Recently, it seems the model made most of its gains on its buy signals since 2009, and most of its gains on its sell signals in 2008. Why not ignore the sell signals during bull markets?
This market has been manipulated by QE since March 2009. Since then, most sell signals have been only mildly if at all profitable. The model thrives on trends, and thus this explains how it achieved +118.3% in 2009 and +83.8% in 2010 by going long/short/cash using the 3x ETF TYH, though in hindsight, even better returns would be achievable using TQQQ/SQQQ as this seems to provide a more favorable risk/reward quality. That said, since March 2009, most of the model's gains were made on the upside with the exception of the May 2010 flash crash where the model made a lot on its sell signal in a short time, +27.5% from 4/28 to 5/25, by shorting TYH on the 4/28 sell signal.
Despite the mostly bullish market since March 2009, if you look at a bearish year such as 2008, most of the model's gains were made on the short side, which is how it was able to finish 2008 up +38.8% using NO leverage, just 100% long/short/cash the NASDAQ Composite (1x ETF QQQQ makes a good proxy).
It would be nice to predict ahead of time which years will be bullish and which bearish, so the model can just focus on either its buy or its sell signals, but as O'Neil always pointed out when asked where the market will be 'X' months from now, no one has been able to do this consistently, so one must watch the market day to day and make decisions accordingly. That is what the market direction model does on a statistical basis as it receives new information daily.