Jeph Heavy Industries

Symbol	Agent	Bought	Per Share	Shares	Last Price	Sell Price	Value
GRRF	Agent 16	10/28/2009	$4.60	217	$3.48	$4.88	$756.51
GRMN	Agent 15	10/28/2009	$35.23	28	$35.38	$37.23	$1,004.26
GRIF	Agent 12	10/22/2009	$30.07	33	$29.46	$31.62	$979.71
CONN	Agent 10	10/20/2009	$9.84	102	$6.26	$11.95	$636.44
GRMH	Agent 1	10/16/2009	$2.15	465	$1.20	$2.89	$558.14
BUSE	Agent 3	10/16/2009	$4.69	213	$4.00	$6.61	$852.88
GNCMA	Agent 2	10/16/2009	$6.47	155	$5.95	$6.91	$919.63
Invested: $7,000.00 Value: $5,707.56

Agent	Stake
Agent 4	$1,000.00
Agent 5	$1,000.00
Agent 6	$1,000.00
Agent 7	$1,000.00
Agent 8	$1,000.00
Agent 9	$1,000.00
Agent 11	$1,000.00
Agent 13	$1,000.00
Agent 14	$1,000.00
Agent 17	$1,000.00
Agent 18	$1,000.00
Agent 19	$1,000.00
Agent 20	$1,000.00
Agent 21	$1,000.00
Agent 22	$1,000.00
Agent 23	$1,000.00
Agent 24	$1,000.00
Agent 25	$1,000.00

Symbol	Buy At	Sell At	Bail At

What is this Thing, Anyway?

The goal of the Swing Trader project is to to build a low/moderate-risk algorithm for day-trading/swing trading that is 95% automated. This practice is commonplace in modern investment, but not readily accessible to common investors (correct me if I'm wrong).

The overarching process is based on a very simple workflow with the following components:

Agent

An automated "bot" that is allocated a certain amount of money called a "stake". In the first implementation, there are 25 agents, each with a stake of $1,000. An agent can only work on one idea at a given time, and they will always buy the entire amount of their stake.

Idea

This is an investment opportunity. The idea consist of a stock ticker symbol, a date range for which is it active, a buy price, a sell price, and a dump price. The rules are simply:

• If an price of the stock falls below the buy price, it will be picked up by an agent, if one is available.
• If a stock is currently owned, and it falls below the dump price, it will be sold, and the agent will become available for a new idea.
• If a stock is currently owned, and it rises above the sell price, the agent will sell all shares, deposit the surplus, and wait for the next available idea with their original stake.

The magic will be finding algorithms that generate a great pool of ideas. I also want to look at giving the agents the ability to generate their own ideas (or at least use specific methodologies) This will allow ineffective agents to be "fired", weeding out bad algorithms. Also, I want to look at giving the agents their own ANNs trained using genetic algorithms.

Any thoughts on this project? Hit me up. I'd love to talk through any ideas.