Hi AITrader,
thanks, that's a very interesting question.
Actually, the DD phases are those where most of the technical improvements and and conceptual advances are made. Everyone is obviously able to close profitable trades, but the real question is how you deal with DDs, how do you hedge the unfavorable moves, and how do you recover.
In this case the DD was extremely exacerbated (we nearly touched our "limit" of 50%) by some main factors, which ordinarily should not be replicated:
1. Most crucially, the use of RegT margin method with ETFs, which imho should be absolutely avoided because ridiculously strict (we are talking of about 3-4 times less available resources to trade)
2. Insufficient use of protective options.
3. Excessive packet size (which was introduced by error in one of my many game revisions) and it just happened while ERY was experiencing a unusually large move (106%), made even worse by the correlation with TNA and TZA, which with similarly large moves, acted as big amplifiers of the problem.
The experience is anyway undoubtedly useful because allowed me to make a number of important technical refinements to make the application totally able to deal with the situation in full automatism, without the instrument now being "blocked" at all by the order rejections due to margin violations.
For instance, previously I had an optional security mechanism which blocked the instrument after a number of rejections. Or I had another security measure consisting in blocking the order submission while asynchronously waiting for response messages from the broker. All these measures have undergone important modifications to make them perfectly fit to a situation of extreme drawdown and the necessity to deal with continued order rejections and also continuous broker liquidations. Also, I could perfect the mechanism of automatic order reduction on rejection. These are the kind of things that simply you cannot see or refine until you actually patiently experience them, and can actually see how the flow of events develops within the trading engine. It's a patient process of tuning and refinements that can only be made live, while actually trading and obviously cannot ever be even imagined in a simulation environment.
From the conceptual point of view, I think that the reader ready to understand what I am proposing has really seen the importance of the player superposition approach and why storing the past trading information is crucial in order to being able to "unwind" losses.
If you look carefully in what we have been doing in the "recovery phase", we have just been "undoing the losses" by using the still open players, or employing the open players to resize the sell players. So, we are not using some concept like predictive "signals" or other similar nonsense, but simply "reflecting on ourselves", on our past actions (which remain "stored"), makings action to build an "internal consistency".
So there is no attempt to use "illusory" signals hidden in tickdata (I don't expect personally tickdata to provide any "signal" useful to the trading purposes, but I regard those concepts as "illusions" or self-delusions created by humans through a process of backward adaptation of the strategy to the past data), but we simply look at the order cloud and operate to make it "internally consistent", which means in the long term to realize the buy / sell average "crossover".
After all, we have gone down to the "hell" of nearly 50% DD and made most back in a nearly 100% recover in few weeks. How would you trust someone who tells you he steadily makes 50% per year and his max DD is 10% (I receive tons of emails like that) ? And what happens if by chance, when abandoning his dream and simulation world and coming to real-world trading he sees 15-20% DD ? Would he be able to deal with the situation? Does he understand the reasons of the DD ? Does he have the technological and methodological equipment and a real clue about how to to recover, or will he proceed blindly trusting the same illusory "signals" which by the game of mere chance (along with the steady trading expenses) have put him in a losing situation?
[There is another conceptually crucial realization which I suddenly made just yesterday morning, just when waking up
and moved to immediately to put it into coding while still half immersed in a dreamlike world, about the use of past information, as finally realized how to use some other past trading information I have actually been "throwing away" (this has been for long hunting me, and somehow boiling in my subconscious mind), but this is a rather deeper question and will probably explain it in a next post.]
thanks, that's a very interesting question.
Actually, the DD phases are those where most of the technical improvements and and conceptual advances are made. Everyone is obviously able to close profitable trades, but the real question is how you deal with DDs, how do you hedge the unfavorable moves, and how do you recover.
In this case the DD was extremely exacerbated (we nearly touched our "limit" of 50%) by some main factors, which ordinarily should not be replicated:
1. Most crucially, the use of RegT margin method with ETFs, which imho should be absolutely avoided because ridiculously strict (we are talking of about 3-4 times less available resources to trade)
2. Insufficient use of protective options.
3. Excessive packet size (which was introduced by error in one of my many game revisions) and it just happened while ERY was experiencing a unusually large move (106%), made even worse by the correlation with TNA and TZA, which with similarly large moves, acted as big amplifiers of the problem.
The experience is anyway undoubtedly useful because allowed me to make a number of important technical refinements to make the application totally able to deal with the situation in full automatism, without the instrument now being "blocked" at all by the order rejections due to margin violations.
For instance, previously I had an optional security mechanism which blocked the instrument after a number of rejections. Or I had another security measure consisting in blocking the order submission while asynchronously waiting for response messages from the broker. All these measures have undergone important modifications to make them perfectly fit to a situation of extreme drawdown and the necessity to deal with continued order rejections and also continuous broker liquidations. Also, I could perfect the mechanism of automatic order reduction on rejection. These are the kind of things that simply you cannot see or refine until you actually patiently experience them, and can actually see how the flow of events develops within the trading engine. It's a patient process of tuning and refinements that can only be made live, while actually trading and obviously cannot ever be even imagined in a simulation environment.
From the conceptual point of view, I think that the reader ready to understand what I am proposing has really seen the importance of the player superposition approach and why storing the past trading information is crucial in order to being able to "unwind" losses.
If you look carefully in what we have been doing in the "recovery phase", we have just been "undoing the losses" by using the still open players, or employing the open players to resize the sell players. So, we are not using some concept like predictive "signals" or other similar nonsense, but simply "reflecting on ourselves", on our past actions (which remain "stored"), makings action to build an "internal consistency".
So there is no attempt to use "illusory" signals hidden in tickdata (I don't expect personally tickdata to provide any "signal" useful to the trading purposes, but I regard those concepts as "illusions" or self-delusions created by humans through a process of backward adaptation of the strategy to the past data), but we simply look at the order cloud and operate to make it "internally consistent", which means in the long term to realize the buy / sell average "crossover".
After all, we have gone down to the "hell" of nearly 50% DD and made most back in a nearly 100% recover in few weeks. How would you trust someone who tells you he steadily makes 50% per year and his max DD is 10% (I receive tons of emails like that) ? And what happens if by chance, when abandoning his dream and simulation world and coming to real-world trading he sees 15-20% DD ? Would he be able to deal with the situation? Does he understand the reasons of the DD ? Does he have the technological and methodological equipment and a real clue about how to to recover, or will he proceed blindly trusting the same illusory "signals" which by the game of mere chance (along with the steady trading expenses) have put him in a losing situation?
[There is another conceptually crucial realization which I suddenly made just yesterday morning, just when waking up
and moved to immediately to put it into coding while still half immersed in a dreamlike world, about the use of past information, as finally realized how to use some other past trading information I have actually been "throwing away" (this has been for long hunting me, and somehow boiling in my subconscious mind), but this is a rather deeper question and will probably explain it in a next post.]
Last edited: