4. Rise of the machines
84% of all stock trades are done by computers
vs
Rationality
Speed
Capacity
Intuition
Understanding
Speech
- Cognitively biased / emotional
- Human
- Limited
+ 50 years of experience
+ Knows how the market works
+ Native speaker
+ Purely analytical
+ 1 TFLOP
+ Big Data
- None
- Depends on programming
- “I'll be back”
5. Pythonic Algorithmic Trading
Pipeline
Data Analysis
Generate ideas
●
●
●
Tools:
Pandas
Matplotlib
● (Seaborn)
IPython
Backtesting
Test on historical
data whether idea
can be exploited.
●
●
Tools:
Zipline
● Quantopian
PyAlgoTrader
optimize
Forward testing
Paper-trade on
real-time market
data.
●
Tools:
Quantopian
Live trading
Trade with real
money.
●
Tools:
Quantopian (soon)
6. Algorithmic Trading:
Backtesting
Historical stock
data
● datetime
● Price
● Volume
●
●
●
Trading Algorithm
Applies transforms (e.g. mavg)
Makes trading decision
Places orders
● Transaction costs
● Slippage
● Order delay
Performance and
Risk analysis
7. Zipline
●
Trading simulator/backtester written in Python
●
Open-Source (Apache 2.0)
●
Stream-based (prevents look-ahead bias)
●
Batteries included
–
●
Moving average, Sharpe, alpha, beta...
Used in production on Quantopian.com
–
–
●
Contribute back to community
Linus' law: "given enough eyeballs, all bugs are
shallow"
http://zipline.io
11. Conclusions
●
●
Zipline gives you full control over you
backtest.
Copy & paste your Zipline algorithm to
Quantopian:
–
Share it, get feedback.
–
Test it on minute data.
–
Forward test it.
–
Go live (soon!)