EOL is a principal based trading platform, meaning Enron is the buyer (seller) when there
is seller (buyer) who wants to transact on EOL. EOL provides market liquidity by
making the bid-ask spread. However making the spread is not the only revenue source for
running EOL. There is certain information asymmetry beneficial to Enron as the market
maker:
⢠Enron owns EOL trading database that contains detailed information about each
transactions; trades can be aggregated according to different categories, for
example, by commodity, by contract maturity, by counter party, by trading time
interval, just to name a few. The informational advantage will allow us to explore
market inefficiency and arbitrage across different products.
⢠The time series recorded in EOL data base contains valuable information about
supply-demand balance, market directions and volatilities, market correlations
and cross-market correlations, trading habits and patterns.
The EOL Data Mining project is aimed at taking the advantage of the information
asymmetry and market inefficiency so as to predict the market conditions. The benefit of
predictability is obvious, especially in the following aspects:
⢠Predictability means profit. The ability to predict (even in a statistical sense) will
give us an edge in trading and risk management.
⢠Predictability will enable us to control and reduce the risk of market making.
Data Ming is a new field that combines the business insights with computer learning
capabilities. The business insights are translated into certain quantitative state space
models [1] (most likely non-linear time series models). Then the best model is selected
deductively to fit the reality the most. A different approach is gaining popularity, that is
the inductive methodology [2]. The inductive method takes advantage cheap
computation power and artificial intelligence, builds the prediction model by learning the
patterns contained in the time series. In the EOL Data Mining project, we will exploit
both type approaches to build our ultimate Enron Perdition Models.
[1] Weigend, A. S. & Gershenfeld, N. A. (eds) Time Series Prediction: Forecasting the Future and Understanding
is seller (buyer) who wants to transact on EOL. EOL provides market liquidity by
making the bid-ask spread. However making the spread is not the only revenue source for
running EOL. There is certain information asymmetry beneficial to Enron as the market
maker:
⢠Enron owns EOL trading database that contains detailed information about each
transactions; trades can be aggregated according to different categories, for
example, by commodity, by contract maturity, by counter party, by trading time
interval, just to name a few. The informational advantage will allow us to explore
market inefficiency and arbitrage across different products.
⢠The time series recorded in EOL data base contains valuable information about
supply-demand balance, market directions and volatilities, market correlations
and cross-market correlations, trading habits and patterns.
The EOL Data Mining project is aimed at taking the advantage of the information
asymmetry and market inefficiency so as to predict the market conditions. The benefit of
predictability is obvious, especially in the following aspects:
⢠Predictability means profit. The ability to predict (even in a statistical sense) will
give us an edge in trading and risk management.
⢠Predictability will enable us to control and reduce the risk of market making.
Data Ming is a new field that combines the business insights with computer learning
capabilities. The business insights are translated into certain quantitative state space
models [1] (most likely non-linear time series models). Then the best model is selected
deductively to fit the reality the most. A different approach is gaining popularity, that is
the inductive methodology [2]. The inductive method takes advantage cheap
computation power and artificial intelligence, builds the prediction model by learning the
patterns contained in the time series. In the EOL Data Mining project, we will exploit
both type approaches to build our ultimate Enron Perdition Models.
[1] Weigend, A. S. & Gershenfeld, N. A. (eds) Time Series Prediction: Forecasting the Future and Understanding