High Frequency Statistical Pair Trading Model In msec & nsec - risk-free
1 ML K-Means Clustering
2 Correlation
3 Co-integration
4 auto-correlation (optional)
5 partial auto-correlation (optional)
6 band selection for entries
7 deep Q-learning for maximizing rewards
8 https://twitter.com/kerberos007/status/1228379802480103424
1 ML K-Means Clustering
2 Correlation
3 Co-integration
4 auto-correlation (optional)
5 partial auto-correlation (optional)
6 band selection for entries
7 deep Q-learning for maximizing rewards
8 https://twitter.com/kerberos007/status/1228379802480103424
Two good articles:
With the advancing technology in re-inforcement deep Q-learning & Quantum computing, risk-free for hedge funds
http://stanford.edu/class/msande448/2019/Midterm/gr1.pdf
Risk-free with tax cut money
Kudlow would pay us tax to buy stocks. https://www.hindawi.com/journals/complexity/2019/3582516/
With the advancing technology in re-inforcement deep Q-learning & Quantum computing, risk-free for hedge funds
http://stanford.edu/class/msande448/2019/Midterm/gr1.pdf
Risk-free with tax cut money
Kudlow would pay us tax to buy stocks. https://www.hindawi.com/journals/complexity/2019/3582516/
Something more productive
For retail traders
Statistical arbitrage swing trading using Index & Sector ETFs. long one asset; short the other
Tons of highly correlated ETF pairs, but the most important criteria is co-integration, meaning the SPREAD should be mean-reverting >95%
For retail traders
Statistical arbitrage swing trading using Index & Sector ETFs. long one asset; short the other
Tons of highly correlated ETF pairs, but the most important criteria is co-integration, meaning the SPREAD should be mean-reverting >95%
Simplest swing pair trading:
1 the "spread" time series of the correlated ETF pairs must satisfy the ADF co-integration test with p-value < 0.05
2 using the Bollinger %B (20, 2) mean-reversion on the "spread" series for entry & exit as explained for the $VIX strategy w/ hi R/R
1 the "spread" time series of the correlated ETF pairs must satisfy the ADF co-integration test with p-value < 0.05
2 using the Bollinger %B (20, 2) mean-reversion on the "spread" series for entry & exit as explained for the $VIX strategy w/ hi R/R
Secret of my machine deep learning model revealed:
Similar modes predicted tons of other stuff that I posted.
Not perfect yet, but improving as more data feeding into the deep learning recurrent Neurons.. They're learning like a 5-year old kid slowly https://twitter.com/kerberos007/status/1229875054411952135
Similar modes predicted tons of other stuff that I posted.
Not perfect yet, but improving as more data feeding into the deep learning recurrent Neurons.. They're learning like a 5-year old kid slowly https://twitter.com/kerberos007/status/1229875054411952135