Waiting to Blow up...

Discussion in 'Trading' started by Bad_Badness, Nov 21, 2020 at 3:49 PM.

  1. All constructive comments welcome. Here is an generalized summary.

    A discretionary ES trading system, 70 days for 5 points a day. Just finished it last Tuesday. I wanted to see how long before it blew up, because I knew it would. Why?
    • Because average win was 1.35 points and a 64% win rate, 3% loss rate, and the rest where "scratch", i.e. under 0.5 point P/L.
    • 5.4 trades a day, Min 2, Max 12, Mode 5.
    • Averaged down on unrealized looser up to 3x, to get a loss into a scratch.
    • Of course, that will not always work, so the stop got triggered at -25.0 TOTAL points.
    • Stop Triggered twice, or 1-35 days.
    • 12 days (17%) an unrealized loss came within 3 points of the 25 point stop but recovered to scratch. Happened 15 times.
    The "never mind" details:
    • Never went negative.
    • Order size was never an issue. Under 12 contracts.
    • The losses where well distributed in the 70 days.
    To me this seems very suspect to bother coding up. I don't think back testing is that great a determinate of future performance in general. And in this case, the system relies on real time data that is not available in historical data.

    I know this system goes against a lot of dogma, but.... :)
    Last edited: Nov 21, 2020 at 3:58 PM
  2. Big AAPL

    Big AAPL

    Looks good. Hard to code a discretionary system because "discretion" is a human emotion that is nearly impossible to code. I get what you are doing and if those numbers are correct, I might suggest a more realistic stop and reentry at a better price in the direction you were biased on in order to MAX out your profit and limit your losses. I know, it is a tough thing to do when you are sure you are 'Right". You're anchor charts MUST confirm this. I will also average down a loser but the day you let it run away from you with 3 contracts is the day you are on the news with the anchor saying, "before turning the gun on himself..."
  3. MarkBrown


    can i just ask what dictates the averaging down, do you have a rule?

    do you have a rule for any of this?
    murray t turtle and Big AAPL like this.
  4. JSOP


    Ok why do you think it would blow up and how do you think it would blow up? If you can't answer these questions, I would recommend you to do more backtesting then you would know for sure. And if you see that it would indeed have a high probability of "blowing up", then you should go back to the drawing board and redesign it. That's how I would go about doing it.

    One thing that I would suggest is not to average down on a loser, ever. Let the loser be, close it and move on. From my personal experience, averaging down or up from a loser is just not worth it. There is a reason why it's a loser in the first place and you putting more money into it is not going to turn it into a winner. Once in a while yes it does happen, but the odds of it happening is far less than just turning the loser into a mega-loser with more money thrown into it. And if you do see losers turning into winners later on far too often, then there is something wrong with your system in signal recognition. A good trading system should be able to give you clearer indications of losing trades from potentially winning trades.

    My 2 cents.
    Last edited: Nov 21, 2020 at 4:51 PM
    comagnum and Snuskpelle like this.
  5. SunTrader


    Testing done how and on what time frame?

    Keeping in mind anything about tick level is just a guess.
  6. It’s LOSER, not LOOSER

    Good luck
    Bad_Badness likes this.
  7. Well done. One of my pet peeves... along with ... "I can't believe what he payed for that car!"
    Clubber Lang likes this.
  8. @Big AAPL I used the term discretionary very loosely in that the methods is pretty mechanical. But coding in the larger time frame elements I know will take time. Mostly in the error-exception points. As for the reversal, that is the next thing I am going to incorporate. To me it silly to ride down something you know is wrong and more importantly the missed opportunity of riding a good wave, the bounce and the retest and the bounce again, i.e. 5-6 trades in a trending situation for this system.

    @MarkBrown, Yes, generally, a larger timeframe and daily S/R. I.e. the premise time frame is busted AND the larger time frame premise too. I could add the averaging down is less than 10% of the trades, because I use a trailing adjustable stop (TAS) so slightly positive trades tend to get kicked out on reversals. But as you know, TAS is one time adjustable at IB. If I could have N number of adjustments, then I could develop an alternative. In general, I am trying to figure out the mechanics of position reversal versus the averaging down method to scratch.

    @JSOP As for blowing up, it could be as simple as a string of the large stops. Since they are so large, a string of 5 would be pretty damaging. But that would be pretty bad and I just have not had it happen, but it could. Hence the waiting. I agree on the "cut the losses". Implementing them is a bit more nuanced, imo. My current thinking is to proceed with developing a reversal mechanism next. The main problem I encounter is the nature of ES. As you probably know, there are a lot of "flush outs" that occur and price then reverts back to a "mean" short term. In fact, they really need to be expected as the normal, imo. Frankly, it is a bit harder to figure out on ES so I did this method first. Also I can limit the trades to reduce the marginal losers, but one of the premises of the system is getting a higher number of trades. So it is a trade off.

    @SunTrader As for timeframe, I don't use tick bars. Since I am not trying to compete with the professionals and the system is designed to work over a standard internet connection and a retail platform, a lot of that data seems too granular to react to live and get orders in and executed. As for testing, it was my profession and my college education, so I take a much larger and formal approach than most. Given the nature of the system, I am only forward testing in sim before live. I could go into the reasons why, but it is not worth trying in a forum, but suffice to say, the effort versus validation is the main tradeoff. In other words, I am focused on getting traction first and system analysis secondary. After I get traction, I can fine tune.

    So I think this leaves me with the following conclusions.
    1) Implement the reversal mechanism. It will be the 4th or 5th time. I probably just need to KISS. EDIT: I realized I already have a system. KISS, just take the loss and get back into the trades! That will keep the complexity down.
    2) Do the analysis on the number of trades versus loss trades to find an optimal range
    3) Start the coding framework for the parts I know I want. There is a lot of work there.
    4) Do the analysis of the reversal versus average down methods, then take a step back and choose a set of solutions to implement and usage test, stress test and boundary test.

    I will probably post when the March ES contract expires. Also at that time I will be ready to fund this system.

    Thanks everyone for your responses. It helps to have others comment to think through the details.


    PS: Before anyone chimes in with the Sim versus Live, psychological and realism factors, I have a lot of experience live trading (decades and 6 digit swings). So working 4 and 5 digit swings is manageable. Also 95% of the trade "executions" are through the bid-ask threshold, so I am not relying on getting "jump to the top of the order que" magic fills. And I do have real swing trades going and a whole other system being developed there.
    Last edited: Nov 21, 2020 at 7:18 PM
  9. SunTrader


    As well I see were not where (twice) in OP and following post your not you're.

    End of today's spelling bee. :)
    Clubber Lang likes this.
  10. SunTrader


    Then you're guessing and sorry don't care how much experience you have testing other hypotheses beyond trading. You should know each has its own procedure/practices/inputs.

    In trading assumptions have to made where an entry and exit was possibly filled and whether or not a stop might have been hit at any point in between. Wide variations in P/L etc from sim to real world. GIGO.

    If right now accuracy is not important o_O have at it. Otherwise tick or at least 1 minute needs to be used to even attempt to come close to what might transpire.