Algorithmic Trading a large folio with small orders

Discussion in 'Journals' started by fullautotrading, May 3, 2012.

  1. LOL! over $750 in commissions only for this test: i expect to be invited to the party ;-)))

    Exceeded now 2K profit with an avg use of 15K - 20K margins (45 days)

    I am continuing here my (real $$$) test from a previous thread
    since i am now applying the trading procedure to ETFs, some stk and options, instead of futures, as stated in the previous thread title:

    After testing futures, i am testing here (with real $$$) a possible approach for relatively small capital, which many where curioous about.

    In the meantime the application has "matured" and many strategic ideas have become more clear, especially with the suggestions of many of you helping out and actively discussing.

    It appears that the best way to go is distributing risk over a large folio, and rely on the sustained automate scalping/hedging action over price "corridors". I have also added a special "game" to scalp "short strangles" of options, which i am currently testing.
    The scalping activity in a large folio is so massive that the possible instruments with current dd ("investment") don't seem to hurt much.

    The situation is over a month, about 100 simultaneous instruments, mainly ETFs and options (i have been replacing instruments on the fly <b>discarding or adding on the fly new instruments</b> while i test them, about $600 have been removed from view where are the vertical green lines). I am keeping only <b>VM</b> as futures (the "small" VIX), as it is rather small. Trading ETFs and stks with very small sizes (30 shares) and allocating very small risk to each.

    Lately i have allowed to "overload" on volatility, even creating new layers (which is actually a violation of my "rule" of "diversificating", as i was a little anxious to bet the profits on volatility).
    A little "accident" happened Tuesday (May 1) where i was experimenting a new game to autotrade a <b>"short strangle"</b> of options and, by error, i happened to start it on 2 SPY options with a too small (for options) entry distance, while coincidentally SPY hit a 4-year high:
    Anyway, the drawdown was quickly recovered next day. This is again a big violation of diversification principle, but i was a litte impatient to test the new game.

    Now i have resumed the moderate scalping, with a some overload on volatility (VM positions). Looks like that <b>even trading algorithmically requires some discipline</b> ;-)) as we should <b>keep the folio always well balanced and distributing smoothly the risk</b>.
    Again, i let this folio to imbalance a little towards a bet on volatility, and this is somehow a discretionary and <b>very arguable</b> choice within the algorithmical system.

    Trading algorithmically such small sizes has caused a boatload of commission and over 2K profits on top of that. Practically i am in business with IB, with this test (LOL!) The Realized line (greed dotted) signifies the <b>scalped amount</b>, and is continuosly "<b>pulling" up the PNL, while the unrealized fluctuate through continuos new "investments"</b>. Now i just loaded a few stoks too, to take take a look.

    I have been keeping margins requirements rather low about 15K (with an occasional spike up to 28K caused by the SPY situation). I will be exploring more ETFs and stocks, to select a good folio for automated scalping.

    <img src="" width="1000" />
  2. Nice Job Tom! May want to consider moving this game to a broker without ticket charges.
  3. Hi PocketChange,

    LOL yes the commissions might look like a significant amount. ;-))

    However, to be fair, a few considerations are probably to be made. I started up the game with <b>really ridiculous small orders</b>, to see how the system responded, and i was not ashamed to issue orders of size <b>1 or 2 shares</b> only!

    (Now i am using mostly 20-30 shares). Clearly, when one spends $1 commission to buy a couple shares, this is an expected consequence ;-))

    Anyway, if i could normally use say 100 shares for each order (say with a 4x account), i think that the $1 is a reasonable price to pay for this kind of game.

    I think that this kind of diversification <b>feels good</b>. It's a risk that one is much more willing to take. The "granularity" of the futures is relatively big, and if one wants to diversify, the whole game is shifted to a relatively higher level of investement (probably in order of Ms), to be effectively diversified and "layered up".

