CASTrader Blog

Machines in the Market Linkfest & Blogs

Day Trading on Steroids: He's seated amid 24 computer screens, each of which flashes hundreds of numbers. Vos points out the one that matters: a line chart that shows he's made more than $5,000 after rapidly buying and selling millions of shares of stock. ... "I saw the world of computers and trading converging," Cummings said. "So the idea was really to just create a robot that traded like a pit trader." ... Cummings doesn't think his idea is unique. Instead, he was uniquely positioned to execute it -- he could code software with a trader's mind. ...Rizzello said he thinks Tradebot is one of the Nasdaq's top 10 traders, by volume.

Alea, a new to me blog, is all over Tradebots, and it's successor, BATS.  More from Alea:

Market Effects of Autodealing
High Frequency Automated Trading
Algorithmic Arms Race
Algorithmic Arms Race [2]

Chips with everything for traders:  This particular program, he says, is actually a modified version of a previous one he wrote for the Norwegian ministry of defence that could be used to track down nuclear submarines in the murky waters of the fjords...."One has some notion of what kind of sounds a sub can make and we try to spot this sound among an ensemble of signals," he says. "This works on the same principle - it is very much like eavesdropping on one specific conversation in a noisy crowded room. But I'm not going to tell you any more than that."...Computerised trading of one kind or another now dominates London's financial markets. At least 50 per cent of the trades made in FTSE100 stocks originate from so-called "black-box" trading systems, or trading algorithms. Some investment banks put the figure nearer 60 per cent.

The smartest (or nuttiest) futurist on Earth: Kurzweil says he is applying Simons-like quantitative analysis to take advantage of market inefficiencies. And he's confident that, just as he trained computers to recognize patterns in human speech or the sound of a violin, he can do the same with currency fluctuations and stock-ownership trends. The ultimate goal is to create the first fully artificially intelligent quant fund - a black box that can learn to monitor itself and adjust. Although he started the company back in 1999, the fund has only been trading for about a year....How's he doing? Kurzweil won't say, citing SEC rules, nor will his investors. "I view Ray as one of the best pattern-recognition people in the world," says Khosla, when asked why he put money into FatKat. "I am a happy investor in Ray's company. A very happy investor."

Blogs with a quant/machine trading focus:

Tales from a Trading Desk

Dinosaur Technology and Trading

Magmasystems Blog

Coding the Markets

Chris Donnan: Programming - Brooklyn Style (you gotta love that title)

Last, but not least, another new to me blog, Zero Beta, takes a skeptical view of machines in the market:  "By handing over power and money to machines left unchecked we may be setting ourselves up for an ugly scenario.  What if they are THAT good?  What if they can learn? A mild result would be a lack of interest in the financial market system from the ordinary investor.  If expected returns go negative – much like a casino, you will see much less investment and much more hording of money.  There will always be the gamblers but those with the best machine will always win – I believe they are called the house at the Borgata."

Update:  C++ Trader links a Bloomerg article on AI and trading and has some thoughts.

Computer-guided stock trading reaches levels undreamed of in sci-fi: "In his cubicle overlooking the trading floor, Kearns, 44, consults with Lehman Brothers traders as Ph.D.s tap away at secret software. The programs they’re writing are designed to sift through billions of trades and spot subtle patterns in world markets.  Kearns, a computer scientist who has a doctorate from Harvard University, says the code is part of a dream he’s been chasing for more than two decades: to imbue computers with artificial intelligence, or AI."

Machines in the markets linkfest

Cracking the code in Automated Options Trading:  "Tim Brennan built his first automated options trading platform back in 2000, when he was a trader with Morgan Stanley. ... Flash forward seven years and Morgan Stanley is one of the largest options dealers in the country, and one of only seven primary market makers or specialists at the hugely successful ISE."

Designing a Robust Mechanical Trading Strategy: A Best Practice in Trading:  "Utilization of genetic algorithms and similar “hill climbing” techniques are a common way to ensure that your optimization does not give you a single point anomaly, but rather that there are similar input values surrounding your inputs that give similar equity graphs. Walk forward testing is another useful tool that can help you achieve realistic results and see for yourself whether a strategy would have been successful on data that was not optimized (similar to the future)."

