## Introducing Asirikuy S&R: Finally True Algorithmic Support And Resistance! :o)

Those of you who have followed my blog for a while might know quite well that I have spent a lot of time during the past few years interested in the problem of support and resistance and how to algorithmically determine these levels without any human intervention or discretion. Last year this research led me into the creation of a prototype system in MT4 (which was discussed on a blog post here) but the calculation of S&R levels was very computationally expensive and not feasible in practice for the implementation of a truly robust automated trading strategy which could benefit from its output. During the past few months I have been going into this problem again and this effort has led me into the development of a new indicator (which I call Asirikuy S&R) which might be the basis for the creation of new algorithmic trading strategy based on support and resistance.

The issue has always been clear, there are a set of levels in the market traders react to (called support and resistance) which are difficult to define mathematically since the definition of the levels not only encompasses their location but also the level of reaction one is to expect from them. For example when you see that there was a significant high on the past few days you consider it an important hourly resistance but breaking it is bound to bring a much weaker continuation than the breakout of a much more congested weekly level that has been tested 6 times. In essence what I am attempting to convey is that support and resistance is not only about location but about the way in which traders perceive the different levels to have varying degrees of importance (this is why support and resistance is such a powerful trading technique).

Many attempts have been made in the past to define these levels mechanically but all fall short of coming up with some sort of way to address the importance problem. For example Dr. Mel Widner designed two indicators (WSO and WRO, covered on a previous blog post) which tell us how price is relative to recent support and resistance levels. This type of indicator addresses the location problem to some extent (by using high/low values) but the actual relevance of these levels is never calculated and this is precisely where this indicator falls terribly short. All approaches which have attempted to use fractals, high/low values, etc all attempt to define the location problem but fail to realize that S&R is not only about where the levels are but what the levels mean to traders.

In order to solve this problem I implemented a Delphi based DLL which used my previous ideas based on the counting the times in which levels have been tested through a defined period in the past. My idea here was to get the S&R relevance of any price level by location – within the past X bars (usually thousands) – how many high/low values have been within a +/- X pip range from this price level. In the end what we have is a normalized percentage value which tells me how many of this high/low values as a percentage of total bars have fallen right next to the price queued price. This way I can have a very clear idea of not only the relevance of this level as a support or resistance level but I can also get an idea of how important this level is relative to others. Certainly the amount of information conveyed by this indicator is great (and something which could never be easily seen manually) and its interpretation is extremely interesting.

The image above plots the S&R relevance of each bar’s closing price relative to the past 5000 bars on a 4H chart on the EUR/USD. You can see how different patterns emerge as the indicator starts to give information that other traders would see as well. For example a consolidation period is always established when we reach areas of high relevance and we can tell that consolidation is happening as relevance starts to increase in a upward slope. As soon as we have a breakout from this important consolidation zone the S&R relevance drops because the price levels reached after breakouts are not important S&R levels. Now you can see also how when relevance remains close to zero we have sharp moves which are ended when we start to see an increase in the relevance value (marking a consolidation zone). Consolidations that start from zero are new areas which haven’t been tested a lot in the past while consolidation zones that start from much higher are previously tested zones which are already very relevant.

The indicator gives us some information which cannot be found through a simple attempt at the determination of the S&R levels, we get a true picture of how relevant a given price level is relative to the “big picture”. In this way we can easily spot relevant congestion zones and we can also predict which breakouts are bound to be more meaningful than others. The indicator also allows us to know when a trend has slowed down (buildup of congestion) and therefore it allows us to build good exit strategies which cannot be built through simple chart observation (as very extensive observation would be needed to grasp all this information). The Asirikuy S&R indicator is therefore my first true practical achievement for the calculation of S&R levels which became possible thanks to the much faster speed of calculations within the Delphi DLL.

Certainly there are lots of ways in which this indicator can be improved and changed to make it even more powerful. For example the comparison of values against fractals instead of simple high/low values is bound to be a powerful addition (which will increase meaning power) and the ability to increase the look-back period further or to build a multi-timeframe indicator which can show the relevance of a level on several time frames is also a target in sight. Through the following few months I will start to develop algorithmic systems to use this indicator, probably our true mechanical and powerful approach at the trading of S&R from a completely automated perspective.

### 2 Responses to “Introducing Asirikuy S&R: Finally True Algorithmic Support And Resistance! :o)”

1. Mihaly says:

Daniel,

In my opinion the fractal and multi time frame option would be more desirable that look back for thousands of bars. In my mind the S&R levels are defined by people and based on recent events. I just don’t know why a 2 years old S&R level would be relevant today. But who knows, that is why testing is good.

There are two more options to consider:

Fib retracement levels often usable as S&R.

Trends have S&R, called channels and those are equally powerful, sometimes going on for years on the daily time frame and they are equally tradeable.

Just my 2c. Mihaly

2. Maxim says:

Daniel,

My sincere congratulations on your achievement!
It’s nice to see how ideas become reality!

Maxim