Mission Statement

We are a group of stock traders and investors whose aim is to democratize day trading by providing an accessible consolidation of resources and human capital, which are represented by algorithmic charting and the processes that are entailed to develop them. The discipline we follow is that of “compound trading”, where our algorithms are used by our subscribers to perform small, low-risk trades, resulting in large gains over time.

This is achieved through live broadcasting, newsletters, alerts, and chat rooms, where members can discuss and view our team of developers refine and apply our algorithms in real-time.

Problem Space

Our lead trader has been involved with the markets for nearly three decades. Up until now, it has been widely held that equities related to crude oil do not have predictive models in place which have been used with any reliability. This project started in part out of personal need, and it became apparent over the past months that there was enough power in the following processes that they could be leveraged intraday.

Hence, we define our problem space in terms of representing otherwise complex equities for an audience which is largely composed of investors who hold career positions or are entering retirement, and whose interactions with the market run intraday.

Statistics

Since our inception, we have made it a point to achieve a high level of transparency without sacrificing the integrity of our paid services that that our subscribers are privy to. When we are able, our profits and losses are posted to a Google Sheet, which contains statistics for each equity that we trade in. Please email us directly if you are interested in viewing this ledger.

Charting: We keep a record of the charts produced by our algorithms on Trading View.

Traffic: Hundreds of unique impressions every day, with international readership represented by virtually every economic zone in the world. Since our formal site launch on October 23, this has amounted to 3,466 visitors in total.

Our subscribers come from Canada, the United States, Ireland, the United Kingdom, Sweden, Germany, the United Arab Emirates, and more.

Methodology

The endpoint of the algorithms are probabilities for price/time action, as outputs on a chart, which provide visual indicators for a trader to use when trying to discern market direction. It is critical to note that these algorithms are methodically generated by hand, and that “formula” involves a complex mix of:

  1. Price and trading history.
  2. Traditional charting indicators.
  3. Historical volatility.
  4. Nature of price actions.
  5. Algorithmic sorting and matching calculation (Gale and Shipley, 2012).
  6. Fundamental heuristics (“rules of thumb”).
  7. Decision tree algorithms and disciplines such as Fibonacci numbers, Euclidean algorithm (a chart is considered a 2D space).

While these do fall under well-known methodologies used by the trading community, the variation on the theme here involves an assumption: each commodity or equity has a “natural state” (in terms of its random time series). If a trader can extract and identify “anomalies”, the algorithm – or process – which combines (1) to (7), will generate targets as shown. So, the problem space begins shifting to an intersection between geometry and probability. As we continue proving this process in a live trading environment, which has hereto been “pen-to-paper”, the natural step is to script the sub-processes into a computer algorithm. Thus, the “sharing economy” aspect of our business model comes into play.

Preamble

The following sections divide, in chronological order, our algorithms being applied across every equity that our day traders target. The format of these calls are generally structured as follows:

  1. Probability indicators are calculated and posted for some x-number of days according to the next time/price cycle.
  2. Fundamental trends are highlighted as necessary, with recalculations posted.
  3. Geopolitical or major market breakthroughs are tracked and highlighted.
  4. Charting is shared leading up to target hits.
  5. The final posts for a given cycle are either a target hit or a miss, and are logged over social media/back-end spreadsheet.

I. EPIC the Oil Algorithm

The development process for EPIC began with an extended proof-of-concept phase over the summer of 2016, with real-time calls being published to Twitter from August onwards. These calls included both successes and failures. In each instance, the following segments are time-stamped in sequence, showing alpha targets (red circles) being predicted by the algorithm, and then follow-up charting showing those targets being hit. Over time, the geometric processes being applied become more refined, as evidenced by how “noisy” the charting becomes. Beginning on August 3, 2016, the first set of posts our lead trader made involved the idea that a rally would occur if $USOIL hit 41.20. A series of tweets were made leading up to 41.20 as alarms for EPIC were going off.

Below is a photograph of the standard charting used for the buy trigger call. The blue arrow on the bottom left shows the alpha target being hit, which had been predicted by EPIC, and the region where crude oil had been trading at the time. At this point, the target had been hit at 7:00 PM EST that previous evening, so our lead trader watched (we do not trade extended hours or futures), waiting for the back-test to fill at 41.20. This happened at 11:00 AM EST the next day. The next blue arrow in this time series points to a significant spike, when 41.21 was finally filled at 1:00 PM EST. From thereon, we considered the first round of calls made by EPIC to be successful, and formally launched a proof-of-concept phase.

wtibacktests

To further elucidate the significance of this call, it should be noticed that crude oil was in an undeniable downtrend, and conventional knowledge argues that a rally was not to be expected.

widevew-aug3_2016

Refer to this chart below, zoomed out previous to these calls being made. On June 8, 2016, crude oil was trading in the 51.50 range and had been in a downtrend leading up to August. In the time-frame following from August up to mid-October, crude oil began trending like so, following from the initial calls above:

wideview-detailed

The following sections continue providing a sequential series of charting and calls (written as posts over social media) as evidence of the algorithm’s predictions. It is meant to be read in order, highlighting calls often made days or even weeks in advance, with follow-up charting showing those predictions being hit. On August 5, crude hit the 41.20 area on a re-test and bounced back again, confirming the targets the algorithm honed in on.

