February 22, 2019

RE: EPIC Crude Oil Algorithm Machine Trading Software Advisory Specific to Trade Frequency Protocol “Throttle” and IDENT Program Description.

We are now near two months of running the machine trading software for crude oil futures contracts (CL).

During the testing phase, which will continue for some time as we adjust code instructions, the execution of trades by the program is “throttled”. Meaning specifically that the frequency of trade was specifically limited to the highest and then was adjusted to a higher win probability threshold.

The result of the initial testing achieved near 100% win side trade accuracy, but less than optimum trade frequency. Increased frequency may (will in our estimation) return a higher ROI – assuming the win rate percentage achieved is high enough. This of course is a complicated calculation within the code that reflects the win side average per trade return vs. loss side average per trade. In short, the loss side is programmed to be less (limited by way of higher frequency executions – tight stop triggers) and when trade is on win side the trade profit is increased via trade size that is progressively trimmed as the trade is in progress. Refer to private Discord oil trade chat server for real-time discussion between developers and traders for more detail.

An article is available at this link that displays some of (actionable by a human trader executing trades manually) the oil machine trading results, much of which was throttled considerably and much of which had human intervention – in other words, had the software been released to execute totally autonomous the returns would have been considerably higher – but we are testing. The highlighted trades returned a 63% increase on the “large account” test for the one month duration. This achievement was specifically to the alerted trades, not the higher frequency machine trades.

The “alerted trades” meaning that which could be considered actionable alerts to our subscribers. The machine software executed many times more trades but our current alert system platform (Twitter private member feed, Discord private chat server, Oil trading room live broadcast) cannot for the most part distribute alerts fast enough for the higher frequency trades to be considered actionable by a human executing trades manually (a trader digital platform is on the team WIP to remedy this). In consideration also is that the higher frequency trade protocols could easily be reverse engineered to expose the proprietary protocols under our IDENT program – this remains a discussion point internally and how the higher frequency trades will be shared is in question (more on that at a later date).

The “throttle” was initially set to approximately 20 x and over the course of fifty days progressively lowered to function near 10 x with less and less human intervention along the way. 20 x for example would result in twenty times less trade frequency than would otherwise be if the software was not “throttled” at all.

Today (Feb 22 at 2:11 AM EST) the code was adjusted considerably to be “throttled” to be less than 10x and will be lowered progressively over the next 7 trade days. The win rate vs. return as it would be calculated over a month is the achievement bar (goal) in focus. More on this objective and other clarification in near future updates.

Market condition will also result in variance of execution frequency as will holiday weeks specific to the model(s) divergence.

The main takeaway: In to next week the frequency of trade will be considerably higher with an objective being to find the most optimum throttle setting to achieve the highest return. The win rate is expected to near 80% and not near 100% and the return on equity on a monthly basis to increase considerably.

The 63% monthly return (monthly return in this instance meaning account equity size increase as it relates to alerted trades only) is a favorable start, however, our team believes 100% + return per month is consistently attainable (on average over a year) and in a perfect machine executed world 300 – 500% being optimally possible. For now our objectives are to achieve consistent wins at higher than 80% with a frequency of about 60 to 120 trades per month with a return averaging 100% per month (the bar).

The IDENT program is a protocol specifically to order flow identification of market participants achieved by way of historical pattern recognition of between 20 – 40 entities that we consider largest and approximately 200 entities that we consider important enough to attempt to track. The entities are prioritized in what we describe as an “alpha” order. The IDENT program is in large part the topic of this recent article at this link that describes the influence of machine trade in the crude oil trade market as experienced by our lead trader and is in large part the reason for the “intuitive like” nature of our software protocol.

The IDENT program seeks to enter trade direction with prioritized alpha order flow and exit in the same fashion. It is a proprietary process and the instruction set within the code architecture will in large part remain private.

For more information on how our development has progressed, refer to this link that will immerse you in a series of articles written from first hand perspective of the day to day trading of our lead trader with crude oil futures.

Thank you.

 

 

 

 

 

 


I Have Been Down the Rabbit Hole. 

