TST Combine - Entering Algo into TST

Discussion in 'Automated Trading' started by Spectre2007, Jan 21, 2018.

  1. Thanks Spectre2007,

    I am using CQG demo free data with NT. And for one strategy (Intraday price action and 3 EMAs for trend) it shows great results for 2007-2008, but 2009-2015, the results is just not good. I am starting to question can I really trust free CQG demo data. They provide on 1 minute OHLC data.

    CQG demo data goes back to 2006.
     
    #41     Feb 6, 2018

  2. EMAs only work when you can filter signals through 1) consolidative 2) trend. Meaning if it’s a consolidative period, your profit targets are within the excursion. You can’t run EMAs through the whole day, the chop during consolidative periods will give you poor results unless you take a set amount of tick predefined profit. During trend phase you can let the profit targets be at the next cross.

    So than you have to break down the day into:

    Consolidation- small profit targets
    Trend- profit target at next cross
    Off- period can’t be defined, system is off

    With volatility at peak levels, excursion of price or waves of price have huge price differentials.
     
    #42     Feb 6, 2018
    Xela and SimpleMeLike like this.
  3. What the testing showed is, need to aim for homeruns, entries should have expectancy. But when they scratch, the dollar quantity of scratches should be made up by multiples. A high dollar amount winner would make up for multiple small scratches.

    It takes a ton of discipline to let your profits run for multiples. My inherent human nature weaknesses, are compensated by being programmed out by the algo. I strongly believe every platform should have a password protected code, where hard constraints are put in, and during the course of trading, the trader would not be able to change the parameters within the system he trades. Meaning shorts above the Opening Range would not be allowed. Or longs below the opening range should not be allowed. Below was attempt at trading tick charts couple years.

    There is a underlying bias in the ways the ticks move. Sometimes the brain can hone in on it. Like trading in the zone. But what ends up happening is, you become extremely confident to the point where you think you can't be wrong. And the loss multiplies to huge levels. Its better to have a low W:L ratio with risk management than to have a very high win ratio, with no risk management.

    c2test.jpg
     
    #43     Feb 7, 2018
    SimpleMeLike likes this.
  4. Great write up Spectre2007,

    I can share my comments on "let your profits run". When I manually trade (semi discretionary and semi mechnical), and I exit the trade for a profit, I get happy. Then it goes up for more profits after I exit, my first thoughts are "I soo wish i had just stayed in that trade". I convinced myself to always have a reason for exiting the trade. For example, I exit for profit because price was at a well known resistance, so lets exit profits. Or ema cross, price coming back down near entry, lets exit. Another thing is, we both know, with day trading, rarely the market makes big moves, especially with ES. For me, I like to measure my trade by reviewing the numbers/stats. I ask, am I getting good entries, RR>1 (or 0.8ish), is my profit targets good, when i get stopped out was it fake out or is price going in the opposite directions. I don't worry about letting profits run anymore. I just want to be on the right side of the trade and doing things right.

    This is why I like programming a strategy to answer the question 1. well what happens if let all setup run to market close 2. what if trail by X ticks. 3. what if I breakeven (need tick data for this with NT, lol) 4. what if this, what if that.

    We can not argue with algo after market close. All we have is the statics and backtesting results of what has worked in the pass until drawdown is hit in the future. Atleast this is my thinking.

    I let the stats decide if I should let profits run or not when dealing with automated trading. But manually trading, whatever chart shows me at the time, I take the profit if the profit target is reached.
     
    #44     Feb 7, 2018
  5. Thanks Spectre2007,

    Are you saying that :

    1. One algo (trending system) can not trade the ES market everyday?
    2. Multiple systems are needed for different market conditions (trending, consolidation) daily? If this is the case, if day trading, I wouldn't know if today will be trending or consolidation.

    Would not be much easier to let the algo run daily untouched rather choppy, trending, consolidation, ?

    Regarding choppy and trending system, just set system to only take 2 trades per day. if two losses occur in one day for trend system, then I would think not a trending day. Try again tomorrow.

    Each system I test for TST is only allowed 2 trades per day because of the daily loss limit. I have not tried a scalping systems yet. For this, i will need to use NT free 1 year of tick data they offer.
     
    #45     Feb 7, 2018
  6. The only way one algo can trade a trend system is to have trend filter to turn it off during certain times of the day or a filter built in where it flags itself that its a consolidation market and not to expect run away profits. Volatility usually implies trend. So you need a volatility measure built into the system. I use ATR AND STD.
     
    #46     Feb 7, 2018
    SimpleMeLike likes this.
  7. 4 trades, small profits. Consolidating market. 4trades.jpg
     
    • log.jpg
      log.jpg
      File size:
      76 KB
      Views:
      6
    #47     Feb 7, 2018
    SimpleMeLike likes this.
  8. Thank you so much Spectre2007, Now i understand what you mean.:) Wow, so all the trend systems I tried to back test in NT, I never considered a trend filter or volatility indicator. I just test the trend system considering all market conditions throughout the historical back test.

    In other words : if ATR > X (X known number I believe is possible trending market), then trend system can take the trade?

    Because the for trending system negative profits occur during choppy (low range days) price action days. Thanks
     
    #48     Feb 7, 2018
  9. Great work Spectre2007. Great profits.

    So your algo identified the consolidating (range) market today and adjusted to take small profits or trade range setups?
     
    #49     Feb 7, 2018
  10. I wanted to post this to show all the combines entered. The various combines I tried various things. But the reason most of these combines failed:

    1) lack of stop loss
    2) lack of algo to take over when screens unattended
    3) lack of discipline at times
    4) taking profits too quick
    5) not treating it seriously
    6) modifying code

    For all the reasons above, I hard coded and tried to remove as much of the bad habits as possible. I have a high degree of expectancy. When I stare at tick charts, I can tell in most markets what the underlying bias is.


    topstep.jpg
     
    #50     Feb 7, 2018
    SimpleMeLike likes this.