AI and Google

AI and Stockpicking: "This is not to say that quantitative models do not work in prediction. The ones that work for me are usually characterized by these properties:

• They are based on a sound econometric or rational basis, and not on random discovery of patterns;
• They have few or even no parameters that need to be fitted to past data;
• They involve linear regression only, and not fitting to some esoteric nonlinear functions;
• They are conceptually simple."

AIStockBot (Open Source): "AIStockBot aims to become the greatest Technical and Fundamental Stock Analysis program written. Using Artificial Intelligence, it should pick stocks better than you. Includes Portfolio & Document Management Systems."

Quantlib (Open Source): "The QuantLib project is aimed at providing a comprehensive software framework for quantitative finance. QuantLib is a free/open-source library for modeling, trading, and risk management in real-life."

ProfitPY (Open Source): "ProfitPy is a set of libraries and tools for the development, testing, and execution of automated stock trading systems Specifically, ProfitPy is a collection of Python packages, modules, and scripts that work together to help you implement an automated stock trading program (or programs)."  Works with Interactive Broker.

Genius Trader (Open Source): "GeniusTrader aims to be a full featured toolbox to create trading systems."

Open Java Trading System (Open Source): "The Open Java Trading System (OJTS) is meant to be a common infrastructure to develop stock trading systems."

Merchant of Venice: "Venice is a stock market trading programme that supports portfolio management, charting, technical analysis, paper trading and experimental methods like genetic programming. Venice runs in a graphical user interface with online help and has full documentation. Venice runs on UNIX, including Mac OS X and Windows."

Genetic Programming with Syntactic Restrictions applied to Volatility Forecasting (PDF):  "This study uses Genetic Programming (GP) to discover new types of volatility forecasting models for financial time series....The in-sample optimization and out-of-sample tests are performed on 13 years of high frequency data for two foreign exchange time series. The out-of-sample forecasting performance of these new models are compared with the corresponding performance of some popular ARCH-types models, and GP consistently outperform the benchmarks. In particular, GP discovered that cross products of returns at different time horizons improve substantially the forecasting performance."

GStock

I just found out about GStock. It is billed on their website as follows: "GStock is a Virtual Supercomputer formed by many computers calculating a huge amount of investment stratagies with one goal - To pick the best profit making strategy for each stock." 

They continue: "With its humongous computing power, GStock tests over a billion investment strategies PER STOCK. Strategies are tested on years of trading data, and eventually produce the best investment strategy PER STOCK. Notice that it is not one single strategy that is applied to all stocks as done by lame investment tools and stock picking services which try to convince you as if one trading strategy for all stock is sufficient .. as if, technology companies stocks behave like oil companies stocks. Companies, markets, core investors and operations are different - using the same strategy for all stock just doesn't make sense. That is why we find the best strategy that fits like a glove for every single stock. and it works like a charm!"

An interview with the co-founder, Oren Rossen, reveals: "We have a donor who is extremely wealthy and who believes that the next big thing is supercomputing based on grid or distributed computing. He is willing to pore large sums of cash on a project that would both have a technological edge and be commercially viable. However, we don't have a clue what can be done commercially with so much computing power."

A press release says: "A recent test of GStock’s massive computing system over a two and a half year period yielded an average 5.1% return per trade over a 53-day period, with 21,000 of the 30,000 trades yielding profits - a 70% success rate."  This is astonishing if there are no caveats.

{digital}alchemy is skeptical: "I wouldn't put too much stock (pun intended) in the buy and sell signals, but GStock is an interesting idea. It's got a nice user interface so it's a useful site if only for that."  They point out that your can download past trading alerts.

I have no opinion on the merits of this concept one way or another except that it's a cool idea, and needless to say, I'd love to know what these guys are doing, exactly.

Machine learning and machines in the market linkfest

A Markov Chain Analysis of Blackjack Strategy.

Naive Bayes classifier vs. logistic regression

Machine learning videos

Portable alpha - rise of the machine:

Automated algorithm-based investment strategies have been gaining ground with institutional and high-net-worth investors over the past few months. But do they pose a real threat to traditional fund managers? Gareth Gore reports

What if you built a machine to predict hit movies?

