Trading with automation (with IB) - II - spreads

Discussion in 'Journals' started by fullautotrading, Jan 8, 2018.

  1. As a curiosity, here is the chart of all orders fired on ES (red=sell, cyan=buy):

    #21     Feb 6, 2018
  2. We are on our 37-th day (counting also holidays and weekends) of trading and up to about 62K profits. In total 662 fills, and about 3.3K commissions.


    All the fully automated part has been performing greatly. For the "surrounding" discretionary strategies, significant drawdown, instead on "manual" entries on options. Also some drawdown on the manual entries on the ETFs, which have now turned mostly into small profits (e.g. VXX):


    Our current profit would be much larger (over 100K) without the "manual" entries, however, for our account the drawdown on options (short strangle) does not pose worries, as we can rollover forever and "eat" all losses pretty soon both by time decay and rollover contango.)

    The fully automated layer with largest profit so far is ES (open player view):


    its profit history:

    Last edited: Feb 15, 2018
    #22     Feb 15, 2018
  3. Our drawdown on the CL options is slowly being reabsorbed (currently around 29K), while the PNL is fluctuating around $150K (42 solar days) [we have made 811 fills with about $4K commissions]:


    All the fully automated layers are doing quite fine with the current game version (especially ES):


    I am contemplating to add a few more option positions to complement the main algorithmic game. I have also added a few more instruments like NQ [Nasdaq 100] (active) and HH [gas], GC [gold] (still inactive).
    #23     Feb 20, 2018
  4. Out of curiosity: why have you selected HH instead of NG? NG seems much more liquid (ignore the pun), with a much larger daily trading volume, and a larger variety of contracts to choose from.
    #24     Feb 20, 2018
  5. You are absolutely right. NG is much more liquid. In fact, it's the one I am currently trading (see the initial folio, and the yellow frame):


    I have loaded HH just to take a look at it, since it is a while I am not following it. (Usually NG does not perform as well as CL and SI, it's just less rich of "fluctuations".)
    #25     Feb 20, 2018
  6. I see. I expect that HH and NG have a high correlation as both are Henry Hub gas contracts. I didn't investigate this, but eye-balling the charts gives me this impression. The only distinction between the two instruments is that HH is financially settled, whereas NG is physically settled. Also the maintenance margins for both products are identical.
    #26     Feb 20, 2018
  7. Hello HobbyTrading,

    yes, as you note, the two instruments NG and HH have a high correlation. Even my "codirection" matrix (picture below) shows the maximum value (even though I have just launched it, so there has been not enough time for "accurate", long term, evaluation):


    From the trading point of view, "high correlation" does not mean (in our case) that the entries generated on the instruments will look the same. Actually, it can even happen that, at a given time, one instrument is in profit while the other (highly correlated) one is in drawdown. This is due to the fact that the entries are essentially influenced by the price curve "microstructure".

    (Also, for this reason, one can find preferable to use a "copy" (or "clone", as I call it) of the same instrument for experiment on hedging mechanisms.)
    Last edited: Feb 21, 2018
    #27     Feb 21, 2018
  8. The platform has been busy hedging the erratic moves and today the PNL is happily fluctuating around $200K (44 solar days, 871 fills, $4.8K commissions):


    after some rather large swings up and down (reached around $90K) yesterday:


    I have increased the "packet size" of almost all instruments, so there will be an amplification (in absolute terms) of PNL fluctuations up and down.

    In the meantime, I found a few hours to add some extra feature to the platform to improve and make more flexible the capability to transfer the trading information from a layer to another one.

    For instance, now one can "recycle" a layer (for instance previously trading an option expired OTM) to reduce the global resources used or, before option expiration, in addition to "ordinary" rollover to a next expiration, one can skip even to a different strike [same expiration] (thus ripping some higher decay rate). If interested I will show some examples of that.

    [In principle, one could even move an instrument to a different underlying (eg. NG to HH), which might initially seem rather abstract, but it might probably make sense, for instance, to switch from a stk/etf to some other, ranging in a comparable range, and with higher volatility or, in general, more suitable price dynamics. After all, all we care about is to extract money from a "suitable" price curve, and the "name" of the instrument that curve belongs to is immaterial.)]
    #28     Feb 22, 2018
  9. We have started a new trading week and the PNL is still fluctuating around $200K ($192K right now). Last week we had however significant PNL fluctuations and drawdown (which at its peak even brought the PNL under zero):


    Those PNL fluctuation were in part expected, as we increased the packet size, however also the mkt volatility and especially the addition of three new layers of NQ have contributed a lot:


    as in fact all losses are now concentrated on the new NQ layers (see picture above).

    All the options (that we are playing around the main algorithmic game with continuous rollovers) are slowly bringing some extra cash with their decay, which does not hurt).
    #29     Mar 5, 2018
  10. A good news. We have just surpassed our previous high-water mark of $224K, touching $246K. Currently with around $240K profits (almost 57 solar days, 1,352 fills, $8.5K commissions:


    Looking at the most profitable layer, it should become well apparent, now that we have a large "order cloud", what the algorithmic strategy essentially is doing:


    that is, we try to exploit all possible "inner" price fluctuations, and every time an order cannot be closed profitably (this could take the role of a traditional "stop loss" order) we simply "memorize" it as an open player (this is essentially the "trading information" we where talking about in the initial posts) and close it at a later time.

    Getting rid of the constraint of trades being performed in a strict sequence "open"/"close", we are thus usually able to unbalance the win loss ratio in our favor.
    And note that "prediction" is not minimally involved in this process.

    To see all trades, this is the same layer, showing all order executed, that is the "winners" and the open players (also waiting to possibly become "winners" sooner or later, when the price breaks new lows/highs):


    From this one can understand why we "need" (for better results) instruments which, structurally, do not have large "drift" or "decay" (or else it might happen that we could not ever close, in the future, a growing front of open players).
    Last edited: Mar 6, 2018
    #30     Mar 6, 2018