Slashdot Mirror
Mysterious Algorithm Was 4% of Trading Activity Last Week
concealment sends this excerpt from CNBC:
"A single mysterious computer program that placed orders — and then subsequently canceled them — made up 4 percent of all quote traffic in the U.S. stock market last week, according to the top tracker of high-frequency trading activity. The motive of the algorithm is still unclear. The program placed orders in 25-millisecond bursts involving about 500 stocks, according to Nanex, a market data firm. The algorithm never executed a single trade, and it abruptly ended at about 10:30 a.m. ET Friday."
Felix Salmon on high-frequency trading and its part in the current financial crisis.
Listen to this 13 min BBC programme/essay at: http://www.bbc.co.uk/programmes/b01n1thw available for the next 12 months
Artificial intelligence is the study of how to make real computers act like the ones in the movies.
It may not affect the prices directly. However it might affect the prices indirectly, by influencing the decision making of others (especially other algorithms).
So it is possible to create a large volume of "trades" without actually ever buying or selling anything? I am surprised that isn't gamed on regular basis
It is and this is the basis of high frequency trading... though on Wallstreet they call it "providing liquidity". It works like this:
Alice wants to sell 1000 shares of Acme Corp. She places an sell order for 1000 shares at $25.00 on the exchange, but she also places a minimum bid of $23.90 on the sell order. This minimum bid what Alice is willing to accept should someone counter-offer but is suppose to be secret, only the sell price will be published.
Bob is looking for 1000 shares of Acme Corp. He wants to place it in his portfolio for long-term growth, but he thinks it is currently worth less. Bob places a general buy order at $24.40 on the exchange. For the sake of simplicity we will say that is his only price, though he too could have a maximum bid he is will to pay.
So there is a sell order at $25.00 and a buy order at $24.40 pending on the exchange, nothing trades. Now Bob could make a buy offer to Alice at $24.40 and the trade would go thru, or Alice could make a sell offer to Bob at a lower price and follow thru. In a perfect world the exchange would figure it out and match the orders... but that doesn't happen without further action on the part of Alice or Bob.
Eve is a high frequency trader... Actually, Eve is a high frequency trading program at MegaTraders LLC. and has spotted that there are buy and sell orders for Acme Corp on the exchange. Eve places a bid at $24.99 for Alice's share, the exchange accepts, and then Eve immediately cancels the bid order. Eve has just learned that Alice is will to sell for less than the sell order posted. Eve then continues placing bids on Alice's stock, $24.98, $24.97, $24.96, etc., each time immediately canceling the buy when the exchange accepts the bid. Eve gets down to $23.89, at which point the exchange does not accept the bid for Alice's stock. Eve has just learned that Alice is willing to sell for as little as $23.90 and all of this has happened within 10s of milliseconds.
Remember all those articles on Slashdot about high frequency firm X laying their own fiber directly to the exchange to cut milliseconds off transit time? Having custom L2 firmware on their switches and no firewalls on their trading links to cut milliseconds off transit time? This is why they do it, so they can submit hundreds/thousands of buy/sell/cancel orders on a single stock within a fraction of a second to learn pricing differences between orders that otherwise should be secret.
So Eve now knows that Alice is will to sell for $23.90 and would perform the same procedure against Bob to discover his highest buy price. Once found Eve can now see a price difference advantages to herself. Eve buys the 1000 shares from Alice at $23.90 and then immediately sells the shares to Bob at $24.40, pocketing the $500 difference. On Wallstreet they call this "providing liquidity", anywhere else this would be considered insider trading and illegal. Multiple all this by several hundred firms with special inside access to the market place, each running their own competing Eve programs, and you quickly realize how the market can go into turmoil within seconds....
A few questions:
- Why are you allowed to cancel orders ? At an auction you owe the money once you've raised your hand.
- Why isn't there a fine on traders who happen to cancel more than X% of their orders ? X being in the order of 1.
- Why aren't transactions or even 'reservations' (which is what a canceled order looks like to me) taxed ?
Just to be clear... The above is a grossly oversimplified example of HFT. Thanks to the new world of online trading, an order isn;t really processed until both sides have confirmed the order. In the old days with a hundred guys in a trading pit, someone offering to buy at $150,000/share would obviously be wrong. Mistyping the same in an electronic order that autocompleted could have disastrous consequences, so there has to be a way to cancel a request. Why aren;t there fines or taxes? Well ask NASDAQ/etc.
In the real world the above is happening on trades with a difference of a fraction of a cent. In many cases it may be that Bob has a buy order at $25.01 and Alice has a sell order at $25.00. Eve has millisecond timing and can simply enter the orders faster than any human trader could possibly react. Eve is also taking in every market fluctuation and stock move the virtual instant it happens... meaning a well built Eve can anticipate the bounces in a stock price based on buys/sells and just announced news, before a human could recognize that news. But at the end of the trading day, Eve has no position in the market. Eve has only served to suck money out of the market by acting as an (unwanted) intermediary.
And although slight, it might slow down the system. Maybe 4% will not result in a noticable change, but could it slow the system down enough to give someone an advantage? (..) Yeah I know, that's pure speculation.
That's exactly what I read in some article might be the point:
“My guess is that the algo was testing the market, as high-frequency frequently does,” says Jon Najarian, co-founder of TradeMonster.com. “As soon as they add bandwidth, the HFT crowd sees how quickly they can top out to create latency.” (Read More: Unclear What Caused Kraft Spike: Nanex Founder.)
Translation: The ultimate goal of many of these programs is to gum up the system so it slows down the quote feed to others and allows the computer traders (with their co-located servers at the exchanges) to gain a money-making arbitrage opportunity.
"When I first heard Daydream Nation it quite frankly scared the living shit out of me." -- Matthew Stearns
If it's advantageous to sell, you sell and make money, there's your liquidity
That is most definitely not liquidity. Just because you want to sell something doesn't mean there is a buyer. Or it might mean there is a buyer but they want a big premium to do the deal. Liquidity is a measure of the ease with which buyers and sellers can find each other and agree to a price. Our recent financial crisis was in large part a crisis of liquidity. Big banks needed to be able to borrow money and everyone was afraid of lending to someone who might be insolvent so there was literally nowhere to borrow from. It's not just an all or nothing proposition either. If a stock is thinly traded, the spreads are going to be huge and it will be really expensive to buy that stock. If it is difficult to find sellers it likely will be difficult to find future buyers as well. The more trading that occurs in a stock, the narrower the bid/ask spreads because it is easier (and less expensive) for buyers and sellers to find each other. More trading = more liquitity = lower transaction costs.
I've had illegal manipulation of a stock I was shorting result in a margin call against my account. There was no opportunity to "wait a day" for the mess to subside; part of my position was closed at the manipulated price by my broker. I've also had a stop loss order triggered by someone pushing the price around with that as its intended effect, a few seconds before the close proceeding a positive earnings announcement.
The idea that an individual investor will survive a crash long enough for the volatility to end without damage is a very optimistic one. Position entry isn't the problem; yes, you use a limit orders, etc. The problems are on the exit side. Forced closing before the return to normal conditions and making safety stop loss orders impossible to use that just two of the many ways individual investors can get hammered by temporary price manipulation.