Inductive reasoning in trading may be the preferred method for developing systems. One reason, as argued by Doyne Farmer in Cracking Wall Street, is due to Local Predictability. The concept is based in chaos theory, and references the idea that there may be an underlying order in the short term that enables prediction of events that are not equally predictable, if at all, in the long term.

Farmer was a founder of Prediction Company, which was purchased by UBS. There is an interesting presentation Prediction Company created a few years back, titled The Business of Model Based Trading. Both the article and the attached .pdf are worth a quick read.

In the referenced article, there are a number of analogies and layman interpretations of the interaction between physics, math, and finance. Below are two passages that I enjoyed:

He likes to use a favorite example when explaining the anatomy of a prediction. “Here, catch this!” he says, tossing you a ball. You grab it. “You know how you caught that?” he asks. “By prediction.” Farmer contends you have a model in your head of how baseballs fly. You could predict the trajectory of a high fly using Newton’s classic equation f=ma, but your brain doesn’t stock up on elementary physics equations. Rather, it builds a model directly from experiential data. A baseball player watches a thousand baseballs come off a bat, and a thousand times lifts his gloved hand, and a thousand times adjusts his guess with his mitt. Without his knowing how, his brain gradually compiles a model of where the ball lands — a model almost as good as f=ma, but not as generalized. It’s based entirely on a collection of hand-eye data from past catches.

And:

While running from lions, or investing in stocks, the tiniest edge over raw luck is significant.

Thomas Helstrom created a MATLAB program entitled Artificial Stock Trading Agent, or ASTA. Unfortunately, the code is no longer available as he is turning it into a commerical product. Still, it seems that what he had was a mechanism for both creating and testing automated strategies, encompassing one ability that many modern retail platforms don’t: testing a portfolio instead of a single issue.

The program, from what I’ve read, was a collection of about 300 .m files. The paper that served as the introduction was written by Helstrom and Kenneth Holmstrom, and titled Parameter Tuning in Trading Algorithms Using ASTA. While I don’t have a copy of it available, it is included in the 1999 publication Computational Finance.

Much of the content of that paper is also included in the ASTA User Guide. The paper, followed closely, shows the methodology for the creation of the ASTA system, and it is quite an interesting read.

While there are a lot of great views and insights in the paper, one of them was interesting from a NN/AI/data-mining algorithm development aspect:

The time series formulation based on the minimized RMSE measure is not always ideal for useful predictions of financial time series. Some reasons are:
1. The fixed prediction horizon h does not reflect the way in which financial predictions are being used. The ability of a model to predict should not be evaluated at one single fixed point in the future. A big increase in a stock value 14 days into the future is as good as the same increase 15 days into the future!
2. The equation treats all predictions, small and large, as equal. This is not always appropriate. Prediction points that would never be used for actual trading (i. e. price changes too small to be interesting) may cause higher residuals at the other points of more interest, to minimize the global RMSE.
3. A small predicted change in price, followed by a large real change in the same direction, is penalized by the RMSE measure. A trader is normally happy in this case, at least if, say, the small positive prediction was large enough to give a buy signal.
4. Several papers report a poor correlation between the RMSE measure and the profit made by applying a prediction algorithm, e. g. [Leitch and Tanner 1991] and [Bengio 1997]. A strategy that separates the modeling from the decision-taking rule, such as the one in 23.4, is less optimal than modeling the decision taking directly [Moody 1992]. Arguments 2 and 3 both give some explanations to these results.

It will be interesting to see if this comes out as a commercial product or not. According to his website, the site hasn’t been updated since 2006, so I don’t really know where it is. Any insight would be great - please comment if you know.

Advanced Trading has a couple of interesting articles regarding the demand for quants, and the related trends on Wall Street.
Demand for Quants
How Traders Achieve Alpha
I-Banks and Trading Floors

A couple of interesting quotes from these articles:
From John Comerford, EVP and global head of trading research at Instinet,

Level 2 data is about 30 gigabytes a day. We’re dealing with data that’s closer to what they deal with in the biosciences and the genomes and not what people deal with in standard relational database technology.

From Paul Alapat, managing director and head of quantitative services at Amba Research, speaking on the two types of work quants perform

Alpha-generation strategies that involve a lot of testing and validation work until they are comfortable that a certain strategy will produce returns over a certain benchmark. [The second area is managing, measuring and monitoring risk].

From Arlene Rockefeller, head of global equities at State Street Global Advisors,

If you are doing something everyone knows how to do, such as matching an index, you can’t charge extra for it. Further, it’s more difficult to maintain competitive advantage because of the fluidity of information — constant innovation is required, Rockefeller adds. “When everyone was excited about quantitative investing, everyone copied everyone else,” she says. “Now those strategies have the potential of being reclassified as beta.”

From Will Cazalet, director of long-short equity at AXA Rosenberg,

We look at what the market as a whole pays for each component — fixed assets, pension liabilities, goodwill, cash — then there is the earnings forecast model out over a number of years. We are not comparable to a lot of quant firms that use factors such as earnings-to-price, momentum, relative strength, dividend yield and historical factors. We don’t do that — we are really more fundamental.

A couple of interesting points - the article Demand for Quants discussed how banks and funds are looking for people with skills in at least two of trading, mathematics, and technology. Business acumen is also particularly useful as many coming from a pure math background, or backgrounds in other studies such as engineering and physics, lack that baseline knowledge.

This is pretty consistent with what I see, even though I am relatively disconnected from the physical street. Minimal experience is key as well. Two years, even if not particularly glamorous ones, open a lot of doors. Some employers are hungry enough that they are looking at side-work a candidate may have done. An acquaintance of mine recently was asked to show a “model” he created in his spare time when applying for a job unrelated to his current role. I have had similar experiences.

There is also a lot of talk about the demand for MBAs. This is something that the Wall Street Journal covers in-depth. Now that I am finished, we will see how the landscape looks.