BATS runs several servers and networks that help its customers route orders with low latency and high throughput to several exchanges around the world.
To enhance throughput, and to ensure that its machines do not get bogged down by a sudden flood of orders, BATS relies on multiple servers each of which handles a subset of stock tickers. These subsets are defined alphabetically - so that it is easy to route the right order to the right server by just simple pattern matching.
When share prices are very low, servers like BATS do not like to deal with innumerable number of orders for small quantities. For instance, if someone places orders for shares of YHOO at 1 share per order, and floods the exchange with 1000's of orders to buy 1000s of shares, it can actually bring down BATS servers. Think of it as some sort of DDoS attack on BATS. To handle this, they set multipliers - saying that any orders would be multiplied by 100 or 1000 depending on the share price. That way, a single order would be for a minimum of 100 or 1000 shares, even if Quantity is set to 1.
In this case, the first server handled tickers from A through BF. And when BATS went public yesterday, BATS own ticker was handled on the first server. Being the first day of listing, even a relatively obscure stock like BATS generated a lot of orders and executions - threatening to bring down the system.
To handle this, someone tried to update the multiplier on BATS - to 1000 - on the first server - but because of fat fingers (or because of some sinister motive) hit return before the full update statement was typed in. Some thing like "UPDATE SecMaster SET Multiplier = 1000" instead of "Update SecMaster SET Multiplier = 1000 Where Code = 'BATS'".
Boom - all orders on that server got multiplied by a 1000 - so even a normally innocuous order to sell 100 shares of AAPL at market, got converted into an order to sell 100,000 shares of AAPL at market.
The thing is, systems like BATS have automated credit limit checkers - to ensure that no one trades more than what they are authorized to trade - so most of these multiplied orders got rejected by the credit check algo. However, in some cases where the customers had massive credit limits and relatively low credit exposures, the multiplied orders were approved by the credit checker - and sent on to the exchange. That's why the impact was seen only on some stocks and not on all stocks traded by that server.
The reason why BATS shares fell even though these errant orders were cancelled.
Because of the very nature of the markets, it is not possible to Cancel ALL the errant executions. Someone who ended up buying a 1000 shares of AAPL, could have sold some S&P futures or some other stocks to either hedge his position, or whatever. If the AAPL trades alone are cancelled, it exposes an innocent trader to potential loses on the other sides of the transaction. And because of the sheer volume of trades, it is literally impossible for anyone to keep track of which trades on other instruments have to be cancelled! Quite obviously, just asking is not going to help - because no one would report profitable trades as being related to their AAPL trades - and only loss making trades would get canceled (and thus impact some other innocent party on the opposite side).
To avoid these issues, exchanges cancel only the most ridiculously priced transactions - that are far away from the regular market levels. The underlying assumption is that anyone who already had orders so far out, or who managed to place orders super quick to take advantage of error orders, has to be sophisticated enough to deal with the consequences of having these orders canceled.
Why these Multipliers became important
For any sensible person, attempting to change a multiplier in the middle of the day is an extremely risky and even stupid proposition. So why on earth do some trading systems support this functionality?
This functionality was added to several trading systems as a response to the "Livedoor Shock" that hit the Tokyo Stock Exchange on 17th Jan 2006. Livedoor was a prominent Dot Com company and its share price was trading at around 250,000 JPY. Because of the high stock price, the Lot Size of this company was set to be 1. That is, you can buy or sell 1 share as a minimum order. If the stock price was 1000 for instance, the lot size would be set to either 100 or 1000. This ensures that the minimum trading unit for all stocks would be a decent amount between 100,000 JPY to 1,000,000 JPY. This is the broad idea - but because of changes in price, the minimum order size could be significantly lower or higher than those limits.
In early 2006, there was a raid on Livedoor by Japanese regulators - and this caused the stock to tank. Within a few days, the stock hit about 200 JPY (a fall of over 99%). At that time, the TSE did not have the ability to increase the multiplier during the middle of the day. So the Lot Size still remained as 1, because of which the minimum size of the orders on LiveDoor was just 200 JPY ( less than $2).
Because of this low order size, there was a literal flood of orders to buy Livedoor by speculators, and this was matched by a flood of orders to Short sell Livedoor (because the sellers expected that the stock was anyway going to zero). The number of transactions increased dramatically - and the TSE was actually forced to shut trading about 2 hours before scheduled close of the day - because they hit the max transaction volume for the day!
After this event, different systems have been put in place by different exchanges and trading systems. Some people decided to dramatically increase the capacity - supporting hundreds of millions of transactions per day. But some people realized that irrespective of such measures, there could be situations where whatever maximums were set, just might not be enough.
It is for this reason that some people implemented support for middle of day Multipliers. In some exchanges, this is implemented as a brief shutdown for a few minutes, when the multiplier or Lot Size is changed, and then the system resumes. In some cases, to avoid even brief shutdowns, another ticker is introduced which is multiplied, and the first ticker is gradually turned off - so that all orders on the first ticker get rejected. And in some high frequency systems like BATS, the server sends out messages to all connected clients regarding the switch and switches the multiplier on the fly.
