nice site, intridea for a consulting/development company I really like their site’s style and structure… :)
matt's notes
indiehead/stock-tracker @ github
Game theory approach:
Operations on Time Series Data
* A typical framework is that of the FAME system, since it embodies an excellent understanding of the special properties of time series. FAME stands for forecasting, analysis and modeling environment
FAME information systems, Ann Arbor Michigan.
www.fame.com
* Data Preparation (i.e. interpolating and time scale conversion) -- curve-fitting
* Queries (e.g. moving averages and sums) -- aggregates over time.
* Forecasting (e.g. statistical or data mining-based extrapolation) -- regression, correlation, Fourier analysis, and pattern-finding.
Forecasting
# Autoregression uses a weighted sum of previous values to predict future ones. There are also seasonal autoregressive models.
# These and other models are incorporated in time series products such as FAME, SAS and SPLUS.
# In options finance, the basic approach is to assume that the price of an equity is based on a random walk (Brownian motion) around a basic slope. The magnitude of the randomness is called the volatility. In a result due to Norbert Wiener (he worked it out to shoot down bombers over London), for this model, the standard deviation of the difference between the initial price and the price at a certain time t rises as the square root of time t.
Steps in a Typical FAME Session
* Specify frequency. Say monthly, starting at January 1, 1996 and ending at the current time.
* Create sales and expenses time series by importing these from a file or typing them in. Specify that these are flow type time series.
* Create a new time series:
formula profit = sales - expenses.
* Create a fourth time series with weekly frequency on inventory. Specify that inventory is a level type time series.
* Convert the first three time series to a weekly frequency (by dividing the monthly values by 4.2 or by constructing a cubic spline to make the sales, expenses, and profits curve look smooth).
This interpolation depends on knowing that sales and expenses are flow-type values.
* Now, use autoregression to predict future time series values.
Because of what is known about stock market time series data (random walks:source)
Possibly not so random http://www.castrader.com/2007/03/why_are_informa.html
Finance Applications of Game Theory
* Higher order beliefs
http://www.gametheory.net/news/Items/110.htmlwww.fame.com
* Data Preparation (i.e. interpolating and time scale conversion) -- curve-fitting
* Queries (e.g. moving averages and sums) -- aggregates over time.
* Forecasting (e.g. statistical or data mining-based extrapolation) -- regression, correlation, Fourier analysis, and pattern-finding.
Forecasting
# Autoregression uses a weighted sum of previous values to predict future ones. There are also seasonal autoregressive models.
# These and other models are incorporated in time series products such as FAME, SAS and SPLUS.
# In options finance, the basic approach is to assume that the price of an equity is based on a random walk (Brownian motion) around a basic slope. The magnitude of the randomness is called the volatility. In a result due to Norbert Wiener (he worked it out to shoot down bombers over London), for this model, the standard deviation of the difference between the initial price and the price at a certain time t rises as the square root of time t.
Steps in a Typical FAME Session
* Specify frequency. Say monthly, starting at January 1, 1996 and ending at the current time.
* Create sales and expenses time series by importing these from a file or typing them in. Specify that these are flow type time series.
* Create a new time series:
formula profit = sales - expenses.
* Create a fourth time series with weekly frequency on inventory. Specify that inventory is a level type time series.
* Convert the first three time series to a weekly frequency (by dividing the monthly values by 4.2 or by constructing a cubic spline to make the sales, expenses, and profits curve look smooth).
This interpolation depends on knowing that sales and expenses are flow-type values.
* Now, use autoregression to predict future time series values.
Because of what is known about stock market time series data (random walks:source)
Possibly not so random http://www.castrader.com/2007/03/why_are_informa.html
Finance Applications of Game Theory
* Higher order beliefs
http://www.gametheory.net/news/Items/110.html
Information Theory combined with game theory
again…yeah………lol. this article puts it perfect
Adam speculates on three possible causes of the “markets do X because of Y†headline phenomenon: laziness, conventional wisdom, and bias.
All true, though to be fair to the journalists we like to tease so much, the root cause of all this is probably just our humanity. Hume knew as much; social scientists even have a name for the cognitive bias. And it’s not like you can sell newspapers or drive traffic by going with headlines like, “Markets sell off hard for reasons that are epistemically opaque; in entirely noncausal but correlative news, oil rises dramatically.â€
“
A fun game on a slow Friday afternoon: take the front page headline on Yahoo Finance, and perform some desirable action if the headline makes a dubious assumption about causation. Move to the next headline, and repeat. We were going to suggest this as a drinking game, but we don’t want everyone drunk before the market’s even closed.
