Algorithmic trading for hedge funds: hedging techniques and application to a folio

[fullautotrading]

Because the method makes sense conceptually and "losses" are part of the method. It's called "investing".

(Not all are so intelligent as you, to have a "method" which entails no losses. The slight problem there is when, in addition to making no losses, the method also makes no sense, to a scientifically educated mind.)

 

[fullautotrading]

I am currently focusing on the recovery of ERY alone (two layers). Obviously, it's not possible at this time to recover (trade) TNA and TZA too with the current margin method, so their recovery will be addressed later, when we have finished with ERY. The DD has improved a bit, to about -19%.

(In the meantime, I have also added a few ERY puts, just in case there is a relatively quick reversion of the price and we might make up some of the losses "trapped" in the buy players which could not close.)

When we get the chance to start a new run (with the "folio margin", of course), I will also discuss another approach (more conservative) to trade these "drifting" instruments, in brief, consisting in allowing only automated scalps against the "long-term drift" (or, in any case, allowing to hold a position on that side only) and letting the main task to ride the "long term" drift to suitable option configurations.

In the meantime I have also used this occasion to further perfect all the automation mechanisms to deal with the margin rejections under extreme DD, and I am now pretty happy about the app behavior under that respect. [The "real trading" activity has an incredible number of tiny but crucial details, concerning the asynchronous management of the orders and events, which, of course, are not even considered in pure "backtesting engines", and that, actually, respect to mere backtesting facilities, represents an incomparably larger development effort (say, like days vs years), imh experience.]

 

[fullautotrading]

Ending the week with a bit more recovery (to -13.98%) due to some downward move of ERY. We could close some older sell players, which has reduced a lot the margin (right now just 63K). Accordingly, the G-L curve (which is in $ instead of EUR, because all instruments are trading in $) has gone momentarily down because now, feeling the "pressure" of all the buy players ("losses") above, which have remained open due to the margin issue.

If the price continues to go down relatively quickly we might take advantage of the puts we have bought for this purpose and the application would automatically increase the sell players (in this case they would be in phase we have been calling "protective") sizing them using a percentage (a parameter of the game decided by the manager) of the "open" buy players. My personal feeling is, instead, that the price might shoot up again to around 20 in the next week, which is a possibility I would anyway welcome, as, even though would put us again in a situation of "margin violation", it would allow to close some of the open buy players, while if the price moves down fast and massively, it may even never go back (due to the drift) to the level it peaked to this period (well, of course until the next reverse split).

This upward move of ERY was quite significant (probably the largest I have seen in the last 5 years or so), sweeping a range of over 106% in relatively short time. Of course if we had the folio margin and could close the buy players on reverse, it would have been a formidable ride, but anyway, let's deal with what we got.

At this point, the worst scenario would be that for the next month the price remains about here without going anywhere and then it starts falling slowly, according to the instrument drift.

 

[fullautotrading]

New week. Not much happening these first two days, with ERY hardly moving at all (moved a bit up). Just very few scalps.

Clearly, the plan is now to patiently let "recover" one instrument at a time (first ERY, then TNA and TZA), as there would be no enough "resources" to trade them simultaneously.

 

[OddTrader]

Hi jcl366, Can you blow up ? Sure thing. But if you are willing to embrace some drawdown (I personally would be willing to go to 100% with my own "risk capital"), have decent capital and suitable account with suitable margin method, have some patience to wait good moves, protect with options, and do the right automation (right software implementation), you can be actually "larger" than mkt, and systematically extract the "deserved" profits ("deserved" because you had patience, capital, sustained some DD and had the right technology and method). ...

Thanks for a very interesting thread.

Was the 100% DD mentioned above based on any back-testing with historical data? Actually, how do you know that Blowing-up can be a "Sure thing"?

Hi trend2009, [The novel part is the fact that we actually use past trading information to recover and "recycle" some of the losses due to hedging orders (thus replacing the concept of stop loss, with a more useful idea, where we physically store this information to inform the future strategic actions).]
Hope it's concise enough: let me know if also clear enough.

I think your writing is much easier to read than JH's . Comprehension is always my own incapability problem!

Hi jcl366, The approach I am currently using is not based on "signals", or any attempt to adapt to past data but it merely relies on hedging capabilities and capital, maximizing their action. There is no past data involved, except the trading information I create myself, and of course structural data (drift, etc.), whose nature is objective and of "mathematical" nature [which of course if one is willing, might even be replaced by some predictive analysis]. So the point is not "methods are curve fitted delusions, but yours is not?". But the exact opposite: they are not "mine" because they "are curve fitted delusions".


