› ›

Factor Timing

aj165602aj165602 Posts: 105
edited December 2019 in Statistical Arbitrage

The aim of this journal will be look at the effectiveness of 'factor timing' approaches.

Over the last couple of weeks, I have finally got to the stage where I am happy with the way I combine factors, so my research area is now focused on capturing time-varying risk premiums.

To start with, I'll post a benchmark naive diversification strategy that does not attempt to 'time'.

Comments

  • aj165602aj165602 Posts: 105

    I got the dreaded "Backtest unable to complete" for this backtest, but I managed to exit the vast majority of positions on the final day, so the core statistics should be accurate.

    The hash for the test is:

    73aced134aa92e38e01f5230b745b1e5

  • aj165602aj165602 Posts: 105

    The strategy maintains a $10m notional throughout, so there is no compounding.

    The PnL for the 3-year period is $1,650,439 on 15,151,411 shares for an annual return of:

    0.333 * 1,650,439 / 10,000,000 = 5.5%

    and a Net Edge / Share of:

    0.5 * 100 * 1,650,439 / 15,151,411 = 5.45 cents.

  • aj165602aj165602 Posts: 105
    edited December 2019

    The Trading Report is not working, but by downloading the Excel file, I am able to obtain:

    Sharpe Ratio: 1.65
    SPY Beta: 0.00

  • aj165602aj165602 Posts: 105

    The next stage is to attempt factor timing, making sure to use data only available to a portfolio manager in real time.

  • jh909952jh909952 Posts: 5

    Just curious if your model is using factors only (I'm assuming data which I would call "fundamental") or is it a combination of factors and a price related model such as momentum?

  • aj165602aj165602 Posts: 105

    Just price and volume data.

    I've since moved on from this approach, however, as I now write down a risk model and attempt to be orthogonal to it at all times.

Sign In or Register to comment.