I spend countless hours on statistics underlying the system I outlined so here’s some background. The model is similar to card counting in Blackjack where one bets when odds favor you over the house. My model bets when stocks are more likely to go up than down.
The obvious but central idea with statistical trading is that it's memory-less. With odds 75% positive (so 25% negative) after a losing trade, the odds are still 25% on losing the next trade. Over short term periods one will sometimes lose. For instance, there were losses in 3 of 18 three-month periods in the last four and a half years; There were no losses in any of the 9 six-month periods but in 2, the first half of 2014 and the first half of 2017, returns were barely positive. Attached are six-month returns.
One question I’ve received was: Why not just buy-and-hold (B&H) the S&P. Column 8 shows it’s returned 67% (actually even more since this is log not percent return) during the period whereas the model returned only 19%? The answer is risk. The model had 102 trades so market exposure was about 433 hours (4.25*102) whereas B&H’s market exposure was 39,420 hours (24 * 4.5 * 365). The model returned 30% of B&H’s return with only 1.1% of the risk exposure.
Another question was: Have most afternoon’s been profitable for the S&P during this period? This is a tougher question. Yes, the S&P rose on most afternoons so continuing the blackjack analogy we had a loaded deck.
Column 7 shows the S&P rose 18.1% on 1134 afternoons. (56% off these afternoons were positive.) But still the model’s return of 19.2% winning 76% of the time on 102 days is significant. Enough that even with a neutral or biased deck, ie the S&P flat or dropping in afternoons, I feel returns would be positive.
A crucial question I created is: How has the model done lately? The answer is: It wasn’t worth the effort, ie, 0.2% return from 11 trades over 6 months (row 8)! Given this, before engaging one should decide, is this "game over"? Or is this a short term result that’s to be expected occasionally?
I vote for the latter arguing: Everyone needs some S&P exposure so why not get it on historically favorable afternoons. With the added proviso, since the exposure is so short, why not leverage up.