Really ? Such regretfull occurence.
Trading algorithmically a folio without stops (with IB), real $$$
- Thread starter fullautotrading
- Start date
Really ? Such regretfull occurence.
I shall be honored.
For real? I thought that was a joke...
Normally in such a situation you analyze the system, find the problem, improve it and go online again. It should also be possible to run such a system without this extreme capital requirement, by using cheap assets such as Forex and CFDs.
Well, He's aiming the strat for hedge funds, which a large portion do prefer
to hold equities rather then forex and cfds.
Sure AItrader, here is our update for this week start.
Metals giving hard time moving up more and then up and down in a small range. I have been hedging extensively trying to grab as much profit as possible on the moves up, which are those causing drawdown (DD), clearly. PNL at about -14-15% of capital.(You can see 2 spikes on the right extreme on the green G-L curve, which corresponds to 2 main rounds of massive hedging.)
From this useful experience, it seems to me that the best sizes for hedging ("protective") players (in this case buys), are from a minimum of 1/10 to 1/4 of the other side (in this case, sell players) (depending on how early one starts hedging.).
[Note that is is just one form of "use of past trading information".]
A key point is that the size of each player must not be relatively too big and they must be properly distanced, in order to be easily closed when necessary. This holds in general.
As we well know, we have "too many" of those metal ETFs in the folio, and as said in a previous post, I have "shut down" all the other layers (in profit) to focus all the attention and capital on these layers.
The instruments which at this point in time need to watched carefully are NUGT, GDXJ and USLV, which have attracted most of our investment. Clearly they are all short.
There are also the "old" metal layers, which we were testing previously (DUST, JDST, DSLV, ...) but those are not a priority now, and actually JDST is just in the "death row", as we have seen has become practically untradable (algorithmically) because frequently not shortable.
The situation is still interesting, but not worrying. In a next post, I will show in detail the order clouds of these 3 instruments which, at the moment, concentrate most of the DD. It turns out that the "best" instruments, among those metals, are NUGT, and GDXJ (where "best" here obviously is "relative to our purposes of scalping and hedging). USLV most of the time has rather large spread. NUGT is significantly more volatile than GDXJ.
The stocks we tried (WMT, BRK B, ... ) are totally "sleeping" and, in practice, a total waste of time, as they keep $$$ tied and move too little.
And, as promised, here are the 3 "bad guys".
The screenshots show the "player mode view", that is only the players which could not close in profit yet. (These have been taken right now, while mkt is closed)
The red horizontal line signifies the break even point for the open players. That is obtained by combining all the open buy and sell players and computing the resulting average.
So it gives a very useful visual representation of how far is our target. Clearly, this line will rise a bit at each buy scalp made by an hedging ("protective") player. In particular, it will go down when the hedging player open (buy, in this case) and, then, will go up again (a bit more) when the player closes (sell order). Also, overloading with sell orders on reversion (if there is one) the way down, will also cause this avg line to "come up", to "meet earlier" the possibly falling price.
The screenshots show the "player mode view", that is only the players which could not close in profit yet. (These have been taken right now, while mkt is closed)
The red horizontal line signifies the break even point for the open players. That is obtained by combining all the open buy and sell players and computing the resulting average.
So it gives a very useful visual representation of how far is our target. Clearly, this line will rise a bit at each buy scalp made by an hedging ("protective") player. In particular, it will go down when the hedging player open (buy, in this case) and, then, will go up again (a bit more) when the player closes (sell order). Also, overloading with sell orders on reversion (if there is one) the way down, will also cause this avg line to "come up", to "meet earlier" the possibly falling price.
Hi Fullautotrading,
thank you a lot for the update.
If I remember correctly, in the past, you (as the investment manager) have manually forced Gbot to overload on some assets (possibly the ones related with metals?).
Would it be possible to show the two investments components in the equity line shown by Gbot?
That is the investments for which the Investment manager is responsible and those for which Gbot is responsible?
This could also be a technical improvement for the next release of Gbot?
Thanks!
thank you a lot for the update.
If I remember correctly, in the past, you (as the investment manager) have manually forced Gbot to overload on some assets (possibly the ones related with metals?).
Would it be possible to show the two investments components in the equity line shown by Gbot?
That is the investments for which the Investment manager is responsible and those for which Gbot is responsible?
This could also be a technical improvement for the next release of Gbot?
Thanks!
Thank you AItrader. One crucial point that I want to stress is that, in my view, a fund management platform must not be something "magical", that you turn on and then you can enter in brain-sleep mode
, leaving every conceivable nonsense to it.It's quite the opposite. The algorithm is there to serve and do exactly what a fund manager would like to do under the given circumstances.
Again, we must definitely get rid of that kind of "Magical Thinking" where we have a black box doing something magic, that is beyond us to comprehend. There is no "magic", and there are no magic predictive "signals" in the tickdata (sorry about that
).It's quite the opposite: a full transparent set of rules that do exactly what the fund manager wants to do.
Clearly, it's the interaction between the fund manager and the algorithm which evolves in time, as the fund manager gets insight of hedging mechanisms and also according to his desired level of control of drawdown and so on. There are many parameters that needs to be sized: for instance, we mentioned the relative size of hedging orders in the previous post. Now, it's clear that there will be an "acceptable" range for those values, but it is also true that different managers may have different preferences, depending on a lot of factors, both subjective and objective (like the available capital, or the structural characteristics of an instrument.)
So, finally and ideally, there must be no difference from what the application does and what the fund manager wants (well, clearly a fund manager who has understanding of hedging techniques). And the fund manager must always have the "priority", that is, if the application is not doing exactly what he wants or if he wants to momentarily modify its behavior, he can at any time jump in and gently steer it to what he likes. (And I took particular care in allowing this kind of interaction is such a way that is immediate, easy and effective at any time.)
In time, the fund manager must have the platform to work exactly as a manifestation of his will, and anticipate any move he would like to do (clearly, that may take some time and "reciprocal adjustments" for that to happen in full harmony). [An analogous "journey" also regards the developer (that's me
) who, in time, become more effective in perfecting the mechanisms to allow that, clearly also based both on educated feedback and own experience.]Here is a "flashcard" on the hedging mechanism, we have been seeing in the previous post (hoping that one picture tells more than a thousand words: let me know if it helps)
