Robotrading: CT + Trending Strategy on folios of futures

Discussion in 'Journals' started by fullautotrading, Oct 11, 2010.

  1. hi

    have been doing more changes to the folio monitor (in the picture an instance just started 10 hours ago with the new panel).

    In particular, i have added other metrics, and removed some other ones.

    I have created a new indicator (for the moment called <b>"trending index"</b>) which seems to be quite effective for automated strategies, and, infact, by using it, i am obtaining some significant performance improvement (reward/risk).

    In the next posts, i will describe the changes. [I am almost done for this update]. This is a result of several contributions and discussions (email/skype) with other quants and traders and i am very grateful to all for the great support.

    Tom

    <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3194885" width="1200" />
     
    #191     May 26, 2011
  2. So let's see what we got so far. Let me know possible errors you see or suggestions.

    <b>PNL</b>
    As said this is a global view of the PNL of the various instruments. Each colum is representing the actual value. PNL is always <b>net</b> of spread an closing cost (it represents the amount actually pocketed by closing at the current instant).

    <b>PNL_DailyAvg</b>
    I have added this one. It's as above, but in form of <b>daily average</b>, instead of total.
    Within the first 24 hours is defined to be equal to above. After that, defined as PNL/DaysElapsed (hope this is intuitive).

    <b>PNL_MDD_Ratio</b>
    This one is the <b>"Efficiency"</b> we discussed in the previous post. I have adjusted the computation by considering at numerator the daily avg profit. So it is: 100K * PNL_DailyAvg / | Max DD ever |. The DD at denominator is taken without sign. The sign is that of PNL. It represents the <b>avg daily PNL for each 100K of max DD ever</b> ("efficiency" of the strategy on the given instrument / efficiency of the "investment").

    [will continue...]

    Tom
     
    #192     May 26, 2011
  3. As an implementation detail, for the efficiency of the previous post, in the case there is no drawdown (denominator = 0), i assume the index to be either 0 if the PNL is 0, or 100K if the numerator is positive. Let me know if you have a brighter idea to deal with this special case.

    <b>PNL_MPOS_Ratio</b>
    This is also an <b>"Efficiency"</b> measure. The difference with the previous one is that the denominator is the maximum absolute position ever seen (instead of max dd).
    This can be meaningful for margins considerations too.

    (this one is depicted in the picture).

    It represents the avg daily PNL for each base "packet" (in this case 1 futures contract).

    Tom

    <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3194958" width="1200" />
     
    #193     May 26, 2011
  4. Some of the following quantities are pretty straighforward and probably don't need much explanation
    <b>
    Real
    Real_DailyAvg
    Real_MDD_Ratio
    Real_MPos_Ratio
    </b>
    are just the same as the previous 4, with <b>Realized</b>, in place of PNL.
    <b>
    Comms
    Comms_DailyAvg
    </b>
    are the commissions, and the daily avg of commissions. Then 2 ratios, of PNL and Realized to commission follow:
    <b>
    PNL_Comms_Ratio
    Real_Comms_Ratio
    </b>
    Actually now that i am looking at them, i am feeling i want to "invert" them, as it is probably more meaningful to have the commissions for each unit of PNL or realized.
    I think i will define instead:
    <b>
    Comms_PNL_Ratio = 100K * tot Commissions / tot PNL
    </b>
    to express the <b>commissions for each 100K of PNL</b>. And similarly for the Realized.
    <b>
    DD
    MDD
    </b>
    are the <b>current drawdown</b> and the <b>maximum drawdown ever seen</b>. Drawdown here can be thought as the <b>"investment"</b> being absorbed by each instrument.
    so MDD is the maximum investment ever made on that instrument. (Here, the drawdown is defined as the distance from the highest local maximum the lowest local minimum of the PNL.)

    The following three:
    <b>
    Pos
    MPos_Signed
    MPos_Abs
    </b>
    are simply the current positions, the max absolute positions and the max signed positions.

    Now come the most interesting and newest metrics. The <b>"speed"</b> of the instrument.
    We will discuss that in the next post.

    Looking forward to suggestions (new metrics/corrections), in the meantime ... ;-)

    Tom
     
    #194     May 26, 2011
  5. unco

    unco

    Hi Tom

    What about a sharpe or sortino ratio for each product (and may be global portfolio).
    Not really useful tho, just there for information.

    Another typical HF ratio is round turns per million under management (in a year).
    But that would imply the user to put a capital as input... Also allowing to calculate folio ROR, DD, etc
     
    #195     May 26, 2011
  6. Hi unco :)

    yes, right, i though about the first two. Infact as you can see i have those in the simulation (random or past tickdata) facility.
    While for simulations i have a definite return for each trading day and therefore a series of returns, where to compute the returns, their avg and std, in a dinamic situation the question seems a little different.
    This is the reason why i am computing the <b>PNL / |MDD|</b> ratio, as expressive "of Efficiency". (Apart the fact that is anyway more interesting, because compare the PNL with the worst case situation). Same for the Sortino ratio.

    Unless you guys have suggestion on how to compute those in the dynamic context. Clearly, i could discretize, taking the various days, but the ratio would probably be not much meaningful or interesting - as you rightly point out - until we run for several (std need at least 2 observations) days (and often, when in profit we have just weekly restarts).
    So looking for suggestions ...

    As to the second one, we simply could "assume" the <b>Capital = max drawdown ever seen (MDD) </b>.

    How would exactly be, in that case, the suggested metrics ?

    Tom
     
    #196     May 26, 2011
  7. The following metrics are 4 speeds. Each one followed by the corresponding averages.
    <b>
    Speed_Abs_AsScalps
    Speed_Abs

    Speed_Signed
    Speed_Signed_AsScalps
    </b>
    Those ending in "_AsScalps" are expressed in "scalps" (in this specific case $250), the other are expressed in the current currency ($).
    For instance, Speed_Abs_AsScalps is the number of "scalps" per hour. Speed_Abs is the $ amount "traveled" per hour.

    Let's take a look at the picture below. These are realtime "speeds" (in $ / h).

    <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3195259" />

    We see how this is intuitively similar to the "volatility" concept, except that the speeds are much more meaningful to a trader, giving the exact amount currently "traveled", by the instrument, per hour. One can actually have a concrete idea of the "risk". While the "volatily" is more difficult to grasp (apart that as dispersion of returns around the avg).

    Tom
     
    #197     May 26, 2011
  8. Lately i am focusing on the hedging layer strategy, so not posting much...

    This week started nicely (a 22K in about 3 days).

    <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3204852" width="1400" />

    Tom
     
    #198     Jun 8, 2011
  9. Now that we have studied a little the "personalities" of the various instruments, will continue on a <a href="http://www.elitetrader.com/vb/showthread.php?s=&threadid=222126">new fresh</a> thread with a more manageable folio.
     
    #199     Jun 14, 2011