Algorithmic Trading a large folio with small orders
- Thread starter fullautotrading
- Start date
Tom,
this has been a long read. I went trough all your threads since 2010 in some 20 hours of continuous reading and little sleep.
However, it seems that your threads on ET have dried up.
Have you moved the discussion over to LinkedIn?
this has been a long read. I went trough all your threads since 2010 in some 20 hours of continuous reading and little sleep.
However, it seems that your threads on ET have dried up.
Have you moved the discussion over to LinkedIn?
Hi windchaser, yes thanks.
Well in part some concepts are also being discussed also on this Linkedin <a href="http://www.linkedin.com/groups?gid=4394344&trk=hb_side_g" target='blank'>group</a> too, but periodically i restart a new thread (i think this was the 5-th or so) here on <b>ET</b> focusing on new strats and strategic development, as weaknesses are found. The <b>ET journals</b> are a fantastic way to stir discussions on these topics, while at the same time providing a practical example. Discussion alone can be "weak", and everyone can say anything and its contrary, when not supported by facts. But looking at the real application can be all another thing.
In other threads i have been taking a look at futures contracts for instance, and so on. Some threads are distanced by months or years, while i was focusing on new developments (once I even had to recreate the nickname because i forgot what it was
) ).For instance, i am now exploring the idea of combining the player layers (scalping) with <b>option</b> layers. The underlying idea is that options can provide a good complementary action to scalping. This clearly requires to trade more compact folios with larger sizes (considering that the minimum multiplier for options is 100).
Since the scalper can be "hurt", when using a contrarian game, by large unilateral moves (which periodically do happen), a layer of options may provide a form of protection. In general, scalping + options can complement each other.
This is an area which could be explored. I put forward some preliminary ideas <a href="http://www.datatime.eu/public/gbot/strategies.htm#ProtectingWithOptions" target='blank'>here</a>, but there is a lot to practice and develop, and i think it is worth exploring.
You have to look at these raids as the "installments" or "episodes" of a never ending journey. I have no "truths" here, clearly, I am just one who is researching and exploring always new methodologies to systematic profitability.
All the feedbacks of a large number of persons participating in the discussions and actively testing (or using and supporting) the application (along with my own trading activity) are a big driving force which helps me stay on acceptable and reasonable paths, and build up ...
Tom-
You have contributed an incredible amount of original and groundbreaking work by designing, building and developing the GBOT. Additionally, your 100's of posts here on ET spanning many threads and many years discussing your thought process and fine tuning of GBOT represent Herculean effort.
I started using the GBOT a couple months ago after having read all posts in your multiple ET threads several times.
It would be great to have you start posting again here occasionally.
One topic of could be discussing the current folio games (CTPlay, PureCT, TightCT and TPlay) and perhaps posting parameters for some new folio games targeted to either different time frames or different volatilities.
For example, I have a folio of 30 ETFs that has been running 24/7 for 6 weeks using the default CTPlay setting and 25% of the ETF's have never traded. In another folio of lower beta ETF's also using CTPlay, only 50% of the ETF's have traded.
Using default settings, I do get very good results with the high beta 2x and 3x levered ETF's. I also get very good results with low beta ETF's, but they just don't trade much.
You have contributed an incredible amount of original and groundbreaking work by designing, building and developing the GBOT. Additionally, your 100's of posts here on ET spanning many threads and many years discussing your thought process and fine tuning of GBOT represent Herculean effort.
I started using the GBOT a couple months ago after having read all posts in your multiple ET threads several times.
It would be great to have you start posting again here occasionally.
One topic of could be discussing the current folio games (CTPlay, PureCT, TightCT and TPlay) and perhaps posting parameters for some new folio games targeted to either different time frames or different volatilities.
For example, I have a folio of 30 ETFs that has been running 24/7 for 6 weeks using the default CTPlay setting and 25% of the ETF's have never traded. In another folio of lower beta ETF's also using CTPlay, only 50% of the ETF's have traded.
Using default settings, I do get very good results with the high beta 2x and 3x levered ETF's. I also get very good results with low beta ETF's, but they just don't trade much.
Hi CoopGekoLivrmor,
Thank you for your kind words and enthusiasm.
