Algorithmic trading
October 14, 2023
Hypothesis #
I’m not sure this will work.
this might work, because
- market might be efficient
- but people are not rational
General Strategy #
need not perfect information nor perfect algorithm
I’ll perform the trading manually and see if I can automate some of my work.
Goal #
personal #
money #
beat the stock market by a noticable magine
- when stock market yields 5% return, I want to achieve 5.5% return (10% relatively)
study #
ai, finance, decision making
broader perspective #
openness #
benefit of openness #
- collaboration
open upto #
- I can make a living comfortably
Workflow (essential qualities to have) #
hypothesis - predict - see outcome - evaluate loop #
which data to use #
at what level, data is available in time #
- company / sector / country / world
integrate human judgement with rule prediction #
- probably more efficient to use human judgement
spell out the decision step #
Describe how a decision is made on what basis.
There are many hypothesis / assumptions i’m making when making a decision. I might not be able to spell out all, but when somethings work or not work, I need to look at the assumptions that I made to evaluate and increment
- emotions probably the most powerful factor that governs my decisions but overlooked by myself.
- hypothesis
- prediction rule
record decisions, I might need to review them #
decision rule, or thinking might be good only in certain situations #
For instance, when a stock price rise, it could indicate the company is doing good, or it could mean the stock is getting riskier. When I make a rule, I assume situations that the rule might be applied favorably Problem is that I can not see what I don’t see.
- test result shows that the decision rule worked or not under the specific situation which I don’t fully understand
Study plan #
Study curriculum This is being updated.
The prerequisite is
- calculus
- stat 101
algorithm integration principles #
Don’t know how much it will work. Don’t lose focus on big picture, algorithm is just for aid. Be aware that putting energy into it because it captures some idea without clearly specifying the assumptions would result in a wasted time.
On the other hand, when we decide to implement algorithm, desired properties are
Decisions as of now #
Update decisions as I learn