RFT Talks #1

I thought that now would be a good time to bring back RFT Talks as I had been quiet for the past week or so but that does not mean that I was not busy, as you may be able to see from the home page of the site

Ironically however I may not have time to record the talking part of the RFT talks be we will see

There has been a lot going on with the site, my programming and trading along with bringing in various technical upgrades that should enhance recording, images etc to raise the standard of the content produced

Content Production

Lets start with the content…….. a number of the videos I produced were well a bit rambling in places, some in fairness to me were build along videos but some weren’t. I also found that although video was an easy format for me that in order to follow along, especially if English is a second language that having the steps written down and interlaced with a couple of images makes it an easier format to access

I also found that by writing the posts and getting the structure in place that the videos had less erms, so’s and filling words, so it is probably for the best to change the workflow in this way. So the videos will be there but accompanied by more writing

Finding Content

An area that a couple of people had mentioned was that it was hard to know if you had seen all of the content on a particular topic so I have made a couple of changes to help with this. For one I am running back through the videos and adding more tags so that the posts on similar topics can be found more easily if for example you wish to see all the posts relating to linear regression, once the tagging is complete you will be able to find them all easily. A couple of areas such as this and some of the new pages will take a couple of days to finish off.

Secondly I have added a new menu item called CODE LABORATORIES AND TEACHING LABS under which there is an item called CONTENT MAP, when clicked it will take you into the classroom part of the site

The lessons are split into three new levels:

Foundations – this is content for members who sign up for a free membership covering getting started type areas of coding and ProRealTime

Builder – this is an intermediate level for those who have decided that they want to get started building their own algos with content covering basic coding, testing and portfolio construction

Developer – the developer level has the build along series, work in progress system files and the super in depth content that will bore anyone not truly interested in developing their own systems

When the Course is clicked it will open a curriculum screen

In this screen you will find all of the lessons for the appropriate level grouped by topic, along with various progress bars etc

At the moment when clicking into a lesson it will tend to have a link to post that will need to be read and perhaps an action to perform in the exercise sections

As new content is posted the courses will be updated so it may be that some never reach 100% but we will need to see

For the moment I do favour keeping the main site content as blog posts as this allows for discussion in the embedded forum posts and allows for tagging and more search functions than if the site only had courses

What embedded blog posts I hear you say? Each post will now have a forum thread automatically created and embedded at the bottom of the post, for more information check out this article

Membership Levels

You will have noticed above that there were three levels, foundation, builder and developer that are parallel to the original names of beginner, intermediate and developer

Using a trial for the developer membership though it has had success has been blighted by people who just want systems taking a trial, downloading all the systems and then leaving while people that may be very interested in coding saw a high barrier to entry as all the content was behind the pay wall

To address this I am going to make the content that was free anyway a membership called a Foundation membership with basic getting started content

In the past the intermediate and advanced content was all contained the in the developer membership but I will split this into two with those levels being Builder which will have more advanced tutorials and content and Developer which will be largely the same as it is now. My focus for the developer membership has not been to deliver finished system each month but to run through examples of building systems and using the RFT code lab to support your own code development. Indeed I was really happy to see this month that members began to share their own Phoenix systems and mods of systems from the site that we can now test. Those of you who followed through with the momentum article can see how well those systems (10 min and 2 Min) are performing at present

It can not be ignored that the majority of the people involved in ProRealTime are looking for prebuilt algos, myself and I know others who code their own systems will still look to other reliable systems for diversification if the price is right, by adding variations of systems that have been developed on walk throughs in the developer content and other systems to the ProRealCode market place and eventually on this site I can serve that market and create more revenue that ultimately is returned to producing content for members. That being said if the implementation of the builder level proves a success then this is much more preferable to renting code.

One of the goals that I would love to achieve this year is for a developer member to get a quality code through testing, get their own brand/mythical creature designation and being to sell their code or share it with other developer members. Ultimately most people sell code to increase their income to increase their account size allowing their algos to work harder. Having a store full of brands created with members would be a fantastic achievement

Changing Market Conditions

This is something that we had talked about in the would you run them and ProRealAlgos 2M Scalp Mod articles, the concept being that the days of trend following may be coming to a close. The Dax in particular has traded in a tight channel for the last few weeks that has limited opportunities, well I thought opportunities until I explored the Axia Futures (Main Site | Youtube) and Fractal Flow content (Main Site | Youtube). Both services cover trading of Futures with the Axia guys having six and seven figure days but in order to do this they are extremely precise about entries and exits, planning trades, looking at historical levels and minimising risk in the way that you would expect from trades dropping 50 full lots of contracts at once

