Pretty much babysat a commodity ETF all day (ultrabear with CL as the underlying) using Multicharts and the IB DOM to monitor CL, indices and the ETF. Anyone trading oil today knows how that worked out
Finding the combo (MC and IB) absolute luxury, but could use a few more screens to avoid switching between MC work spaces (there are 2 x 27" monitors attached to the trading computer). On the topic of browsers, the "new" trading computer hasn't abended in a BSOD in a couple of weeks, coincidentally since I stopped running IE and NT and did a Windows 7 update. I also removed a cigarette butt from the interior of the computer--possibly mine but no idea how it got there and still no clue what the cause of the BSOD was. In any event correlation does not imply causation.
I wrote some experimental indicators in PowerLanguage to develop the notion of translating color into something a bot can understand and tried to bounce ideas for heuristics off my Gen Z son while giving him driving lessons last night--essentially how to translate what color in an indicator conveys to me when I'm trading (essentially the first derivative, or rate of change) into feature vectors for input into artificial intelligence (AI) engines. He also trades and is enough of a geek to understand what I'm talking about and would like to do more but so far unfortunately is more into texting and cars, looking for a vehicle appropriate to his station once he learns to drive (he's 21. His siblings were driving the instant they were legal). Not sure what to make of Gen Y & Z and their fixation with Beemers and turning cell phones into flashlights
Experimental PowerLanguage indicators are attached for MACD,15 EMA, 50 SMA and Stochastics for sake of amusement. All of them use the previously published RGB/HSV conversion functions and all try to reduce the respective indicator applied to price in 3 time frames (setup or medium time frame, entry or short time frame and confirmation, or long time frame) to a single coloured line. Technically speaking they all work but fail miserably in their goal because the heuristics are insufficient (whence the monologue with my son mentioned above). The heuristics depend on being able to objectify (to the extent a bot can grasp it) exactly what an indicator provides to a trading method. If the heuristics are insufficient it likely means my understanding of my trading method is insufficient, which likelihood is supported by the fact my P&L trading short time frame spot Forex is oscillating around a fixed point at the moment
Tried to write the algorithms in the indicators as functions since functions will be required by a bot ("signal" in MC speak) but it turns out I don't know how to persist variables in MC PowerLanguage (requires global variables) except by defining them in the calling algorithm and passing them back and forth by reference--the way C does. If a 3rd party DLL is still required to implement global variables in MC don't want to go there right now. The function equivalents work, but are uglier than the indicators so not included.
The .PLA file is included for MC users and text for each indicator in a .ZIP for any who would rather view or convert to another language.
Last edited by bnichols; October 23rd, 2012 at 05:51 AM.
Spent almost the whole day learning how to write signals in Multicharts, and transforming aspects of the 5 energies system into experimental signals. BTW, the screenshot in the previous post is in support of this work and designated as the strat workspace, not something a human would trade from
In other news I lost $600 cash (my daily limit) trading gold ETFs today. I tried 3 times to catch a falling knife when I know far better than that. I think I've resolved the issue but tomorrow will tell. In contrast currency trading is going swimmingly, averaging +$162/day over the last 5 days trading 1/2 contract (except when downshifting and stomping on the gas pedal--increasing to 1 or 2 lots--as required to pass a spot of bother), coincidentally since I switched to Multicharts on the cash currency account. Only one hiccup when TWS lost the connection with its servers just as the London market responded to something someone said around 4:30 AM local time (not at my best at that hour anyway) and dropped like a rock when I happened to be long EUR/USD. Unlike IB to drop the ball on tick volume--trust it was a one-time thing. Luckily the drop turned into enough of a downtrend to turn those frowns upside down.
Trading continues to sink in, slowly but surely. I continue to be reminded that even though we think we know a thing (e.g., placing an order will not cause the market to snap out its 10-pip range or come to its senses and start to move in the direction we want it to go) we very often in fact haven't learned that thing very well as evidenced by our actions, and will continue e.g. to place ill advised orders out of boredom, or irritation with the market, especially to place long orders (when every scrap of evidence before us proves beyond a shadow of a doubt price is dropping and will likely continue to drop) simply because we don't want to go short today, and not trading means we'd have make a start on the to-do list maintained by our significant other. Much work to do on ourselves as traders before we sleep.
The more I use Multicharts the more I like it. A little like doing that business trip in a Caddy when the office accountant told you all you could expense was a Neon.
Last edited by bnichols; October 25th, 2012 at 10:35 AM.
Reason: typo / clarification
Developing bots since I last posted, which means paper trading and coding.
Just discovered a new parameter I haven't grasped yet and hence haven't coded yet. When it is coded it will probably be called Trust. There is a randomness about price action (search for value) this AM that I don't recognize, hence don't trust, hence am not trading. Well ok, just now for a few bucks once price commits.