At this point, I indirectly started warning traders in short positions, and reiterated that it was safe to buy oil on every dip until the election. Still skeptical, I erred on the side of caution until more testing could be done.

In the charting below, this re-testing at 41.20 can be visualized at its constituent blue arrow, with four consecutive targets being hit. The next warning I put out over social media was an important test at 42.20. Once 42.20 hit, crude took off to the next level of pricing.

$WTI Crude Price Action

Shifting away from equities such as $USOIL, there are also several examples where I am executing these types of trades on other ETF’s. Specifically, I was looking at oil companies that are squeezing or reacting well to crude oil’s price actions. Here is an example from my feed with reference to $REN.

longren-1337_rising

For context, the next example elucidates a case where oil had run for a number of days beyond what the algorithm had called, and almost every trader was saying it was going to rest or turn right away (implying that you should exit your long positions) at 43.75 resistance, or at latest, before 45.70 major resistance. While this play would make sense with typical charting scenarios, EPIC had a different idea, and the process signaled that the resistance points would not be of concern. These levels were put out to the public, and each one hit almost to the penny.

crude_targethit

The first target hit 44.45 and runs from 43.27, then blows through resistance 43.70 and hits 44.55.

crude_secondtargethit

The second target then hits at 45.10.

crude_thirdtargethit

After that, the third target hit 46.10, and then oil comes off a bit and “rests” for a time.

The following are more notices I put out warning traders about which support levels could potentially be trusted or not. After support was reached, a rally restarted.

crude_testsupport

Testing support of target at 44.55.

Snipes Using EPIC in Spiking Scenarios

 

crude_exactalphahit

Time-price target by EPIC helped discern a spike at 12:00 PM EST for this ETF.

As the month went on, I began tweeting more alpha targets, down to exact times and prices, as there had been enough live trades performed where they were being hit.

crude_pivottest
What this case highlighted was the algorithm’s ability to also demonstrate pivots that traditional charting would not be able to extract. These pivots are instrumental in this particularly compound trading discipline, since they are what ultimately allow for predictable, low-risk gains over time, by performing small trades. The next level shows a target at 48.30 which would give crude a downdraft.


crudeintrapivot
This next chart is an example where I also started sending out tweets about the intraday pivot which the algorithm found during the downdraft, but traditional charting could not. Exploiting these pivots and spike in crude oil allowed me to trade $UWTI.

wticlimbing

The following charts are more examples of targets being hit, as generated by EPIC. wtitargetone wtitargettwo wtitargetthree

By the end of August, I was reporting 82 wins and 16 losses using the algorithm.

At the beginning of September, I was finding more success using the algorithm indicators in other oil-related equities. The following is another example intraday with $REN.

renquicktrade

As time went on, and as we have mentioned earlier, the process to generate algorithmic charting started to become a lot more refined – and thus had a trade off with how “noisy” the charting became. The following is an example on September 5, 2016, where the algorithm’s target hits at 44.55 on $WTI. After squeezing and going on a rally, it blew through two strong trend-line resistance points (thick blue lines). When it hit the next upside trend-line (in blue again), it paused, took off, and went through two previous resistance zones (thick purple).

wtiaggressive

Using these contexts as an instrument, I could then anticipate this spike in $REN:

renpop

renpop2

The pattern continued throughout the month, and the following sections show more targets being called and published in mid-September. By this time, they have been hitting much more consistently, and with higher frequency, than they had in August.

In the next image, the blue arrows show where a call was made and hit in $WTI. The pink lines are for the quadrant work we had been implementing into the charting as a tweak to increase our intraday wins. Notice how the price action gets aggressive as the time/price cycle got close to the time of the target and how the price of crude aggressively dropped to perfectly intersect with the target zone the intersecting lines referred to by the blue arrow).

wtiquadhit

Intraday Edge

The following is an example of how EPIC can be used as an intraday trading edge, which has been the purpose, and as we have found, the best way to use the signals generated by the algorithm.

Below is the result of this call. Keep in mind that there are around 30 or more tweets on EPIC’s feed between the call and this post detailing every step of the rally.

After making this call, I then began sending out alerts that the price of oil was going to turn down. The following set of posts show what crude oil did next.

Target Hit:

Here is an example of a cluster of alpha algorithm targets that converged to one time and price target, and then split out from there, causing complications.

