Version 0.3.3

Overview

• \(\color{#D7A84B}(500,250,150)\) is the length of the back testing periods with a roll forward of \(\color{#D7A84B}500\).
• There are 4 tests which are executed in total for each chosen list or combination of strategies/filters.
• Mathematical definitions: Mathematical Framework

Implemented Strategies

• \(\color{#D7A84B}\gamma _{mv,30}(1.99)\) (Time based Mean reversion with roll of 30, z-score of 1.99)
• \(\color{#D7A84B}\color{#D7A84B}\gamma _{cm v }(1.999)\) (Cross sectional mean reversion with z-score of 1.99)
• \(\color{#D7A84B} \gamma_{m t, 30 }(1.999)\) (Time based Momentum trending with roll of 30, z-score of 1.99)

Implemented Filters

• \(\color{#D7A84B}\eta _\text{regime}\)
• Regime based strategy implementation
• \(\color{#D7A84B}\eta _{pc(1)}\)
• Principal Component neutrality (to order 1)
• \(\color{#D7A84B}\eta _\text{dollar}\)
• Dollar neutrality
• \(\color{#D7A84B}\eta _\text{beta}\)
• (Beta neutrality)

Results

• Cross asset mean reversion and time based momentum trending work well.
  • Cross asset mv alone is slightly effective, showing positive SR in 2 different periods
    Period 3 Period 4
  • Momentum trending alone works very well on some periods, but not as well on others.
    It works well on Period 2 but not on Period 3 meaning it seems to be regime dependent.
  • Combine in a regime filter, they seem to yield the best results so far. \(\color{#D7A84B}\alpha >0,SR>0\) across \(\color{#D7A84B}2/4\) time periods
    • Combined with beta neutrality, dollar neutrality, and PC neutrality in the first component, \(\color{#D7A84B}\alpha >0,SR>0\) across \(\color{#D7A84B}3\) of the periods
      Period 1 Period 2 Period 3
• Time based mean reversion must be reduced as it produces noisy signals. \(\color{#D7A84B}\alpha <0,SR<0\) consistently in these situations across time periods
  • An example is Period 2
• Cointegration works in some periods, but not others, showing it is regime dependent like any other strategy class. Cointegration combined with pc neutrality of the first vector and dollar neutrality leads to positive alpha in all 4 periods, but not reliable Sharpe.
  Period 1 Period 2 Period 3 Period 4