Four years ago, when McCready was working with a similar version of the program at a firm in Barcelona, he ran thirty just-released albums, chosen at random, through his system. One stood out. The computer said that nine of the fourteen songs on the album had clear hit potential—which was unheard of. Nobody in his group knew much about the artist or had even listened to the record before, but the numbers said the album was going to be big, and McCready and his crew were of the belief that numbers do not lie. “Right around that time, a local newspaper came by and asked us what we were doing,” McCready said. “We explained the hit-prediction thing, and that we were really turned on to a record by this artist called Norah Jones.” The record was “Come Away with Me.” It went on to sell twenty million copies and win eight Grammy awards.

Soldier of fortunes:

I scarcely dare ask, but is the computer in the building? "Yes," he says. "It's a big one." He leads me through a locked door into what I can best describe as a large, noisy cupboard. Banks of superpowered servers are churning through calculations of unimaginable complexity. Schroeder and I can hardly hear each other above the whirring of the fans trying to keep them cool. On average, this machine buys and sells $10m every hour. On the floor is a row of plugs - just ordinary plugs. I keep far away from them.

The Pope issues a warning on artificial intelligence:

Pope Benedict told scientists on Saturday that by believing only in "artificial intelligence" and technology they risked the fate of the mythical Icarus, whose home-made wings melted when he flew too close to the sun.

"Contemporary life gives pride of place to an artificial intelligence ever more enslaved to experimental tecnhiques, thereby forgetting that all science should safeguard mankind and promote his tendency to authentic goodness," the Pope said.

Emergence is coupled to scope, not level

Is the Increased Speed of Direct Market Data Feeds Worth the Price?

So for a simple, albeit not light-speed fast, aggregated data feed, you now need ticker plants, feed handlers, normalizers, stream processors and time-series databases. Oy. And by the way, none of this is cheap. Including buying the data and transporting it (which I didn't even mention), you are looking at each feed costing upwards of six figures. Multiply that by a dozen and, along with the expanding number of markets, suddenly you are talking real money.

The Creativity Machine

Stephen Thaler has created a "creativity machine" that can do things like compose songs, make up new words, invent new toothbrushes, design warheads and power a robot army.  Neat, huh?  He's founded a company called Imagination Engines to monetize the technology.  The technology has been called Artificial Intelligence's "best bet" by NASA.

The technology works by using a neural network in a unique way.  Neural networks have been around a long time.  They are basically a connected set of nodes that perform a simple math calculation.  Inputs feed into each node from other nodes or the world outside the neural net, and computed outputs are produced for other nodes or as outputs to the outside world.  Each node weights the inputs it is receiving.  The network is trained by feeding a set of data with inputs and expected outputs.  The training (setting the weights and other parameters) is where the hard part is, but once done, the networks will perform some fairly amazing tasks, like face recognition.  Feed the network a previously unseen set of inputs, and it will dutifully compute an output.

Thaler has managed to take that one significant step further.  He takes a trained network that is say trained to recognize songs, then "takes it to the brink of death" by adjusting the weights and starving the network of inputs.  What the network then produces as output is a "creative" piece that is an amalgamation of what it has learned before.  It doesn't do this randomly, however.  The network knows from it's training what notes sound good together, so it tends to maintain that, even as it is recombining notes in previously unheard ways.

Thaler's company has a product that provides prediction of the next half hour of stock price movement based purely on technical indicators.  I can't tell if this software is one of his creative neural nets or just a plain neural net.  It doesn't seem like the creative aspect would be of much use in the pure prediction game.  There are other neural nets that attempt to predict stock movement, include a free online one.  In essence, any investor is a predictor, be it 30 minutes in advance or 30 years, and your bankroll growth is your score of how well you are doing.  CASTrader will be no different.  In the final analysis, systems like neural nets, CASTrader, and direct reinforcement learning, and even the human brain are prediction machines.

Thaler's company claims it's brain is "vastly more powerful" than genetic algorithms, but I'm not so sure.  Maybe the secret is in the sauce.  I'm sticking with the genetic approach for predicting the markets at least.  We'll see how it goes.  I can always throw in neural networks at a later date.