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

It may perhaps be interesting to take a look at the 2 (currently) "extreme" instruments: maximum and minimum current PNL, to make some considerations and perhaps understand and take some actions on the "losing" one, if necessary. They currently are TNA and ERY (-1.8K right now).

At this time, TNA has still the most successful (non-predictive) scalping action with (currently) about 6.9K PNL, a max of 7.5K and a min of -1.1K. It is working with the so called "Shield+" s/h game, that, as we know well at this point, works with an upward bias, and a "long position constraint". During these 43 days, it went initially a bit down and then went up, thus realizing a nearly perfect configuration (with a "load" phase not too big) for this kind of scalping. Also featuring a price curve "rich of retracements" which allows the scalper to grab several trades up and down. We have been "protecting" it with various "option-share configurations" and currently (after the recent expiration) there are 4 puts in protection (I will add more if it moves at least a 4-5%):

TNA_2.png


The 2 pictures show both the same TNA layer: the top one the open players only ("player-view mode"), while the bottom one all executed orders. Comparing the pictures, it is obvious that the scalping action largely exceeded the hedging action, thus causing (so far) a good result in this case.

In the next post, we will take a look at ERY, which, being negative, is probably more interesting and instructive.
 
Last edited:
Let's now take a look at ERY. This went similar to TZA. Even if PNL is still negative, the situation looks perfectly fine. The instrument has been scalping in a 20% range. There are still a few protective buy players which are still open and that there are responsible for rising the buy-side average a bit too high. Considering the normal behavior of this instrument, here there has probably been "excessive" protection on the way up. It's fine anyway, we either "unwind" those " losses" (close the players) if the price raises ("loss recycling") or we take into account their size in the following sell players. In fact, the instrument is now actually in the "protective" phase on the sell side ("sell hedging-protecting"), because the "scalping-loading" sell players have been closed and now the application must compensate for the buy players open (as we remarked earlier, this is a form of "use of past trading information"). Furthermore, we also have 10 protective CALLs in place. In other words, we have probably exaggerated in "protection", and if we took a little more risk with this instrument we would already be in the green. Anyway it's ok.

ERY_2.png
 
Last edited:
Added another "option-share protective configuration" to TBT. Usual type, with "ratio" R=4, expiration October 17, in 44 about days (we have currently 19 Calls in place for TBT, 5 of which expiring in 2 days):

TBT_OPT3.png
 
It is also interesting to look at the current situation of UCO. This was the instrument which seemed to start a runaway and we have "armored" with a "load" of options. It turns out now that perhaps we have exceed in prudence (of course we do not know the future, and our approach is not predictive), as in fact now the layer turned even positive by itself.

It is interesting to note that it did not require any "reversion" to do so (clearly if it reverses we may make more $$$ by closing the open buy players), but the mere scalping action on a relatively "narrow" corridor has already made up for all the previous drawdown (previously touched -3.1K):

UCO_2.png


Note that the calibration of the size the new scalping players based on the buy players which were already "open" is a form of "use of past trading information". Disregarding (or, worse, not storing at all such information, as it is still customary nowadays) obviously would be an unforgiving loss of precious information (which, of course, may ultimately be the first cause of unprofitability, or at least gross inefficiency, from a conceptual point of view).
 
Last edited:
Today GDXJ is finally moving significantly the way down. Added another "option-share protective configuration" to GDXJ . Usual type, with "ratio" R=4, expiration October 17, in about 43 days (we have currently 16 Calls in place for GDXJ):

GDXJ_1.png


This move came just at the right time, because today we are also "investing" a good chunk of PNL "loading up" on a marked upward move of ERY, so it served a bit to mitigate the DD due to ERY.

As a simple practical "rule of thumb" for adding options, currently I am just watching the manual layer and adding a new "option-share protective configuration" when there is a move of x%, where x = Volatility / 10 (let's call this the "10% rule"). This is a criterion pretty easy to follow in practice because we can just take a look at the green lines on screen (which are spaced 1%) and the realtime volatility (also printed on the screen). (We may refine it later on.)
 
Last edited:
And here is the automatic layer of GDXJ. Top picture shows, as usual, the open players, and the bottom picture all orders. While the price was quickly plunging, the application placed a series of "protecting" sell players which scalped a nice profit, since the "Bias-" game is asymmetric and the protecting sell player has a large size (in this case 60% of the other side (buy) players, about 1,400 shares).

GDXJ_2.png


Finally it closed almost all those sell players, because the price momentarily reversed direction, thus remaining with just -100 shares short for the overnight gap (that was nice). It has in fact (in the "layer settings") a constraint to stay short at least -100 shares (unless protective buy players must be added):

LayerSettings1.png


(Later we will see more in details the layer settings.)
 
We ended this week. Here is a folio update:

PNL47.png


Today was a bit of a "rodeo" (which is good for scalping). As anticipated in a previous post, today we had a pair of TBT options expiring. After IB processes them I will discuss the actions needed to "re-sync" the application. (At this time the app is still in "sync", which means the broker as not yet processed them and removed from the account.)
 
Ok, the 2 TBT options have now been removed. In fact we get from our account view (green $ icon):

Code: 
============================================== NOT MATCHING PRESENT IN THIS INSTANCE ===============================================
Symb                                                                       This G-BOT             IB            Exp
TBT                                [TBT   140905C00063000]                          5              0        Expired
TBT                                [TBT   140905P00057000]                         -1              0        Expired
------------------------------------------------------------------------------------------------------------------------------------

In this specific case, they have expired both OTM, and if we login in account management and go take a look at the IB trade report we can already find a perfect PNL match with the application. So, in this case (no exercise, no assignment) we do not have to do anything, because when we also remove the option layers on our side, we are in sync again with the account.

TBT_OPT4.png


Clearly, since the short option (put) "came" with 100 shares short on the manual layer (the option-share configuration was: +5CALLs/-1PUT -100 shares), we should now close them to remain consistent with our systematic methodology. We can do that when the market opens.
 
Last edited:
I have just removed the expired option layers and closed the 100 shares short on the TBT "manual" layer (belonging to the "protective configuration"). The application is now perfectly in sync with the account, and PNL accurate to the penny:

TBT_Man1.png


Also, I have placed another "protective option-share configuration" (+3Call/-1Put -100, expiration October 17) for ERY, which is still in a "load" phase with a pretty large position.

ERY_OPT1.png
 
You must log in or register to reply here.
Back
Top