I find it hard to keep using systems that in backtests haven’t been working for the past three years.
Most of my backtests look like this:
2001-2007 +25% a year vs market
2007-2016 +10% a year vs market
2016-now (past three years) - not beating the market
There is irony what you say. My best performing models - nobody wants to trade. It’s in waters they simply don’t want to fish. But everyone want models in areas that are over-fished where no consistent alpha remains. But if many were willing to fish the remote waters…the alpha would dry up there as well.
How many people want to trade a sector specific market neutral product with 100 positions? So far the answer is none. (Acutally, the hedge fund Noviscient was interested but I would have to feed them each trade and P123’s current set-up is very tricky for market-neutral stock portfolios which work best in the screener - which is not optimal for exporting actual trades with re-balancing). Which is my best performing model? Same one. Hoping to on-board with WorldQuant soon so these models will likely die a natural death of old age.