I know my journey (of the last two and a half years) few have traveled. I know this to be so, like I know I breathe.

I didn’t know before I went down that hole, but I know now – how much I did not know before I went down that hole. And it’s huge. I knew nothing. Few know anything.

I can’t imagine many other disciplines in the world as off-side as this. Nothing, and I mean nothing in regular market banter, conventional stock exchange media or social media ever deals with the reality of what is down that hole. They don’t even know there is a hole (most).

Why is this topic of any importance to a trader or investor?

The topic? A human trader cannot beat real machine learning software and that will matter before the humans know it mattered.

A human trader cannot beat real machine learning software and

that will matter before the humans know it mattered.

Assuming the 80/20 rule applies to the markets, then 20 percent win 80 percent of the trades and reap 80% of the reward (ROE). I don’t know what the real numbers are – it was likely closer to 90/10 before machine learning entered the public stock markets. I would venture to guess it’s 97/3 or so now. But I don’t know.

The problem looks like this… what happens when the machines (run by less than .001% – a guess) garner 90% or more of the market return? Is that possible? I know it is, I know it may be true now, and if not it isn’t far away. What percentage of current liquidity is machine learning trade driven? What percentage of that is on the win side?

What does that look like in the future? How far away is that future? I will bet that future is here now and it has enormous effect on your earning potential (as a human trader). And I will put forward right here that it will end the way humans trade in global markets about five years out. It already has, the humans just haven’t accepted reality.

So why then is this so important? We’re here to win. We trade to win. To earn. To see return on equity. Return on time. The path forward is critical for any trader that derives their income from trading.

Early adopters will win. They will win because they will develop relationships with firms that will provide solutions. They will win because they will adopt the technology, relationships, financial rewards before the technology is out of reach (financially speaking). Or perhaps, never available at certain levels of society. Never available is more likely.

Early adopters will win.

The market is already littered with technology built on poor science. It simply doesn’t work. But the firms that have figured it out… have really figured it out. How long does it take them to hit market capacity? Will they share the technology? I say no, they won’t be sharing.

There are many questions.

At minimum, it’s prudent in my view for a trader to cozy up to the developers. This isn’t a sales pitch, we’re beyond selling anyone. We’ve completed our first machine learning software – we don’t need to sell you. So why write this article? Because I want to share something I’ve learned in the process of development that I think is of critical importance to anyone that has followed our journey. I feel a responsibility to share. Share what  you ask?

You can’t beat the machines. It’s not possible. And I know. I’m a damn good trader and I can’t even beat our first generation software with crude oil futures trade. I don’t think it has lost a trade this month (maybe it has but I don’t think so) and we’re near three weeks in to the month. And it’s not that it hasn’t lost, it is more about how it wins.

It knows before I do. It enters before I do. It trims in to positions and exits when I wouldn’t. It knows things I can’t know, I can’t see. It sees every line on every book in the library instantly, while I search for the right book on the shelf.

It sees every line on every book in the library instantly,

while I search for the right book on the shelf.

How you ask? It has intuitive like capabilities (as I do) but it can process the decisions (as if intuitively) many times faster than I. It has systematic approaches to trade (as do I) but it can process the decisions on thirteen time-frames considering the historical structure of the financial instrument and how each time-frame relates to the next and which structure or time-frame should trump various trade execution decisions (the rule-set).

It is simply faster. It can process 8000 rules and how each rule relates to the next in each of the structures on each time-frame instantaneously. That instantaneous decision would take me at least 12 months full time of charting, historical back testing and deep thought to conclude that one decision. The machine executed on the decision and left the trade behind before I glanced over and seen it was over. I didn’t have time to acknowledge it was leaving the scene.

It’s not only faster, but precise. It enters and exits with absolute precision.

Yesterday (President’s Day holiday) it executed on one trade. The biggest move of the day on a slow day and it executed before the move happened. It shorted oil, I sat watching thinking why the heck is it short here? Tick tock tick tock boom, oil dropped about 37 points near instant, on a low range boring holiday trading day. And it covered in a flash before I could process why exiting the trade at that juncture of trade on the chart was valid. It is fast and it acts as if it is intuitive. And it is only first generation software.