Most of the trades taken are more scalp trades than longer term swing trades but the key is precision. The PRA Scalp mod to date has under performed and I feel the reason for this reflecting on the content from those sources is that you can’t have a precise exit and an imprecise entry and expect a positive result

If you look at the results from the indices algos that I have been tracking then the systems that have done really well this month that trade indices are those that have a tight stop and close at the end of the cash sessions. Those that are more trend following and particularly those entering positions on highs of days did not do so well. That being said a site member did send in a version of the Traderherz Sniper mod that was out performing both my mod and the original on the back test. One of the key features appeared to be that less trades were taken, it will be really interesting to see if that form continues in the coming weeks

This phenomenon was noted by ProRealAlgos (Code: PRTARB for 10% Discount) who have gone into their back catalogue and are releasing some systems that they must have previously not considered for publication, talking about the need for diversification of entry themes, timeframes and looking at systems with lower win rates but better gain to loss ratios. Some themes that have been discussed in a lot of detail in the developers articles for some time

The algos we’ve chosen to be a part of the ProRealAlgos Cuts collection follow a few characteristics. Here are the 2 main points. 

(1)

They differ from the two other algo types

We’ve felt the need to diversify our algo offer with something more than just trendfollowing algos, with medium to large stoplosses, and average gains smaller than the average losses.

The ProRealAlgos cuts algos all have small stoplosses, that often lead to smaller winrates, but instead have a bigger average gain than average loss. 

(2)

Pre-verified results

Over the years we’ve developed alot of strategies (ALOT!). Most strategies are discarded before even reaching the testing phase. Some algos make it to testing but fail there others are just forgotten in the long list of systems and ideas.

What we’ve done over the last couple of months is to go through every single algo we’ve ever developed. Even the forgotten ones. The ones that never left the idea stage. The two main points we’ve looked at is the performance of that algo since it was built and the algo performance in the OOS of the new historic data period (1 million bars). As expected the majority of the algos we tested didn’t qualify. That’s how it is and should be ! 

Out of the list of algos that qualified, we’ve cherry picked the best performing ones and the one that qualified as different from the two other algo types (as explained in point 1).

So think of ProRealAlgos cuts as a collection of algos based a number of different strategies, markets, timeframes that have proven results aswell as the characteristics described in point 1. They are number of different strategy types e.g. mean reversions, breakouts, trendfollowing and they all differ from one another. “

I think this is something that we are all guilty of and in some ways is a limitation of the data set to an extent. If a system is built particularly on a lower time frame then the time spent in market from the Covid bottom until now has largely been a nice uptrend and especially in the US markets any pull backs have largely been shallow, so it is possible to have an imprecise entry, a wider stop and generate great results. The question is what to do about it?

As a far as I can see there are a couple of options:

Diversification – this can mitigate risks but if we consider the premise of precise and imprecise systems then diversification and overcome some of the imprecision within a portfolio but overall the capital will not be used in an efficient manner

Optimisation – one thing that I did notice from the Mod of a Mod that Sean provided was that it was not that different from my mod, but it was different enough to remain profitable as my mod fell away. If you consider that the sniper mod was created in February then this does beg the question as to ‘can imprecise systems work in a short/medium time frame?’ This is perhaps an exercise to explore in the next couple of weeks is to revisit my version of the mod and see how a reoptimisation would compare to the prior version. From here the question then becomes ‘should we reoptimise  imprecise systems profitable or otherwise to better fit current market conditions?”

From there the next step would be to complete the work on ProRealTime AI and establish the set of parameters that would allow the system to self optimise without being over optimised

Precision – a final option is to improve precision. Is the problem facing several systems that they are taking trend following positions in markets that are not currently trending? Perhaps the imprecise entry can be used with a set of exit criteria that will reflect if the market is currently trending or not, or perhaps if the market is not trending this system should wait until those conditions return. This can be an unpopular option as it can leave months without returns but this is why we have different systems in the portfolio for different market conditions

Taking example from RFT systems:

The Pegasus system uses RSI Divergence for an entry signal, the stop is tight and the risk to reward ratio is high. What I saw when using the system was that there would usually be an entry when there was a gain in the EU session that pulled back before the US open, a positive open from the US then leads to great gains. This is precise but those conditions in which the EU is positive initially but then unsure going into the US entry are really sentiment driven, in some months those conditions occur often and strongly but for around half of the year they occur infrequently

The Manticore systems have a couple of variants (not tagged atm use manticore in the search box to find them all) but they all have the same flaw……. they need around three to four 2% plus days in the Dax in which it opens and moves in a set direction around 2%. When the system works it WORKS, when it doesn’t there is a slow grind of small gains turning into losses over and over again. This system is imprecise but in those market conditions in which impulse not precision if needed great things can happen

What To Do?