Yes, the scalping/hedging part is quite effective. The "sample" games enclosed in the application tend to be suitable for many 2X ETFs and futures. Plain ETFs or 3X may need a slight adjustment of scalping size and entry distances (reduce, or enlarge in case of 3X).
The part that is missing in these threads, and that completes the picture is the <b>options</b> part. They are necessary because the occasional large runaways aren't really for anyone, and somehow also pretty "masochistic" to endure, as simple option configurations will save a lot of troubles, and also greatly reduce <b>margin requirements</b>.
I have sketched a few hints about this part on <a href="http://www.datatime.eu/public/gbot/strategies.htm#ProtectingWithOptions">this page</a>.
Probably, in the future I will make a new "journal" here illustrating this part too ;-))
The very basic idea is that a simple <b>"protection unit"</b> like:
2 calls (buy)
1 call (sell) [to cover previous time decay]
provides, for instance, a basic protection for 1 short "packet", and so on, as the scalping game progresses. This is enough, in the larger scheme, to cap the occasional runaways and let the massive scalping do the rest, and make a killing.
My current view is to select a good folio of instruments. Not too many, but uncorrelated. Having options, with tight spread and trade decent sizes. (Temporary, "non systematic", correlations are handled by the application).
Here is a <a href="http://www.datatime.eu/public/gbot/GBOT_Tickers.txt" target="_blank"> selection of instruments</a> (ETFs, Futures, etc.) [the application sample ticker file].
Please, let me know other interesting candidates that are missing here (I will update periodically this file).
That brings up a few questions:
1) What is the optimal FOLIO SIZE with regard to number of instruments? 10-25 right? Or would a folio of 4-8 highly diversified instruments tend to perform better. The smaller folio would not offer cointegratiion benefits which may exist in a larger folio.
2) How best to select instruments for folios? Clearly while we want to create a group of instruments as uncorrelated as possible, and I assume we want to group SIMILAR BETA (or Volatilities) together?
3) In a levered ETF folio, what are the pro's and con's of including both BULL and BEAR ETF pairs in the same folio......for instance if we have FAS in a folio would it make sense to also have FAZ?
1) What is the optimal FOLIO SIZE with regard to number of instruments? 10-25 right? Or would a folio of 4-8 highly diversified instruments tend to perform better. The smaller folio would not offer cointegratiion benefits which may exist in a larger folio.
2) How best to select instruments for folios? Clearly while we want to create a group of instruments as uncorrelated as possible, and I assume we want to group SIMILAR BETA (or Volatilities) together?
3) In a levered ETF folio, what are the pro's and con's of including both BULL and BEAR ETF pairs in the same folio......for instance if we have FAS in a folio would it make sense to also have FAZ?
I think the guiding criterion is that the instruments must, first of all be uncorrelated. Then they must be liquid and have options. All those you can find with those features will be ok.
There is no point in using 2 "systematically" correlated (directly or inversely) instruments, as it only makes the folio management more cumbersome.
When I programmed the application I tried to make it completely "fool-proof", but it's better not too use it foolishly introducing correlated stuff) : this way the cointegration features will limit their action to the "temporary" or random correlations which may and do take place.
The "systematic correlation" should not be introduced. In case use larger sizes on the chosen instruments. This because we want to distribute our investment <b>on "essentially different" opportunities</b>.
Avoid "bear" instruments because they tend to scalp much less (and often slide to 0). Avoid anything has seen "splits" in the past. (Avoid futures with excessive contango.) Avoid the wildest 3X (or limit their usage in particular extreme circumstances, where "amplification" of moves can be useful).
Always use options as explained on my site.
There is no point in using 2 "systematically" correlated (directly or inversely) instruments, as it only makes the folio management more cumbersome.
When I programmed the application I tried to make it completely "fool-proof", but it's better not too use it foolishly introducing correlated stuff
) : this way the cointegration features will limit their action to the "temporary" or random correlations which may and do take place.The "systematic correlation" should not be introduced. In case use larger sizes on the chosen instruments. This because we want to distribute our investment <b>on "essentially different" opportunities</b>.
Avoid "bear" instruments because they tend to scalp much less (and often slide to 0). Avoid anything has seen "splits" in the past. (Avoid futures with excessive contango.) Avoid the wildest 3X (or limit their usage in particular extreme circumstances, where "amplification" of moves can be useful).
Always use options as explained on my site.