    Personally, i am not really unhappy with the commissions, as they seem to be a natural <b>price to pay for the "luxury" of diversifying</b> (and probably sleep better... for those who sleep ;-))
    After all, execution is very <b>smooth and fast</b>, connection continuous 24/7 (including holidays and saturdays), and <b>simultaneous universal access to all mkts</b> (which is essential for investment diversification).

    Clearly, if IB offered me a further discount, i would not refuse it ;-)))
  4. This is an example of the automated scalping action: the one which actually scalped better (clearly, by chance). I added it later to the folio, infact it scalped directly 20-30 shares (blu=BUY, red=SELL).

    Each scalping "layer" is constrained to a max position of 3 "packets" (where 1 packet is now defined as 30 shares).

    <img src="" width="1000" />

    I am currently testing a special scalping "game" for <b>short</b> strangles of options.
    It sounds like a nice idea scalping automatically options (delta, vega, theta, ... ) ... to be verified better ;-)
  5. Today (end of this week) was characterized by the incredible plunge of oil. I was just looking at CL when it started falling from about 106 to about 97.5 in just a few minutes. A spectacular fall:

    In my large folio i have several ETFs directly related: UCO, USO, DUG, DIG, SCO etc.

    Interestingly enough even this huge move just "tickled" the whole folio. And this makes apparent the benefit of spreading risk.

    In particular the case of SCO (ultra short for crude oil) is interesting. Even though it took a big hit, loading up all the packets allowed on one layer, at the 3-rd packet the algorithm began a <b>"quick scalping oscillation"</b> (see the cluster of trades on the top right), which already recovered a good part of the drawdown. I have observed several times this behavior and it seems pretty effective.

    This is pretty interesting as it makes understand how massive the scalping action can be. And we are talking of a big move. The scalping action of the entire folio is so <b>massive</b> that even sharp vertical moves on single instruments are easily sustained. In time they either reverse or the scalping of that layer, or additional layers superimposed, make up for the dd.

    <img src="" width="1000" />
  6. So, ending up with almost 3K profits in addition to almost $800 for IB (commission), and having kept margins generally under 20K.

    Even the 6% today moves went almost unnoticed, lost in an ocean of small, massive scalping action. It looks like diversification is definitely a key concept.

    It's all about making a large game, a grand scheme that distributes risk as much as possible over a myriad of possibilities, while sustaining a continuous scalping action over multiple "price corridors".

    A question of proportions ...

    <img src="" width="1000" />
  7. Lampert


    Hahaha, some people would think that was a Freudian slip.

    Tom, I have been following your posts and would like to congratulate you (and your expert advisers), on your tremendous progress with the G-Bot trading system.

  8. LOL! very funny Lampert: it must be so, as i did not notice it, even after the elapsed time for corrections (which is when i usually get all my errors ;-))

    "Greed is good, greed is right, greed works..."
  9. Today was a fun day. While yesterday i saw very few scalps, today it was practically a machine gun ;-))

    While i test i add and remove instruments on the fly (the green vertical lines indicate where i have removed some instruments, along with the positive PNL and the corresponding commissions).

    Some instruments seems to scalp better than other, due to their "dynamics" and game parameters.
    So far those giving more satisfactions are guys like: VM, fas, qld, mdy, erx, ery, agq, bgz, dig, faz, qid, skf, tna, uwm, uyg, xop, etc...

    If you have some other candidates which might have similar behavior, let me know i will give them a try.
    I am also trying now a few stocks (i am not a fan of stocks anyway.)

    Unfortunately, the account seems to be limited around just a max of 100 different instruments (so IB is self limiting its own gain in commissions :))

    Fortunately, many different async layers of the same instrument "count" as one ;-)) and that is a great thing for diversification.

    <img src="" width="1000" />

    [Today i also "redistributed" so those working with the new version can give me an hand with the new games. Especially the new option scalper ("ShOPT3"), which needs to calibrated.]
  10. mickson


    Hi Tom,

    It has been way too long on my side and i wanted to congratulate you on the road you and your alter ego (G-Bot) are travelling.

    I look forward to being able to getting back into the testing seat in the not too distant future.


    #10     May 8, 2012