Here’s a post that shows the trade;

A trade or two prior to yesterday’s was the same way. I couldn’t believe what I was watching.

Here’s another example, it knows over and over again where the real move in a time frame is before the move.

And an example of how precise it is in comparison to my personal trade executions;

It’s first month (January) it rang up 63% in oil trade account gains and it was yoked to a human at all times that throttled its executions by about 10 to 1. What am I saying? It could have traded up to 10x the return (assuming the same win rate and ROE on each trade would have transpired had it been non-yolked).

It’s first month (January) it rang up 63% in account gains.

Over the last few days we’ve allowed it autonomy – to a point. It can execute with autonomy when it is executing but we still have it throttled to about 5 to 1, this will be slowly adjusted / released. This week is a holiday week so there is issue in the structure of the models so it will be a slower than usual machine trading week for our model, but next week will be a mad house – a slaughter house. I’m not exaggerating. It doesn’t lose. Okay, it does – maybe, but rarely and for next to nothing for loss when and if it does. If it’s wrong… its out and fast.

If you think this is exaggeration, visit our public facing Discord trading chat room (click here) and randomly ask anyone to step forward and tell you I’m wrong (members I am referring to). Or check out the alert feed yourself. I win around 90% of my personally executed trades in oil (yes documented), it (the machine learning software) wins something nearing 100%. But its wins are the real meat of each trade. My trades are choppy. It harvests the move in a way I cannot. The compound return effect on the difference is astronomical. Click here for a recent article I posted about compound trading gains in oil trade.

We have been working day and night for over two years to develop machine trading software that really works. Why do I say “that really works?” Because most is garbage. Below are the reasons why most algorithmic trading is a waste (how ours was developed) and why you the human trader cannot beat real machine trading execution in the stock market.

Take four people (average at any given time in the development process), have them work 60 – 80 hours a week for 2.5 years. That’s about 14,000 hours. If the methodology used is right you then have at least another 14,000 hours to refine the software rule-set along with constant updates etc. We have now completed the first 14,000 hours. What has that experience taught me? Primarily that a human trader will never beat machine software. And that doesn’t consider machine software that becomes intuitive like – AI, Artificial Intelligence.

Here are just a few (and I mean a few) reasons why you will never out trade the machine – specific to the methodology of development we used to develop our first machine trading software for crude oil futures contracts CL. Yes, I am going to tell you how we did it. Why? Because I won’t tell you how our intuitive development is implemented, that is proprietary and always will be. Here’s the nuts and bolts:

  1. ROI – First month the machine garnered a 63% increase in trade account size and that was only the actionable oil trade alerts. Not the machine HFT returns. This means the software executed many more trades than what was alerted. We can only alert what a human can type fast enough to alert that could be actionable for our membership. Next on our list is a real-time feed for our members (automated alerts). In other words, the machine gained much more than 63% on its trading account, but that is private and always will be. What is public is what we alert that is actionable by a human trader executing trades manually. We are judged on what data can deliver as actionable to our clients.
  2. INSTANT EXECUTION OF DEEP TECHNICAL KNOWLEDGE – Take thirteen time-frames of crude oil (the charts) and find the structure of the financial instrument on each time frame (this is not systematic machine trading, this is an intuitive like process). Not to mention the time involved and cost. The human can’t process decisions on thirteen time-frames (both systematic and intuitive like and as they relate to each other, as described earlier in this article).
  3. INTUITIVE EXECUTION, INSTANT – Each time frame structure as it relates to each other. Continuing specifically the intuitive like component of point 2 above… imagine constructing the models for the structure of the financial trading instrument on thirteen time frames, that in of itself is a massive undertaking. Then being able to almost intuitively determine how intra-day trade relates to each and which trumps the other. This is massive.
  4. FAST EXECUTION. The machine can out execute any human with orders in and out and trailing and on and on. Complex structures of entries and exits and more instantaneously. This is critical for return on each trade.
  5. PRECISE EXECUTION. Have you ever watched crude oil trade on the one minute chart? Precise execution with orders that change frequently in a flash of a second increases returns on each trade considerably. You can’t imagine the importance until you’ve been down the hole.

