I apologise in advance for the length of this post but these are complex and important issues central to your project's success and cannot be discussed in short post soundbites. Please do not feel you have to reply to every post though and please do so in your own good time - when you have time.
Per your posts above I'm happy to concentrate on the discussion of how to 'automate' the selection of systems to be traded at any given time now.
You say in your post above:
'...I do not change the previous combination. That rarely happened in the 11 combinations we traded. Almost every time, I add more systems/contracts to the existing combination...'
I don't know what the definition of changed is where you come from, and you may consider it to be semantic but in doing what is underlined above it was 'changed'
Anyways, whatever:
Just a summation then to make sure I am clear, mostly when the systems traded combo was changed it was by enabling more systems and/or more contracts to be traded by the prevailing systems trading in the combo.
So, from your posts above you consider a system combination to have failed when;
a. 200% of relativized max historical drawdown for that combo has been reached -or-
b. When the 'uncle point' comes re the investor (s) apetite for drawdown/risk.
whichever is the sooner, b. rendering a. an acedemic point should it be reached sooner.
Further it is the case (from reading your posts above) that 200% of actual (not relativized) max historical drawdown has been reached by the prevailing system combo;
'...Yes, here is the problem with our combination. We were expecting the systems to not exceed twice the maximum historical drawdown, but they did already...'
So the point at which a prevailing system combo thereby acting as a trigger to change that combo (so long as that point is above the uncle point of the investor (s) has been decided by you. I would argue that probably leaves you insufficient a cushion to turn things around by acting on this trigger to change the prevailing systems combo for a new one, before the uncle point' is reached.
Notwithstanding the evidence that the future tends to be worse than the past, once a system combo (or indeed individual system,) starts to exceed the known historical metrics, that is probably the point at which it should be changed / (disabled) until it starts performing again.
In choosing when to change the systems combo - thus far you have tended to do it when a new a/c hi has been reached by the prevailing combo (please correct me if I am wrong) or at the very least the preaviling systems combo has not been changes whilst in a period of drawdown ?
If the systems combo has been changed as a new a/c hi was achieved/period of drawdown was ended then you have either been lucky that the new combo went on to achieve a new a/c hi until this last combo or have been doing something right, or a combination of the two factors..whatever the case I question whether adding more systems to a combo that is clearly working is the right way to go about things, over and above adding more contracts to the systems already trading in the prevailing systems combo that have achieved the new a/c hi.
Even if you automate the systems selection for trading either via automation or via a manual change from a pre-determined fixed rule set; so long as all 120 systems that you know to be profitable in the back test (and most in the forward test) are not trading together all at once any such choice of which systems to trade and when - howsoever intelligently derived, leaves you open to the vaguaries of luck in such a choice to a lesser or greater extent.
If you do intend to change a prevailing systems combo it may be better to consider disabling individual systems as their performance hits a pre-determined lo as suggested above and opnly to disable them at this point. If a new system needs to be added to the prevailing combo to replace the disabled system then I would look for the next most 'stable' in the back test and forward test period currently not trading .. Consistency and longevity is key to trading and stability has to be a consideration.
In summary then and with an eye on automating the systems selection procedure perhaps you should only change a prevailing systems traded combo by;
a. adding more contracts (if trading capital available - within the more conservative money management parametres) to existing systems in a previaling systems combo at a new a/c hi, -or-
b. disabling a system (s) in the prevailing systems combo should a pre-determined more conservative negative performance metric be reached, and in so doing either
i. allocate the trading capital to the best performing system currently trading -and/or-
ii. to the next most stable system currently not being traded.
I need to give more thought to what may constitute a 'next most stable' system as discussed above and consider the factors 1-8 you mention in your post #2934, when I have done this I will post again in due course.
BBmac.
Replying as I read.
My point about not "changing" a combination is because you've said in previous posts "why did you change a combination that worked?" and my reply is always the same: we're not
disabling the systems that have worked, but we're adding more systems now that we have more margin and can increase profit without increasing risk (by increasing diversification).
You have to agree that, given your accusation of dropping a combination that works, it makes sense to reply that the term "changing" does not imply that we're dropping any systems
of the previous combination, but merely adding new systems. So, not just in my world, but even in the world you're from, you must agree that using "change" can also be wrong,
depending on the context. The combination does "change", but the systems from the previous combination do NOT change, stay the same, keep being traded, and often are even doubled,
so they are not dropped as the combination changes. What I mean to say is that you previously used the term "combination change" to imply the "dropping of systems from the
previous combination", so I could not accept that term, in the way you used it.
Yes, good point. We cannot turn around. Once a combination fails, it is over. However, after detailed analysis, after months of waiting, we decided to disable a few systems in the
first combination, which caused a change in the combination, and saw us go from #1, to #2, to #3, to #4, as we dropped 5 of the 10 systems I had initially selected. But the
dropping of systems is a slow process, because we need the proof that the individual system has failed. We'll be wiped out much faster by a combination that has delivered a
tripled max relativized drawdown, than we'll be able to identify which system has failed. In this case, combination #11, no system has failed. It's just that all their losses
happened simultaneously, which is something we could not expect. You see if we had to be prepared for all the individual drawdowns to happen at once, then we would not need a
profit cushion of 40k but 300k. My systems have an average drawdown of 5k. Multiply that by 50 and you get 250k. Then add to it that we're trading 2 contracts on a few systems.
We've never meant to be ready for all individual drawdowns to take place at once. We've only meant to be ready for the combined drawdown to be twice as bad as what has been in the
past (non-relativized), and along the way, we've also tried to take relativized drawdown into account (but this was done after the combination #11 was already selected).
"...until it starts performing again". Disabling a system combo until it starts performing again is not something we can do. None of us (3 people) think this way. It is either we
trade it from the start to the end, or we don't trade it at all. We do not agree with the logic that you wait for something to rise and then start trading it, because there's no
way to know if it will continue or if after a loss there will be gains. Otherwise trading would be easy: you just wait for ES to rise, then go long.
We have scaled up this last time after a high, but it wasn't a principle until now. You see: capital has only been increased 3 times. From 15k 1) to 30k, 2) to 45k, and 3) to 160k. So this might have happened after a good run, because of people being happy with my systems, and certainly it was not after a long drawdown. But it's never been established as principle. These 4 different levels of capital would make you expect 4 different combinations, but they have been much more, because I found ways to scale up without adding capital, due to the systems not trading all at once (they have different schedules).
I think the idea of waiting for drawdowns or waiting for new a/c hi should not be used, because you never know what is next. I think one should scale up while ignoring altogether how well the systems are performing. I mean: you should not time your scaling up based on the fact that you're in a drawdown or not.
Your considerations on how to pick the systems are good, but I feel i am more advanced in this area, so we would have to start from those 8 factors that I mentioned previously and find a way to combine them in a formula that returns a final choice. I am also good at synthesis, so I will probably do the whole thing myself, and your main contribution after all might have been achieved already in your insisting that I should do it. I mean: I am probably going to do it by myself, but thanks to your reasoning and so on.
Now, whether i do this by myself and totally freely, or I do it with the investors, according to their wishes, depends on whether we'll reach "uncle point" or not. If it all ends due to exceeding 42k of drawdown, then I will be on my own again, and this will affect how I work on this automation.
But let's also mention that I am burned out and that I don't have the stomach to do much more work on excel.
