What

First and foremost, this is an unproven method of choosing when to enter and exit a position. Since back-testing is used to produce the models future results may not reflect or produce expected returns.

2014-06-30

New Models Built - Paper Trading Starts Tuesday.

Based on the tests run over the past 3 weeks these new models should trade a bit more actively than the models built earlier in the year.

Here are the results using a start date of 2013-01-01

SymbolBacktest $Buy&HoldPrevious CallDateCurrent CallDate
bib41,595.6330,541.38SELL2014-05-06**BUY2014-05-08**
erx17,557.9925,108.92BUY2014-06-26**SELLToday
fas17,769.7323,387.34SELL2014-05-15**BUY2014-05-19**
midu14,679.7924,197.09SELL2014-05-06**BUY2014-05-08**
spxl21,619.7824,564.77SELL2014-04-28**BUY2014-04-30**
tecl22,020.0721,418.19SELL2014-04-25**BUY2014-04-29**
tna24,235.5123,740.55SELL2014-05-06**BUY2014-05-08**
ugaz13,693.1211,548.47SELL2014-06-24**BUY2014-06-26**
ugld13,026.253,751.80SELL2014-04-07**BUY2014-05-27**
urty25,515.9823,922.07SELL2014-05-06**BUY2014-05-08**
** - before model build date of 2014-06-30

2014-06-16

Change to Model Builder

1)   Optimized cross-over point for a buy call, prior to this builder change, used to check that the previous day's number was less than the value and current day's number was above.  It did the opposite for a sell call.  The new builder only checks for current day above (buy) or below (sell) the optimized value.

2)  This is an odd-one - the buulder would average out the days, sell optimal and buy optimal values for any combination that created a return greater than the buy and hold value.  The new method selects the specific day count and sell/buy optimal values for the combination that creates the best return over all iterations.

2014-06-11

Reset

I have discovered a flaw in my design or a bug in the models' software.  As such it will be necessary to rebuild the models.   I  will restart this whole process on the first trading day of July.