Trading the Pac Rim market this morning and looking at price vs the signals I use, and after 3 years trading EUR/USD exclusively, I'm still seeing price action I've never seen before. For the time being it remains inscrutable.
One can enter by peppering the market with orders, and with a long/short bias, but so far the fact remains I have no clue where the market will be when I wake up later this morning.
Someone is arguing with someone else about where the EUR should be right now. Can't begin to program a bot to handle domestic disputes. Shorts are winning at the moment. I can imagine someone shouting, "Well then what is it worth".
IMO I'd have to go to school 2 or 3 times to program a bot to capture this action. Leaving this at 1.2900 spot EUR/USD. Made a few shekels.while writing this.
Believe the randomness ("Trust" issue) I ran into last post re EUR/USD was due to stop hunting, which like news can artificially bias price--one skilled trader with megabucks driving price to its likely conclusion by triggering stops. Maybe s/he has a big order to fill
ETA: in any event long on paper at 1.28965 assuming all the folks buddy took money from all night want their money back.
Cocooning since the last post, paper trading CL to work the kinks out of a new "long/short/chop" MC PowerLanguage indicator based on Barry Burns 5 Energies system. As mentioned previously I hit my daily loss limit last week of October in an oil ETF with CL as the underlying and swear that is the first and last time that is going to happen . Shudder to consider the cost if I'd been trading the future directly.
The indicator should run with any instrument although I notice the sub-indicators it depends on react somewhat differently (more responsive?) to movements in CL than in spot currency, so some tuning parameters might be required. As usual the main issue remains detecting chop.
The plan is to turn the indicator into a function and wrap a signal (strategy) around it to turn it into a bot in order to see whether it will optimize, and perhaps deploy it if it's robust enough. The algorithms such as they are lend themselves to implementation with an AI technique but remains to be determined whether that would be overkill.
In any event and in actual fact the whole exercise is mainly intended to force me to learn something about CL before trading it in a cash account again
More cocooning with the 5 energies bot preliminaries since the last post, hopefully a few useful results to share as they become available.
What's interesting to me and hence what gives me the strength to persevere with what might seem arcane quant theory is that so far the tangents I take off on all seem to lead to the same ground. In that respect concepts I gained working some time ago with ARCH and GARCH models of financial time series in the R language pop up in the literature on the Hurst exponent (due to Harold Edwin Hurst (1880-1978) the British hydrologist and sometimes called the Hurst coefficient, which person not to be confused with J.M. Hurst, the stock market cycles researcher) and via the concepts of H.E. Hurst to concepts introduced by Mandelbrot (one and the same, who introduced fractals).
The commonality between ARCH, GARCH and the Hurst exponent is how far financial time series are random processes and whether & to what extent from time to time they may deviate from random processes, especially the extent to which a time series exhibits "memory" (i.e., the extent to which present events are a function of past events). For me (aside from determining my wife's mood) this translates immediately into a study of S/R levels. While it's one thing to know what S/R levels are and trade them manually, it turns out for me it's quite another to teach a bot to recognize them, and rules of thumb aside (e.g., "if it doesn't break after 3 attempts it's not going to"), especially to create an algorithm that returns the probability price will respect an S/R level.
The commonality between the Hurst exponent and the fractal dimension for the purpose here is defined mathematically as
D = 2 - H (equation 1)
where D is the fractal dimension and H is the Hurst exponent.
For completeness, in the limited reading I've done (reading list available but will have to make it available) at least one questionable source defines the relationship between fractal dimension and Hurst exponent as
D = 1 / H
just as some equally questionable sources derive fractal dimensions for financial time series as being < 1 (IMO this is impossible, signifying a result that despite its practical application to whatever the researchers are working on can't be defined as a fractal dimension as commonly accepted).
At the outset (once I got over the simple math relationship between the Hurst exponent and the fractal dimension) however it struck me the formula does not do justice to the practical difference between the 2 measures for hydrologists, financial market theoreticians and traders.
Traders toss the notion of fractal about as signifying a relative time frame, financial theoreticians consider the Hurst exponent a measure of market memory (or in terms of statistics, degree of heteroskedasticity) and hydrologists think of the Hurst exponent as a measure of indentedness or meandering (which calls to mind both the only paper by Einstein I read that made perfect sense, "On the Meandering of Rivers" and the coastline of Norway, still regarded by hydrologists as the best example of indentedness, claimed to be the invention of Slartibartfast in HitchHiker's Guide to the Galaxy, whom coincidentally I presently resemble.)
I plan to post some code snippets for utilities (e.g., like drawing stochastics based HH/HL/LL/LH lines on charts and 5 energies bot thought process indicators in MC PowerLanguage, which seems to me like pseudo language, so should be easy to transcribe to real programming languages) and the trading-platform-independent Visual Studio 2010 project files for the Hurst/Mandelbrot research once I get the bugs out of the VS 2010 project.
ETA: Between then and now we trade !
Last edited by bnichols; November 11th, 2012 at 09:31 PM.