I'm hoping others can point out things I may have missed with the system I have been designing/testing. It's a very basic trend following system applied to a 5 minute chart. I developed it based on what fit into my day, my personal trading style and with drawdowns that are tolerable. It is intended to be a system that I can follow without second guessing. Commissions are accounted for, as is a small slippage. When applied to 25 momentum stocks the results for the last 5 months are : 119% profit, 178 trades, 49.4% win rate, 13.5% Max Drawdown. When Applied to only a select 5 stocks (FB, TWTR, TSLA, DUST, NUGT) the results are: 134% profit, 87 trades, 55.1% win rate, 13.2% Max Drawdown. Obviously the second configuration is better. The problem is it doesn't trigger every day and can be hard to watch and wait for. I'm still working on the portion that will text me when a signal is triggered. Eventually my goal is to have it fully automated. The drawdowns between highs are also more drawn out which may be harder to tolerate. Right now when this is happening (like this last week) I console myself by being proud of the fact that I have lost money according to plan. I've also tested it against stocks in a downtrend and it performs within expectations. DUST and NUGT are bear/bull etfs which move opposite to each other and both generate positive returns over the test period. I've also ran simulations of randomly selecting trades instead of just taking them in order and the system returned positive results, although obviously less trades results in lower returns. At this point what I really need is more historical data for the stocks that meet the volatility criteria. I'm afraid I may have inadvertently curve fit the system, although it has been running for two months now with results that are within expectations. Can anyone see anything I may have missed or potential problem I may run into? Also a good source for 5min historical data would be appreciated.
What language are you coding this in? I am assuming the text portion is going to be separate... I used this for an application I built. Works well. http://www.eztexting.com/developers/rest-api/sms/code-examples/p-h-p.html Although this is the more "professional version" and will cost money I am pretty sure sending <10 digit phone here>@vtext.com (Verizon Example) See here: http://www.popularmechanics.com/technology/how-to/4318204
I'm using Amibroker. So far I've been able to do most of what I want, although it can be challenging. If needed I can write a plugin in a c derivative. There is already a built in system for sending emails so I'll most likely do the test messaging through email like your verizon example It's not as graceful but for a small number of phone numbers it is nice and simple.
You really haven't given any info. From what you did give the results are positive... bet everything you have
Last 5 months the market has been extremely reasonable, last slight shocker was in 2011. I never trade custom lists, it should work on the whole universe or a list like Nas 100 or SP 500.
A dividend paying blue chip trades completely different from a tech stock and although there are some techniques that may work on both, there are just as many valid ways to trade them individually. Possible but I have tried to avoid it. Being a trend following system it works best with stocks that have a tendency to make large unexpected moves. I prefer highly liquid stocks or etfs so that I can get filled quickly with a signal is triggered. This is the criteria I have used to select the stocks for trading. There are actually some that perform better but are not as liquid as I would like.
Not in my opinion and you'd be introducing selection bias. If you pick stocks based on how well it worked before, that's a big problem.
Do not under estimate the value of the same rules working on a basket of stocks. And I agree with you, that you want to trade stocks that move, so eliminating the dividend payers is valid. The basis of a system is selection bias. You need to find the rules that work, or the stocks that work. AT&T isn't suddenly going to be a mover, so eliminating it is perfectly fine. The only issue is your short test period, where the market has floated all the boats. IMO you need to test from say 1997 on. That is the selection bias you need to be concerned about.