[/b][/b]

 

[fullautotrading]

Hi OddTrader,

> Actually, how do you know that "Blowing-up is a Sure thing"

The sentence was "Can you blow up ? Sure thing".The "sure thing" is intended to be referred to the "possibility" to blow up (e.g., small accounts, too large sizes, etc.).

> Was the 100% DD mentioned above based on any back-testing with historical data?

A 100% DD cannot ever be referred to "back-testing with historical data", because if one would see that, under those circumstances, he would obviously modify some parameters (for instance sizing, spacing, "signals", etc.) to avoid that. That is the essence of curve-fitting (or interpolation of the past).

100% DD is just a possibility that the future might reserve, for instance to someone attempting to trade a folio with a tiny account (eg. < 100K), or with relatively too big packet size, etc. It's in the realm of the possibilities. However, the subjective probability may be small, and actually vanishes as the risk capital, and experience increase, and there is more room to play meaningful games and hedging maneuvers.

In any case, the 100% DD in my post was rather referring to the psychological attitude towards the money being traded, which imho must be considered "risk capital", a work tool, because being emotionally attached to the money used for trading, almost inevitably leads to sure loss.

In reality, when you are near 20% DD, it already becomes pretty uncomfortable to trade (usually you are already receiving margin violation warnings), when you are around 50%, it is a herculean task or, worse, a "mission impossible" to get back on track. It's actually in DD that the usefulness of the "players" and "loss recovery mechanism" (use of past trading information) becomes really evident, imho. They allow you to do what, would be practically impossible otherwise. Clearly, there are people that don't see that, because they are so smart that their equity curve goes always up with no significant DD. Obviously, having never seen a real DD in their (fantasy) world of strategy "adaptation" through backtesting, they do not see or understand the need of any recovery mechanism and of maintaining the past trading information, while, instead, they hope that, "by magic", some new "signals" will make them back all the money gone in stopped trades.

> I think your writing is much easier to read than JH's ... Comprehension is always my own incapability problem

Thank you: that's quite humble of you. Actually, to me, maintaining a thread on ET, apart the pleasure to often meet highly valuable persons, is also an exercise for my English, in addition to the instrumental need to refine the trading methods, and the huge app code. It's always challenging trying to express yourself in a non-native language, so I guess that, in case of misunderstanding, it's just fair to assume that the fault is mine

 

[OddTrader]

Thanks. Not having enough time for the whole thread yet, I still need much learning to understand its values.

If possible, it would be nice to see a (monthly/yearly) performance summary report of your (simulated) trading covering last 10 months/ years.

Good luck!

 

[fullautotrading]

Yesterday ERY did not move a millimeter . Today, instead we had some activity. It rose to about 19.5 (pretty much what I expected) and then, very suddenly, just 30 minutes before the close bell, it dropped to about 18.3 (a 6% move downward was made in about 9 minutes!). This has given the possibility to clear a few of the buy players in that area (see yellow circle and compare the area with the previous-post screenshot), and relieve a bit the "pressure" from the buy side. If it continues fall tomorrow, we can start feeling some more relief: today I received "only" 9 emails of margin violation warnings . I did not touch TZA at all today. Instead, I occasionally reactivated TNA ("auto" mode) when there were margin available left from ERY to let it scalp a little bit. Now I put it back to sleep ("flat and manual").

 

[fullautotrading]

End of the week. We have recovered to about -11%. We need some more patience to entirely recover ERY. [Once done with ERY, I will address, one at a time, the other 2 main instruments which have many open losing players (TNA and TZA) remained open.]

Since the overall upward move against the drift was pretty large (about 106%), I think it makes sense to place some "bets" with long puts to try to recover some of the losses trapped in the buy players remained open due to the margin issue. Some of these will probably expire worthless (in hindsight, the expiration of 21/11 was probably too risky because close for that strike, and also the position a bit excessive. Anyway, I have added also some other expiration dates.) The time decay could have been neutralized, as we have done before, with short ATM options. Anyway, since in case of the price going up (as it actually happened this week) we could recover some buy players, we might consider that the bet was in part "paid" with the money recovered by closing those buy players.

 

[AItrader]

Hi Fullautotrading,

thanks for your continuous updating of your journal which I find interesting as always.

I am wondering what you have learned and what then you have changed (if anything) in Gbot algorithm
taking into account the period when the profit went from 0 to -300K+.

I am wondering because now Gbot is trading well and recovering the losses.
But actually it means that Gbot strategy is making profit and it is important to avoid the
to be trapped in the same situation that lead to the negative performance as before.

So how is Gbot now going to face a similar negative market situation as before?

Thank you for your time.
Regards,

AITrader