There have also been (and continue to be) precise situations where our algorithm has been able to predict a larger series of convergence and divergence points intraday.

Evidence to Date

Compound Trading’s pre-launch phase started mid-October, and ran until the beginning of December, when we formally made our services semi-exclusive (meaning interested users are still free to watch for a period). The following examples are prominent pieces of charting, intraday, which are meant to demonstrate the improving accuracy of EPIC as time goes on. As per usual, commentary about cycle resets and geo-events precede the posts which have been highlighted here.

October 20, 2016:

The price of crude on September 22nd was meant to rally from 43.70 to 51.53 on $USOIL, but it was a major time/price cycle termination instead.

In the following chart, we see a direct target hit on October 19th, and the termination continued in advance of October 20th. The termination continued through to the next lower two targets on the right-hand side on October 21st, with the price cycling through that quadrant. October 21, 2016: After EPIC’s call for 1:00 PM EST had been reached (the lower quadrant of the targets from October 20), the rig count for that day pushed the probability that the lower target would be reached, and there was another direct hit.

In cases such as these, however, we are conservative in calling these “ballpark” oscillations direct hits:

October 24, 2016:

In-between target calculations, this example highlights the use of traditional indicators (such as the blue trend-line) in the algorithm, with crude bouncing off of one at 8:44 AM EST.

Following this trend, I made a call at 12:06 PM EST that crude would need to hold support to operate within its quadrant for that cycle. However, crude lost critical support (in thick blue) by 12:19 PM. For the following day, this signaled a new set of anomalies, and hence a type of checkpoint where I typically send out new probabilities for algorithm targets during that week. October 25, 2016:

New targets:

That day, however, was a target miss. Revisions were made for the next trading day.

October 26, 2016:

To make up for the mistake from the 25th, it was important to highlight how critical traditional indicators are.

After readjusting, there was a near-direct hit, with signs that the market direction was going to be dancing about support at 49.27.

October 27, 2016:

At this point, a diagonal trend-line was forming support, and provided a good buy trigger. The following is a worksheet which shows the price moving to the upper segment of its quadrant after hitting a lower target at 10:30 AM EST on October 26th,

October 28, 2016:

Following its trend from yesterday, the price began trading between two alpha targets, and poised to hit the lower one at 1:00 PM EST.

The next posts show a play-by-play right before a downdraft made a direct hit on the lower target zone, which had been calculated days in advance. Other featured posts include price corrections, and affirming my buy trigger after the target hit.

October 31, 2016:

By this stage, alpha targets were being calculated weekly, positioned as outputs for time/price actions on Tuesday, Wednesday, and Friday.

Whenever recalculations are made, these are reiterated throughout the evening following worksheet adjustments.

The resulting chart appears as follows:

November 2 to 4, 2016:

Getting into the trading week, crude broke resistance and broke through between the upper target zones on November 1st, and the following shows a price action right before a target that had been calculated for November 2nd.

Here is a direct hit for the price/time at 10:30 AM EST on November 2nd, with another price action moving towards the last target for November 4th.

November 8, 2016:

Target hits, predicted by new calculations:

Perfect Hit, November 9, 2016:

November 10, 2016:

Between November 10th and the weekend, I was waiting for the price action to be accepted or rejected along one of the lines in thick blue (resistance). This happened on November 13th:

One of the more idiomatic features of EPIC is its ability to leverage traditional indicators, such as trend-lines. In this case, I highlighted the channels or trends in which price was moving, showing what one of my worksheets looks like before layering some more “complex” facets on top of this charting.

November 25, 2016:

Direct target hits:

Smaller time-frame analyses:

As time went on, there were contexts in which different types of traditional indicators were required to explain variance away from some of the algorithm targets (in the case of misses). This is an example of an “optimization problem”, whereby an individual applying an algorithmic process has to know when to react to a changing condition in terms of applying the correct indicator.

After considering Fibonacci support levels, the recalculated targets made direct hits.

After exiting its position, the price of crude went from a direct target hit into a breakout.

This worksheet highlights a support test right before price lead into another direct target hit.

By the time these calls were being published, Fibonacci levels were now being considered in terms of re-weighting future targets.

This provided deeper precision with newly layered indicators.

Exact target hit, to the penny:

Overnight positions, showing price action towards target.

Closeup of the same price action.

Target Hit:

Another example of overnight price action heading towards target.

II. ROSIE the Gold Algorithm

Our second most popular, and more fleshed-out offering, is ROSIE, which targets equities related to gold. Our lead developer made the decision to experiment with similar processes found in EPIC at the beginning of October, looking for the same type of edge that he had found using algorithmic charting in oil equities.

As with Section I, the following posts come in the form of tweets, with photos attached pointing to major price actions. However, we elected to highlight some of the more significant target hits here, rather than providing a completely detailed survey as above. Our aim, instead, is to show how multiple layers of complexity were achieved from simple premises as this algorithm underwent development.

