CS Senior Seminar More Thoughts #2
Overview
	Implement R.El-Yaniv's competitive analysis for online trading 
problems.  The model takes a set of input variables and output a 
recommendation to the trader as to whether he or she should seize the 
opportunity to trade or remain silent and wait for the next promising 
chance to come.  There are two directions the implementation could go in 
at this point.  First, a comparison between El-Yaniv's model and other 
models aimed at solving the same problem could be established to serve as 
a theoretical verification of El-Yaniv's model.  Second, the 
implementation could be further furnished to become an application.  In 
this case, other aspects of the problem could be factored into the model 
to reach a more realistic conclusion, such as risk handling.  
	In the first case, I'm thinking about setting up two/three traders 
with an identical portfolio before trading.  Then an opportunity for trade 
will be introduced by a change in the price of a stock.  The three traders 
will respond by applying their respective solution models to the data and 
react accordingly.  By comparing the net wealth of the modified 
portfolios, we are presented with a result that speaks to the validity of 
the solutions.  Of course, this one test is very limited in terms of its 
thoroughness in trying to prove or disprove the models.  More tests could 
be conducted on different portfolios and their statistical significance 
could be computed.  From an economic point of view, this series of tests 
provide a comparison of the two/three models that make recommendations 
based solely on asset prices to traders.  That is, trading decisions are 
made based on not only asset prices, but also external shocks, such as 
news, or a sudden change to the wealth.
	In the second case, I'll probably need a database to hold the 
historical stock prices.  After some adjustment to the model, if there's 
any to be done, the model will be run against all stocks and come up with 
a list of possible candidates for trade.
	On a technical note, the hoop jumps I can think of at this point 
include how to populate the database in the batch mode and how to 
translate the algorithms into C or Perl.

Outline
-	Briefly introduce the economic background of the trading problem.  
Cover the CS background of the problem in more detail, including its 
development, logistics, comparison with other similar models and known 
limitations.  If going for the application route, mention whether there 
are other things to be considered.
-	Implement the El-Yaniv model.
-	Based on the implementation and the direction it is going in, 
build the portfolios and run statistical checks, or build the database and 
watch the results.

Plan
-	Mid October, Learn about the algorithms
-	End October, Implement the algorithms
-	Mid November, Decided which way makes more sense depending on the 
result of the implementation and do it
-	End November, Write the paper
-	Early December, Modifications