Overfit ... or over-fit?
As noted in this thread and in the In & Out Strategy thread, holding TLT (20+ year bonds) while being out did not work this time. Much of the work in this thread has focused on optimizing the in holdings. More work seems to be needed on the out holdings. A momentum-based selection sounds promising. The continued discussion in this thread should be able to specify a concrete approach at some point.
Has the In & Out fail us?
If we say out is hard-wired as meaning holding TLT then the answer may be yes. At least currently we are not better off than if we had remained invested in the market.
However, I was going back to the first two lines I had written on Quantopian when I introduced the In & Out Strategy back then in 2020:
Intuitively, it might be possible to generate excess returns via cleverly timed entries and exits in and out of the equity market. This algo may be a first step toward developing a strategy that derives optimal moves in an out of the market on the basis of early indicators of equity market downturns.
It seems that the original intention of the In & Out was to create a market timer. Musing a bit more about this, the in & out strategies discussed in this thread appear to have three main characteristics/components:
1. The in holdings selection
2. The out holdings selection
3. The market timer property
For the in holdings selection, several approaches have been proposed in this thread. Regarding the out holdings, these are alternative non-equity assets. As noted above, more work is needed to optimize this component.
But how did the In & Out do as a market timer telling us when to avoid equities?
Actually not that bad. I have a version (the v8) running that told me on 26 Nov 2021 to get out of equities which I did. At first it didn’t look like the best idea since equities reached new highs in Dec, but then starting with the new year it increasingly looked like a great move. Now I am waiting for the signal to tell me to get back in. (To generate the In & Out signal, I am using an approach similar to what Matthew Wang has proposed in the other In & Out thread.)
Long story short: The out holdings selection has disappointed. However, I am actually impressed by the market timer property. After the discussions on overfitting, ‘it uses too many signals’, ‘it shouldn’t react to individual signals but wait for multiple signals to fire’, and ‘with four parameters I can fit an elephant, and with five I can make him wiggle his trunk’ etc., the relative precision of the market timer is remarkable to me.