In the early development phase, ROSIE only looked at traditional probability models to make trading decisions:


Note in the following example the type of interaction we have seen in terms of other day traders contributing to the development process.


As with EPIC, ROSIE also leverages fundamental charting techniques as part of its process, such as support/resistance and trend lines.

In the following representations, a grey box is used to show a price target generated by the algorithm.

While not as “noisy” as EPIC, the algorithm generates convergence patterns which are similar to those found in oil equities.

Another target hit after re-testing:


After several weeks, Fibonacci levels were layered on top of traditional probability models:


While at this resolution, the “target” may resemble something akin to a trend-line, looking at the charting generated by this algorithm shows that the type of “quadrant” that the price is trading in puts this set of equities and their related instruments into a different type of geometry. Hence, while there are similarities in the design process, this is an example of how new algorithms can be formed using others as their basis.

The post below is a chart which extends back to July to elucidate some of the long-term testing performed to generate new target zones.


At a lower resolution, we can see the price of gold fluctuating at the top of an algorithm quadrant.


Instances such as these represent indicators for possible breakouts, which are sent out as alerts intraday and generated days in advance with ROSIE.


Part of this process also involves tracking other indicators, such as USD/JPY, to anticipate bottoms. Notice how some of the probability levels and geometries involved in this following chart resemble many of the heuristics which have been applied to EPIC, as well.


Based on the results from USD/JPY, users were alerted about a potential bottoming out as we watched the behaviour of price over the weekend. By December 5th, the warning was formalized by using the lower level of the algorithm quadrant on $GOLD as an indicator that price would bottom out.


An exact price-time target was hit, which had been called in July:


This was supported by historical data, combined with the aforementioned traditional indicators applied against it:


The next set of posts highlight a confirmation for going long on $GOLD, based on a breakthrough with USD/JPY. Our lead trader continued tracking USD/JPY as it bounced off a Fibonacci level, signaling broken support through gold’s algorithm quadrant:


The breakthrough in USD/JPY indicated a 90% probability for gold to hit 1133.00, as generated by ROSIE.


Direct Target Hit:


Finally, ROSIE also takes into considerations traditional setups such as the “golden cross”:

III. SUPERNOVA the Silver Algorithm

This algorithm targets equities related to silver. The following pieces of charting highlight the types of heuristics involved in the algorithm, which also takes into account indicators such as Fibonacci levels.

Building off of ROSIE, where we have seen new indicators being layered onto the algorithm as needed, SUPERNOVA instead leveraged these layers from the moment its development started.

We started with looking at floors and ceilings.


Using $USLV as an instrument to affirm when $SLV would bottom out, we were then able to predict a price target, as shown in the arrow in the second post below. Note how the price of silver then traverses to the upper segment of an algorithm quadrant, which has also been used as an indicator to determine when to enter long positions.


As with our other algorithms, fundamental technical analysis involving factors such as moving averages have also been highlighted:


As usual, the following chart shows a consolidation of these layers into an algorithmically-generated quadrant by which our traders are able to determine price actions moving towards targets that we able to predict days or weeks in advance:


Using historical data, we are also able to demonstrate market direction using components such as moving averages and previous target hits.


The next examples show price actions similar to ROSIE in which silver bounces off a target price and oscillates within the range of the bottom Fibonacci level:


At the resolution of 15 minute time-frames, we can then surmise what levels silver has to hold in terms of price before it can enter a breakout.


The following segment of posts show our lead trader tracing the behaviour of silver’s price as it defends a Fibonacci level up until our predicted price/time target is hit, which had been predicted in July.

The next set of algorithms are still in early development, but largely combine notions such as converging/diverging series and Fibonacci levels, with fundamental variables such as moving averages.

IV. $DXY – US DOLLAR INDEX

The following image shows a basic notion involving price converge, highlighting self-similarity in the US dollar index.

While we have found some relations between these convergent series and the price actions of USD/JPY, our processes thus far have kept in line with traditional charting practices:

V. FREEDOM the S&P500 Algorithm
Observations regarding the S&P500 have been intermittent, although our process has generated important alpha pivots since we began publishing FREEDOM’s calls:


Fibonacci levels have also been used to refine our calculations with the quadrant-based charting that has been seen previously.


This has, so far, resulted in charting such as the image below, where the price touched the top of a quadrant, and then fluctuated within the target zone. Our member buy alert was sent out corresponding to the white arrow.

VI. VEXATIOUS the Volatility Index Algorithm
This is the youngest algorithm in development, and we describe its targets in terms of volatility clusters, meant to be used intraday.

So far, we have kept the charting simple, and put out calls in terms of the likelihood that a specific day will be volatile. For example: