (and the WSJ article) is potentially misleading. The 'Algorithm Trade' referenced was a Mutual Fund company executing a very large sell order in E-mini Futures. The sell order was executed through a trading algorithm which attempted to participate in 9% of market volume. Many people associate 'Algorithmic Trading' with 'High-Frequency Trading' but there is no link made by the report. ETA: An equally correct headline could be "Large Institutional Seller set off May 6 flash crash.
- wglb 16y ago
  Now you have me confused. Algorithmic Traders are quite a different species than High-Frequency Traders. How did the article conflate these?
  - fr0sty 16y ago
    It is the popular imagination conflates them... 1. High-Frequency trading, by definition, done by algorithms. 2. We hear rumors or storyies of Algorithms gone wild. 3. This report blames and 'Algorithmic Trade' as the cause. People are tempted to 'connect-the-dots' and would likely end up with an erroneous conclusion.
    - DannoHung 16y ago
      High-Freq is usually the label used for prop trading. Algo is usually the label used for agency trading. The differences between the algos used isn't really pertinent.
      - lrm242 16y ago
        The differences are huge. The motiviations behind each are entirely separate. They are used by different people for different things. A day trader, slogging it out in front of their computer every day is very different from a floor broker trying to execute customer orders. That is the fundamental difference we're talking about here. Unfortunately, any time the markets are mentioned in connection with algorithm people instantly think "high-frequency trading", and most often they are wrong.
  - lrm242 16y ago
    In the industry algorithmic trading does not equal trading with algorithms. Algorithmic trading is typically used by sell-side firms to place large orders into the market. The algorithmic part comes from how the computer places the orders. The algorithms are used to benchmark the order placement against some known value, for example two common algorithms are Volume Weighted Average Price (VWAP) and Time Weighted Average Price (TWAP). The sell side wants to get the order into the market and by using algorithms they want to reduce the impact that order has on the price. By doing this, they can fill the client order at a better average price than if they just simply placed a large market order to buy or sell. Algorithmic trading, as described above, is not high-frequency trading. They are not remotely similar. Just because computers are used to manage the order flow into or out of the market does not mean they are related. It is important to note that the conclusions here are very close to what the gut reaction of the market was the day of and after the event. A fat fingered order. This is the algorithmic execution equivalent of a fat fingered order. The order, which was huge, hit the market too quickly and didn't manage its impact. Whether there should be rules to stop this from happening again who knows, but the theories around computers fighting it out and high-frequency firms manipulating the market have been clearly deemed false.
- beagle3 16y ago
  Said seller did 2.5% of daily average, and did it slowly. That triggers a crash all on its own. And we're supposed to believe that the SEC has any idea what's happening.
grandalf 16y ago
Consider the true implications of the idea that an algorithm "set off" the market crash: - Traders were happily trading along at a market price they all had tremendous confidence in. - An algorithm decides to liquidate some shares, causing the market price to dip a bit b/c of extra supply. - Said traders immediately lose all confidence in the market price they'd been happily trading along at all day and start dumping their own shares. Conclusions: If the traders were confident in the initial market price (i.e, if they were "right" then they should have started buying when the price started dipping. This would have held the price up. That they didn't reveals a lot about their own strategies. If the fact that the order placed by the algoritm is "to blame" implies that a human wouldn't have placed that order. It sounds to me that the algorithm simply wanted to liquidate and wouldn't have placed such a large market order if it were overly concerned about getting top dollar for every share. This happens with human traders all the time. The bigger the order the more it will move the market. There are large trade desks that help minimize this but they take a big spread for the service. So ironically it was the algorithm that had a principled, deep view of the market and the human traders who were fickle and skittish... exactly the opposite of the typical impression of each. It's like if Kasparov were playing against Deep Blue and Deep Blue sacrificed a pawn and Kasparov knocked over the board.
- DannoHung 16y ago
  This discounts the other algorithms that don't think about price confidence and are only reactionary.
  - grandalf 16y ago
    True... but as we have seen humans too can be overly reactionary.
- chipsy 16y ago
  Not necessarily. That hypothesis rests on the basis that the market is still majority human-traded. Quote the article: High frequency traders, devoid of any real demand from buyers of the futures contract, began to quickly buy and resell contracts "generating a 'hot-potato,'" the report said. So this is algorithms-against-algorithms. I would wager that in intraday trading, human minds play a relatively small role, because mechanical methods break through the emotional barriers that plague pure discretionary trading. Most good daytraders will develop a signal system to guide them, even if individual trades use manual entries and exits.

17 points by jakarta 16y ago

PDF of the actual report: http://online.wsj.com/public/resources/documents/WSJ-flashreport10012010.pdf http://online.wsj.com/public/resources/documents/WSJ-flashre...

fr0sty 16y ago

The title of this post (and the WSJ article) is potentially misleading. The 'Algorithm Trade' referenced was a Mutual Fund company executing a very large sell order in E-mini Futures. The sell order was executed through a trading algorithm which attempted to participate in 9% of market volume. Many people associate 'Algorithmic Trading' with 'High-Frequency Trading' but there is no link made by the report. ETA: An equally correct headline could be "Large Institutional Seller set off May 6 flash crash.

wglb 16y ago
Now you have me confused. Algorithmic Traders are quite a different species than High-Frequency Traders. How did the article conflate these?
- fr0sty 16y ago
  It is the popular imagination conflates them... 1. High-Frequency trading, by definition, done by algorithms. 2. We hear rumors or storyies of Algorithms gone wild. 3. This report blames and 'Algorithmic Trade' as the cause. People are tempted to 'connect-the-dots' and would likely end up with an erroneous conclusion.
  - DannoHung 16y ago
    High-Freq is usually the label used for prop trading. Algo is usually the label used for agency trading. The differences between the algos used isn't really pertinent.
    - lrm242 16y ago
      The differences are huge. The motiviations behind each are entirely separate. They are used by different people for different things. A day trader, slogging it out in front of their computer every day is very different from a floor broker trying to execute customer orders. That is the fundamental difference we're talking about here. Unfortunately, any time the markets are mentioned in connection with algorithm people instantly think "high-frequency trading", and most often they are wrong.
- lrm242 16y ago
  In the industry algorithmic trading does not equal trading with algorithms. Algorithmic trading is typically used by sell-side firms to place large orders into the market. The algorithmic part comes from how the computer places the orders. The algorithms are used to benchmark the order placement against some known value, for example two common algorithms are Volume Weighted Average Price (VWAP) and Time Weighted Average Price (TWAP). The sell side wants to get the order into the market and by using algorithms they want to reduce the impact that order has on the price. By doing this, they can fill the client order at a better average price than if they just simply placed a large market order to buy or sell. Algorithmic trading, as described above, is not high-frequency trading. They are not remotely similar. Just because computers are used to manage the order flow into or out of the market does not mean they are related. It is important to note that the conclusions here are very close to what the gut reaction of the market was the day of and after the event. A fat fingered order. This is the algorithmic execution equivalent of a fat fingered order. The order, which was huge, hit the market too quickly and didn't manage its impact. Whether there should be rules to stop this from happening again who knows, but the theories around computers fighting it out and high-frequency firms manipulating the market have been clearly deemed false.
beagle3 16y ago
Said seller did 2.5% of daily average, and did it slowly. That triggers a crash all on its own. And we're supposed to believe that the SEC has any idea what's happening.

grandalf 16y ago

Consider the true implications of the idea that an algorithm "set off" the market crash: - Traders were happily trading along at a market price they all had tremendous confidence in. - An algorithm decides to liquidate some shares, causing the market price to dip a bit b/c of extra supply. - Said traders immediately lose all confidence in the market price they'd been happily trading along at all day and start dumping their own shares. Conclusions: If the traders were confident in the initial market price (i.e, if they were "right" then they should have started buying when the price started dipping. This would have held the price up. That they didn't reveals a lot about their own strategies. If the fact that the order placed by the algoritm is "to blame" implies that a human wouldn't have placed that order. It sounds to me that the algorithm simply wanted to liquidate and wouldn't have placed such a large market order if it were overly concerned about getting top dollar for every share. This happens with human traders all the time. The bigger the order the more it will move the market. There are large trade desks that help minimize this but they take a big spread for the service. So ironically it was the algorithm that had a principled, deep view of the market and the human traders who were fickle and skittish... exactly the opposite of the typical impression of each. It's like if Kasparov were playing against Deep Blue and Deep Blue sacrificed a pawn and Kasparov knocked over the board.

DannoHung 16y ago
This discounts the other algorithms that don't think about price confidence and are only reactionary.
- grandalf 16y ago
  True... but as we have seen humans too can be overly reactionary.
chipsy 16y ago
Not necessarily. That hypothesis rests on the basis that the market is still majority human-traded. Quote the article: High frequency traders, devoid of any real demand from buyers of the futures contract, began to quickly buy and resell contracts "generating a 'hot-potato,'" the report said. So this is algorithms-against-algorithms. I would wager that in intraday trading, human minds play a relatively small role, because mechanical methods break through the emotional barriers that plague pure discretionary trading. Most good daytraders will develop a signal system to guide them, even if individual trades use manual entries and exits.