I am not speaking to;

I’m not speaking to conventional hedge fund robo adviser software. Most of it is junk designed to rob the masses taking advantage of the casino mind or lazy investor.

I’m not speaking to high frequency trading that leverages machine speed, order flow or execution locality. That is a form of HFT we have no interest in. We want transferable knowledge as it applies to the natural trading structure of the financial markets. Transferable in that the process used to derive the model for one can be applied to another.

I’m not speaking to run of the mill python code some random developed in his/her basement with 500.00 and an idea of how an instrument trades (the systematic process).

And finally, I am not speaking to the news oriented bots running software trades on intra-day media or geopolitical driven events.

The intuitive like component is artificial intelligence – machine deep structured learning.

The intuitive like component, beyond systematic machine trading is where the magic is, this is where the depth is, where the future is and the success of such an initiative lies. The intuitive like component is artificial intelligence – machine deep structured learning. It is (as it applies to where we are now) the early building blocks of autonomous machine learning trade.

Beyond systematic machine trading is where the magic is, this is where the depth is, where the future is and the success of such an initiative lies.

How did we get intuitive like software to work? What does intuitive like mean to us?

Here’s a glimpse… a real trader with decades of experience traded real-time for hundreds of hours live with software developers that extracted the intuitive like human understanding – real-time, asking questions, engaging in the why and how at each tick in the chart.

We lived together inside the trade, inside the natural trading structure of the financial instrument (crude oil) and then we replicated its nature in to the code. Only now… we have manifold times more horse-power.

We lived together inside the trade, inside the natural trading structure of the financial instrument (crude oil) and then we replicated its nature in to the code.

The new brain – the new trader, the intuitive like software… can execute hundreds of times faster on thousands of times the information second by second. Every tick on the chart is a complete new set of rules that need confronting, they need to be examined from every angle, back tested, related to each time frame, a plan derived and a trade has to executed right now. A series of right now. With precision.

The bullet hits before you hear it.

This is why the human will never beat (real) machine learning trade software. The only question that determines how good the software is, is how good the trader was that the developers used to extract from that created the entity (the software) and what process was used to do that. The cost? 14,000 – 28,000 hours at hundreds of dollars per hour – for a generation one package. And that is just the start.

It is a different world and we’re not by far the first out of the gate. But I know we’re running one of the better models out there. That I do know. Because it wins when it goes in to battle.

Any trader worth their salt will soon, if not already, understand they need a plan to engage this new frontier.

I also know that any trader worth their salt will soon, if not already, understand they need a plan to engage this new frontier. If you don’t you will regret not taking the time to forge your future. Now.

Best and peace,

Curt

Learn How to Day Trade Crude Oil Here:

Crude Oil Trading Academy : Learn to Trade Oil – real life articles / examples from an expert crude oil day trader.

Read Our Press Release About The SOVORON™ agreement (click here):

About SOVORON™

SOVORON™ offers a unique trading service providing clients with technology application based Machine Trade of personal investment accounts. To explore their unique client service structure click here http://sovoron.com/ or call 1-849-861-0697.

About Compound Trading Group.

Compound Trading Group is a market data provider to retail, commercial enterprise and institutional traders.

Trader Services – Compound Trading Group provides trader data services such as algorithm development, periodicals, trade coaching, trade alerts and live trading rooms. The near term objective is a digital dashboard environment for traders to fully engage machine trade and intelligent assisted trade data offered only at the commercial enterprise level.

Commercial Enterprise Services – Compound Trading Group develops financial instrument algorithm models and provides connect-ability to its proprietary AI Machine Trading software / data (representing various financial instruments) to private trade and institutional firms. Compound Trading Group also provides enterprise trader training, integration services, custom software / data constructs and strategic partner development opportunities.

Learn more at https://compoundtrading.com.