1 - 5 Minute Timeframe Automated Strategies (E-Mini SP) Before 2009 and After

Discussion in 'Automated Trading' started by Algo_Design_Kid, Aug 1, 2010.

  1. I would like the opinion of anyone else who is solely trading these timeframes and is 100% automated for the ES.

    I have been doing quite a bit of research (++2000 simulations) and it almost seems like as soon as 1/1/2009 comes around the market 100% changes gears and absolutely has changed trading patterns in these short timeframes.

    Anything that works 1/1/2009 (approx.) and after definitely does not work pre 2009 and everything that worked pre - 2009 definitely does not work post beginning of 2009.

    For instance, during the bull run of 2006-2007 you would think long strategies would also work for mid 2009 - 2010. My research has come to the conclusion of this not at all being true, in fact you could turn the equity curves upside down and that is exactly the difference between these years.

    The same for 2003 - 2006 compared to 2009-2010.

    As far as the variables involved I really do not use a lot of "classic TA", but have defined ranges (atr and by point value), the same open and close "triggers", and the same st. deviation bands as well defining when to enter setups and when not to. So the entries and exits are delineated well, *IMO*.

    From 2009 -> last Friday the same strategies that worked Jan. 2009 for example still work great - so they are still outpacing the Sp500 for the year but if I took these codes and tried to implement on any other year they are just outright laughable at best and you could have done better most likely following John Kruk.

    The sample size of trades post 1/1/2009 is 4000-8000 depending on how far I want to tweak the variables relating on how I enter the trade so it isn't a very small sample size. I am finding this puzzling.

    Has anyone trading these timeframes pre 2009 just totally blown up post 1/1/2009 that are 100% automated? It would probably feel like being stranded on a sinking ship in the middle of the ocean as you slowly realized no help was coming.

    I would SERIOUSLY be amazed if anyone on the planet could prove they are trading the exact same way pre 2009 -> present day trading in the timeframe of 1-5 minute bars with 2500 - 4000 or so trades a year. In fact I am going to have to say it would be impossible and I would be willing to eat every hat within a 5 mile radius if someone could show me this.

    But in all seriousness I am just looking for a good conversation, and it wouldn't be the first time I ended up having to salt a hat.
  2. nLepwa


    What do you work with?

    My price distribution model gives a stationary distribution pre and post 2009. I don't see any difference.

    I can see a very slight increase in the trade # for 2009 but this is most probably due to some random factor.

  3. Can you be more specific as to what you mean by "what do you work with"?

    I guess the OP was about as detailed as I can get without literally spelling out what my methods are.

    What specific timeframes are you referencing and what quantity of trades are you looking at?
  4. nLepwa


    I use a variety of timeframes ranging from a few ticks to a few weeks. I havn't seen a change in the (adjusted) price distribution for any of them.

    I don't want to give an exact number of trades but the sample size is significant (several thousands).

    This sounds like curve-fitting to me.

    If the underlying price distribution is not stationary the triggers are meaningless, statiscally speaking.
  5. I see where you are going with the curve fitting. I just find it very difficult to believe I could curve fit 1-3 minutes in the ES getting in and out of positions within an average of 2-3 minutes at the very most. A lot of times it is has unloaded the position < 60 seconds. Doing this greater than 3000-5000 times. Anything is possible though.

    Can you expand on what you are saying about adjusted price distribution please? I am not totally sure I follow.

  6. My view is that volatility changed into 2009 from 2008. Thus volatility strategies mislabeled as trend strategies would start to falter.
    This does not explain however why you had good results in another type of bull market. This may be caused by ineffciencies in your type of trading. IMO <60 entries and exit are too fast, you might execute only the bad trades, making a theoretically profitable strategy non-profitable. Did you forward test the 2009 or is the data different type?
  7. No fwd testing as of yet. I haven't really been digging deep in 1-5 min strategies for the longest time. Before was mainly focused on Hourly candles. Those still hold up (knock on wood). But all methods work really well post beginning of 2009. I am speaking about 20% - 35% gain from exactly 1 year ago and about the same from 1/2009 - 12/2009

    We are talking ranges of 1.5-2 ES points though on 1 minute candles. Same ranges (ATR) - same f****kin everything. Shit is just not "reacting" the same as in the past.

    IMO the 2003-2006/7 bull market traded VERY MUCH THE SAME WAY and if you try to bring that to 2009 you will be on the sinking ship. That is my opinion at least, but 1 user so far has still apparently had luck. I am finding this hard to believe but I certainly do not think my methods could be exponentially greater than anyone else's. The market gives lessons everyday, I am just a meager student at best.
  8. Try to get your hands on historical data from a different source and tell us what you find.
  9. No doubt. I am interested as well. I am wondering even if I am getting non real world preferential fills sometimes. It is all limit / stop orders that are either filled or cancelled at the close of every bar but I am skeptical if the real spread is always being accounted for.

    I have been there before.
  10. Are you continuously changing your system for market conditions in these short timeframes? For instance what length of time will you have the same strategy in play for in regards to your "tick" program?
    #10     Aug 1, 2010