Felix Salmon makes this point well. Financial journalists, for whatever reason, constantly report on correlated events (â€Today, p occurred after news that qâ€) with sufficient ambiguity as to leave the reader with the impression that q caused p. As every freshman philosophy student learns (or used to learn, anyway), correlation does not imply causation. Enough pontification, let’s play…
“Oil prices rose sharply Friday on news that a ship under contract to the U.S. Defense Department fired warning shots at two Iranian boats.†Really? Did the Associated Press talk to traders at the NYMEX who were all, “OMG warning shots, buy buy buy!!!†No. Isn’t it just as likely that after yesterday’s heavy selling in the energy sector, some mean reversion was due? Isn’t it also just as likely that oil rose in price for no discrete or knowable reason whatsoever? The point here is that unless you’ve got a defensible causal chain to report on, it’s sloppy to be casting prepositions all over the place that leave readers with the impression that causality has occurred.
The story goes on to undermine its own headline with an account of how non-unique these saber-rattling sorts of events are. It follows up with the concession that, “On Friday, oil prices were already up before the report on news of a pipeline attack in Nigeria and a looming refinery strike in Scotland,†and then continues anyway with commentary on one of the perennial bugbears of oil reporting, namely shenanigans in Nigeria.
Sadly, the story does not achieve oil cheerleading nirvana inasmuch as it fails to include a quote from T. Boone Pickens.
http://www.condoroptions.com/2008/04/25/the-correlation-game/http://www.lab49.com/files/demos/Lab49iPhone-CapitalScreencast.swf
Portfolio123 - Historical Screener Overview (it's web-based too)
They have a really nice suite of web-based tools for trading analysis/ranking/simulation….
Portfolio123’s Historical Screener is a one-of-a-kind screening tool that allows you, during screen development, to get immediate feedback on the past performance of the screen. Once you are satisfied with your screen, you can run it as a normal screen to get a list of the current stocks for investment ideas. Naturally past performance is no guarantee of future results, but a historically tested screen should give you the necessary confidence that your ideas are on the right track.
Portfolio123 Product Feature page
Portfolio123 uses Trading Systems to manage Model Portfolios. By following the buys and sells in these Models, an investor can effortlessly manage a large, diversified portfolio of stocks, without constant monitoring and research of his/her positions.
The Trading Systems are centered around a quantitative approach known as “Weighted Multi-Factor Ranking System” (or Ranking System) that assigns a rank to each stock in the universe. The inputs available to Ranking Systems include: fundamental ratios, technical indicators, industry/sector averages, and custom formulas. Different investment strategies can be created by varying the inputs and importance assigned to each. Examples of investment strategies include: Growth, GARP and CANSLIM.
Portfolio123 is powered by Reuters Fundamentals which is the data source most respected and widely used by professional money managers.
“
I wanted to give a brief overview of the quantitative stock models I’ve been using for some of my “active” portfolio money. The basic idea of these models is to use either a stock screen or stock ranking system along with buy filters and sell rules.
Most investors are familiar with stock screens which are available over the internet. Some of the better free ones are MSN Money Deluxe screener, Yahoo Stock Screener, and Multex Investor Netscreen. Many discount brokers also supply a stock screening capability.
Ranking systems are more sophisticated than stock screeners. In fact, a stock screen is really just a special case of a ranking system where there are only two outcomes- pass or fail. Most stock analysts use a three-way ranking system of sorts when they classify stocks as buy, sell or hold. Value Line and Zacks both sell proprietary 5-way ranking stocks for stocks, while MSN Money publishes a 10-way ranking with their Stock Scouter ratings.
I use some of the above services, and also subscribe to a service Portfolio123 which allows you to develop and test ranking systems that assign percentile ratings (0.0 to 100) to each stock in the universe, where the top rated stocks would generally have a rating of 99.5 or more.
Once you have selected your ranking or screening system, you also define buy filters and sell rules and a re-balancing frequency. The buy filters can be used to eliminate certain undesirable stocks from consideration. You can filter out penny stocks, stocks which do not trade sufficient volume, low market cap stocks etc. Filters can also be added for fundamental factors such as P/E ratio, industry group etc.
A sell rule is needed to determine when stocks will be sold. The most common sell rules are based on a drop in the ranking system below a certain level or a stock failing the stock screen, but numerous other criteria can be used- stock price weakness- either absolute or relative to the industry sector or benchmark index, stock price strength- selling when stock may be “overbought”, sell rules based on elapsed time etc.
Quantitative models can be backtested over different time periods to see how they would have performed but there are some caveats- For larger portfolios which use small cap stocks, you can get a variation on the Heisenberg Uncertainty Principle. The buying and selling of your portfolio can affect the execution prices significantly, so the backtesting results may not be realistic. You also must make sure to test the model under a variety of market conditions and time periods.
One big advantage that a small investor has over the larger institutions is the ability to use some of these active quantitative models. A large mutual fund managing over 500 million dollars is very restricted in the kinds of models they can use.
http://quantinvestor.blogspot.com/2006/07/quantitative-stock-models.html