I do think that self optimising systems have potential along with continuing to use a portfolio approach in which not every system is expected to create the same return each month

The area to explore in more depth is precision, when you watch professional traders they do not say I want to get in here and then give it a 2% stop loss. They look at market structure, lows, highs, double tops, bottoms and trend lines. This would not necessarily mean new systems but potentially looking at current system and using new techniques to make them more precise

That being said one of the mantras from discretionary trading is that your ‘edge’ needs to evolve as the market changes so therefore it may be unreasonable to expect all systems to be fantastic for long periods

Build Depth Of Knowledge

Around two weeks ago I recorded a couple of longer videos looking at my vast collection of indicators and how successful I thought that they were when trying to use them for technical trading

Having finished the recordings I realised that most weren’t profitable but that did not seem right, why would people use these things if they were not useful? So I started to research each of the indicators in more depth which was a useful exercise but did produce conflicting results for example the head and shoulder pattern is primarily taught that the position is entered as the neck line is broken, however Fractal Flow takes his entries at the top of the right shoulder to give a better risk to reward ratio (Fractal Flow Awesome Graphic Patterns Video)

He also had some very interesting views on indicators, being a student of market structure he viewed a number of indicators as meaningless and peaked my interest in some others. The Linear Regression for example I thought of as another version of a moving average but it is a mathematical formula to find trends in chaotic results “a drunk man walking down the street will walk chaotically but you can nearly guarantee that although you can not know where his next step will be it will most likely be on the pavement”

Moving averages were seen as a away of mathematically explaining price action which makes it more simple to read but losing the important details. “Why should the 50 SMA matter, it gives no indication as to the underlying market structure”

I did explore a number of indicators that I would like to consolidate into an indicator that will produce a score showing the likelihood of the market turning but the core theme was that this professional trader was using Andrews Pitch Fork and other tools that I hadn’t heard of but need to. So lets get all of the available indicators, examine them closely and determine how each can be used to benefit our coding and systems

Revisiting Technical Trading

I had purchased yet another indicator largely out of interest (Volume Profiles) and what it and the research with Axia into how it would be used actually did peak my interest in technical trading again, well that and a guy on Twitter who keeps posting trades that I think could be coded into scalping algos but at this point I don’t want to pay £600 for his course

What I am interested in is using the indicators that I have and the teaching resources that I have purchased in the past correctly. What I don’t want is to be sat watching the 1 minute Dax all day, I code algos for those types of trades so that my time is not wasted.

The harmonic and some of the graphical patterns on time frames of 4 hours to 1 day are of interest but they do need to be combined with a bit of fundamentals and knowledge of news, the Fed speech on Friday being a good example. I saw manual trades that were stopped out on ATR stops, algos that got caught in the chop and all the while I wondered if any of them should have been in the market at the time. For most major news events there is a period of the market range consolidating and removing risk. When the news comes out there is often an impulse move against the news in a sort of short squeeze move, then a larger impulse in the direction that will be come trend followed by profit taking at the next high for the short term money before we move back to trend. Taking some of the premise above do we have risk in the market with an imprecise system in the hope that we can ride out the move? Do we close the positions until the news is over and potentially miss trades/close out swing trades causing losses, what is the best way?

Building Confidence In Results

One of the tasks that I did have following the videos that I had recorded was to perform some forward testing of the indicators as they would generate signals and then disappear as the patterns extend to the point of breaking. What I will be doing for the next few weeks is posting potential trades into the RFT Pattern Trading forum  so that all of the patterns are saved along with my thoughts on profit zones etc and we can see if they can be profitable (Tracked Overall In Edgewonk)

Once rules are established it will likely be able to automate the trades, the one complexity that I do see is that for example on Friday the comments by Powell caused a move in the dollar which in reality gave opportunities across EUR/USD, GBP/USD, AUD/USD but some were better than others, the algos would not choose the optimal pair to take the position in

The new Edgewonk will allow for tracking multiple portfolios (RFT Alpha Portfolio | RFT Pattern Testing Portfolio | Profit Profit Signal Testing) of testing including testing of external signals and systems allowing for both a detailed post and an overall public profile that will update as I in put results

And Breathe…….

3371 words and finished for now, back to creating new posts and videos I hope you will enjoy the journey

Leave a Comment