When there is a single multiplier being changed, the system is capable of handling things - but when thousands of multipliers got changed, the on-the-fly system broke down completely. This is what resulted in the mess on Friday.
The question still remains whether this was a genuine human error, or whether there was a sinister motive involved. Especially considering that this was the first day of BATS listing, someone who did not get a big payoff from the IPO could have used this method to sabotage the company and the IPO.
</conspiracy-theory>
The above post is just hyperactive imagination - The Livedoor incident on the TSE did happen - but I am not aware of any trading systems that change multipliers on the fly! But I couldn't resist a good old conspiracy theory with all loose ends sewed up!
Again... RETURNING TO a small trade transaction tax would raise revenue, put a big crimp in these non-productive financial patterns, and affect only the 1%.
Push an idea that can actually happen and makes sense.
If the government starts taxing stock trades, watch out. While 1/4% may not sound like much, once the feds get a taste of mo money, their appetite will never be satisfied. Remember the itty bitty start to sales tax? And what is it now, 8% in lots of places. I know that's a state thing but the federal government is every bit as greedy. How would you like to be paying 5% tax on your trades? It'll happen once they open this box.
If the government starts taxing stock trades, watch out. While 1/4% may not sound like much, once the feds get a taste of mo money, their appetite will never be satisfied. Remember the itty bitty start to sales tax? And what is it now, 8% in lots of places. I know that's a state thing but the federal government is every bit as greedy. How would you like to be paying 5% tax on your trades? It'll happen once they open this box.
Like anything, there are positives and negatives which need to be balanced.
The market would be more stable without all the activity from day traders and all the manipulation by large trading firms (*cough*goldman-sachs*cough*). Adding a small fee would reduce volatility and reduce the risk of manipulation, ensuring a better chance that the shares would actually trade at what people are willing to pay for them. I don't think it even needs to be 1/4% - even a smaller amount would do.
Do the positives outweigh the negatives? I don't know - I'm not an economist. But it certainly doesn't bother me for it to be considered.
Because of course, programmers are automagically protected against errors and mistakes. No buggy software on Macs. Safari crashing this morning was an illusion.
Comments
To enhance throughput, and to ensure that its machines do not get bogged down by a sudden flood of orders, BATS relies on multiple servers each of which handles a subset of stock tickers. These subsets are defined alphabetically - so that it is easy to route the right order to the right server by just simple pattern matching.
When share prices are very low, servers like BATS do not like to deal with innumerable number of orders for small quantities. For instance, if someone places orders for shares of YHOO at 1 share per order, and floods the exchange with 1000's of orders to buy 1000s of shares, it can actually bring down BATS servers. Think of it as some sort of DDoS attack on BATS. To handle this, they set multipliers - saying that any orders would be multiplied by 100 or 1000 depending on the share price. That way, a single order would be for a minimum of 100 or 1000 shares, even if Quantity is set to 1.
In this case, the first server handled tickers from A through BF. And when BATS went public yesterday, BATS own ticker was handled on the first server. Being the first day of listing, even a relatively obscure stock like BATS generated a lot of orders and executions - threatening to bring down the system.
To handle this, someone tried to update the multiplier on BATS - to 1000 - on the first server - but because of fat fingers (or because of some sinister motive) hit return before the full update statement was typed in. Some thing like "UPDATE SecMaster SET Multiplier = 1000" instead of "Update SecMaster SET Multiplier = 1000 Where Code = 'BATS'".
Boom - all orders on that server got multiplied by a 1000 - so even a normally innocuous order to sell 100 shares of AAPL at market, got converted into an order to sell 100,000 shares of AAPL at market.
The thing is, systems like BATS have automated credit limit checkers - to ensure that no one trades more than what they are authorized to trade - so most of these multiplied orders got rejected by the credit check algo. However, in some cases where the customers had massive credit limits and relatively low credit exposures, the multiplied orders were approved by the credit checker - and sent on to the exchange. That's why the impact was seen only on some stocks and not on all stocks traded by that server.
The reason why BATS shares fell even though these errant orders were cancelled.
Because of the very nature of the markets, it is not possible to Cancel ALL the errant executions. Someone who ended up buying a 1000 shares of AAPL, could have sold some S&P futures or some other stocks to either hedge his position, or whatever. If the AAPL trades alone are cancelled, it exposes an innocent trader to potential loses on the other sides of the transaction. And because of the sheer volume of trades, it is literally impossible for anyone to keep track of which trades on other instruments have to be cancelled! Quite obviously, just asking is not going to help - because no one would report profitable trades as being related to their AAPL trades - and only loss making trades would get canceled (and thus impact some other innocent party on the opposite side).
To avoid these issues, exchanges cancel only the most ridiculously priced transactions - that are far away from the regular market levels. The underlying assumption is that anyone who already had orders so far out, or who managed to place orders super quick to take advantage of error orders, has to be sophisticated enough to deal with the consequences of having these orders canceled.
Why these Multipliers became important
For any sensible person, attempting to change a multiplier in the middle of the day is an extremely risky and even stupid proposition. So why on earth do some trading systems support this functionality?
This functionality was added to several trading systems as a response to the "Livedoor Shock" that hit the Tokyo Stock Exchange on 17th Jan 2006. Livedoor was a prominent Dot Com company and its share price was trading at around 250,000 JPY. Because of the high stock price, the Lot Size of this company was set to be 1. That is, you can buy or sell 1 share as a minimum order. If the stock price was 1000 for instance, the lot size would be set to either 100 or 1000. This ensures that the minimum trading unit for all stocks would be a decent amount between 100,000 JPY to 1,000,000 JPY. This is the broad idea - but because of changes in price, the minimum order size could be significantly lower or higher than those limits.
In early 2006, there was a raid on Livedoor by Japanese regulators - and this caused the stock to tank. Within a few days, the stock hit about 200 JPY (a fall of over 99%). At that time, the TSE did not have the ability to increase the multiplier during the middle of the day. So the Lot Size still remained as 1, because of which the minimum size of the orders on LiveDoor was just 200 JPY ( less than $2).
Because of this low order size, there was a literal flood of orders to buy Livedoor by speculators, and this was matched by a flood of orders to Short sell Livedoor (because the sellers expected that the stock was anyway going to zero). The number of transactions increased dramatically - and the TSE was actually forced to shut trading about 2 hours before scheduled close of the day - because they hit the max transaction volume for the day!
After this event, different systems have been put in place by different exchanges and trading systems. Some people decided to dramatically increase the capacity - supporting hundreds of millions of transactions per day. But some people realized that irrespective of such measures, there could be situations where whatever maximums were set, just might not be enough.
It is for this reason that some people implemented support for middle of day Multipliers. In some exchanges, this is implemented as a brief shutdown for a few minutes, when the multiplier or Lot Size is changed, and then the system resumes. In some cases, to avoid even brief shutdowns, another ticker is introduced which is multiplied, and the first ticker is gradually turned off - so that all orders on the first ticker get rejected. And in some high frequency systems like BATS, the server sends out messages to all connected clients regarding the switch and switches the multiplier on the fly.
When there is a single multiplier being changed, the system is capable of handling things - but when thousands of multipliers got changed, the on-the-fly system broke down completely. This is what resulted in the mess on Friday.
The question still remains whether this was a genuine human error, or whether there was a sinister motive involved. Especially considering that this was the first day of BATS listing, someone who did not get a big payoff from the IPO could have used this method to sabotage the company and the IPO.
</conspiracy-theory>
The above post is just hyperactive imagination - The Livedoor incident on the TSE did happen - but I am not aware of any trading systems that change multipliers on the fly! But I couldn't resist a good old conspiracy theory with all loose ends sewed up!
"The End-Time is Here - 2008 was God's last warning, 2012 is economic collapse & WW III"
Rather ironic Google Ad to place on this thread.
I hope the world doesn't end before the Olympics and the iPhone 5
The Olympics, I don't really care about, though I admire the men and women who have dedicated their lives to sporting excellent.
But iPhone 5... Well, maybe iPad 2013, I'll try to stick around until then.
And I got to finish Mass Effect 3, though apparently they are making a new ending as we speak. So, no rush.
Again... RETURNING TO a small trade transaction tax would raise revenue, put a big crimp in these non-productive financial patterns, and affect only the 1%.
Push an idea that can actually happen and makes sense.
If the government starts taxing stock trades, watch out. While 1/4% may not sound like much, once the feds get a taste of mo money, their appetite will never be satisfied. Remember the itty bitty start to sales tax? And what is it now, 8% in lots of places. I know that's a state thing but the federal government is every bit as greedy. How would you like to be paying 5% tax on your trades? It'll happen once they open this box.
If the government starts taxing stock trades, watch out. While 1/4% may not sound like much, once the feds get a taste of mo money, their appetite will never be satisfied. Remember the itty bitty start to sales tax? And what is it now, 8% in lots of places. I know that's a state thing but the federal government is every bit as greedy. How would you like to be paying 5% tax on your trades? It'll happen once they open this box.
Like anything, there are positives and negatives which need to be balanced.
The market would be more stable without all the activity from day traders and all the manipulation by large trading firms (*cough*goldman-sachs*cough*). Adding a small fee would reduce volatility and reduce the risk of manipulation, ensuring a better chance that the shares would actually trade at what people are willing to pay for them. I don't think it even needs to be 1/4% - even a smaller amount would do.
Do the positives outweigh the negatives? I don't know - I'm not an economist. But it certainly doesn't bother me for it to be considered.
This would never happen under OS X.
Because of course, programmers are automagically protected against errors and mistakes. No buggy software on Macs. Safari crashing this morning was an illusion.
Ahemf.