Just Follow the Model

Yesterday, based on looking at the candlestick patterns and Oscar Carboni's caution about a possible "holiday reversal", I decided to close my long position. I missed a big rally today as a result. At least I didn't lose money. But I shouldn't doubt the model. Now the model is signalling short. However, due to low volatility, I won't take this short signal. There was a similar signal on 9 January that made small gains for a couple of days and then lost big. Another similar signal on 11 May also lost money. So, seems a good point now to just step out of the way, especially as on Tuesday I am traveling to Europe.

The model gained 3.19% on this long trade.

That Worked Pretty Well

The model passed the test. We are now back to a positive return from trading for the month.

Started Trading Again on the Long Side

Following up on this post, I took the next long signal from the model, which was yesterday. I got in at NQ=7418.75, which was almost the low for the day. The market ran up sharply at the beginning of the cash session and then corrected sharply, ending down, but still above my entry point. Today's forecast is long again using the latest version of the model, which smooths one of the signals but using the old version with an unsmoothed signal, it might be short, or maybe I'd invoke the "close to zero rule", which said to ignore a change in direction of the indicator if the indicator was close to zero. Now, we have a clear objective rule. Let's see what happens.

But until volatility shows some sign of increasing, I won't take the next short signal.

NetWorthShare

NetWorthIQ seems to have died. So, following up on EnoughWealth's blogpost, I have opened an account with NetWorthShare. It's surprising that NetWorthIQ didn't make more of their website. I would have thought they could have got a lot of advertising from the financial industry.

Does it Ever Pay to Go Short?

I did some tinkering with the model to avoid the kind of false buy signal that resulted in the stop out last week. I applied Hodrick-Prescott filtering to one of my indicators. This eliminates these kind of false turning points but also eliminates a fairly subjective rule in my decision tree. So, overall that improves the model. This is one step further to a fully objective system that can be automated.

You need to be careful with HP filtering as it uses all the data in computing the smoothed estimate. So in back testing you have to run the filter repeatedly using just the data that was known up to that point.

The model is currently short. But I don't have a trade on. I am thinking to put a trade on when it switches back to long.

In a recent post, I showed that a hedged portfolio levered 1.5 times would track the market when the market does well and track the model when the model does well. Instead of thinking of this as trading plus investment we can examine it as a pure trading strategy. That suggests that it doesn't pay to go short. Just stay out of the market when the model is short and only take the long trades and lever up the returns. In 2018 so far, going short would add to returns though. But in 2017 going short detracted from returns. The model only won 45% of trades in 2017. The average win (1.57%) was almost double the average loss (-0.9%) though so, the expected value of a trade was still 0.2%. When we split trades into long (22) and short trades (24) instead the average long trade made 0.83% and the average short trade lost 0.39%. So, avoiding short trades would have doubled returns, returning 20% instead of 10% for the year. Of course, just going long for the whole year would have returned 32%. But we don't know that will happen ex ante. Levering the 19% by 1.5 times or so reproduces the long-only result.

The question now is whether you can win by going long only in a year like 2008. My intuition is that trading would result in a positive return for the year but that this would be insufficient to hedge the losses in an investment portfolio. It would moderate the downside though.

Testing that hypothesis will have to wait a little while.

But for the moment, volatility is low and so going long only might pay off.

Stopped Out...

My long position was stopped out on Wednesday. Now back to more or less zero profit on this account - still have a positive overall result from the trading experiment. Also, I realised that I still can't really psychologically handle trading overnight futures positions at the moment even at the smallest trade size. I am losing sleep because of it. So, I am going to stop trading for the moment. I need to get a lot of academic work done in the next two weeks before going on another overseas trip. At some point after I am back I will do some further research on trading models. My thinking is I could design a model that would only make trades when the odds were most in favor of winning. The current model trades all the time regardless. The ultimate long-run goal is automation of trading or taking it out of my hands in some other way. To achieve these goals I don't need to be trading continuously at the moment if I'm not making good money at it and  it's having negative effects instead.

I am also at the moment on a trip for a job interview. As I am learning more about the position it seems more challenging and to need leadership skills beyond what I have. I would be shocked at this point, though, if they offered me the position.