BRIEF OVERVIEW OF THE ALGORITHM

---

 In this section, I will try a simple explanation of the Alondra Union algorithm, referring to commodity price forecasts. 

I have published two very enjoyable books on Amazon, in Spanish.  

One deals with the static numerical series, of each of the species, whose hedging data, both futures and options on futures, is published weekly by the United States Futures and Trading Commission (CFTC), which includes some 265 species in total.   

It covers the sections of Agriculture, Energy, Metals, Currencies, Bonds, and Stock Market Indices. That is, there are 265 static series, one for each species.  

In the second book, I explain how a second dynamic numerical series is created for each species, based on the first static series, where an attempt is made to expose a possible price trend for each species.  Let us begin.  In general terms, and my personal opinion, there is a basic problem in the analysis of the evolution of commodity prices.   

They are analyzed as if they were stocks, with moving averages or oscillators, for example.  However, commodities are like icebergs. The important part is "invisible", and at great depth. These are hedges of futures and options on futures. 

Both impact the spot price of each species, sooner or later.  When I started this study, we had entered the strict quarantine stage in Argentina. It was March 2020.  Since we were all isolated, I wanted to dedicate that special time to studying the problem of forecasting commodity prices.  

After several computer simulations, I released a prototype in September 2020, with the static series, and a second prototype, in January 2021, which included the first static series, with a dynamic series attached.  

The static numerical series collects information on hedges, both in futures and options, published by the CFTC, and assigns an average weight to the three market players: Commercial, Non-Commercial, and Non-Reportable.  

The first, the “Commercials”, refers to the actors who own and work with physical merchandise.  They are the producers, collectors, food or energy factories, importers, exporters, mines, and central banks.  

The second is the "Non-Commercials", which are the intermediaries, such as brokerage houses, investment banks, and large financial investors. They are called “smart money”.  

The third, the “Non-Reportable”, are the retail investors who invest in hedges.  I attributed a specific weight to each one, within the general average of the positions, both buyers and sellers.   

From there comes a SINGLE static series, for each of the species.  In the analysis of the COT (Commitment of Traders), there is a biased analysis, for each one of the actors, which, in my opinion, adds confusion into the analysis, since the market is always one, regardless of the actors in play.  

This static numerical series, already in itself, provides very valuable information on the evolution of the net coverage of each species.   

If it is negative, it means that there are more selling hedges than buying hedges, and this usually predicts a drop in prices or at least a stabilization of prices. If there is net buying coverage, we can infer that prices will rise, since there is a tendency to scarcity of it.  

In the analysis of the static numerical series, I was able to verify that it is much easier to predict the evolution of the price of any species, when there is abundance, that is when the static series of the species under study constantly presents a net selling coverage.   

I labeled these species as type “N” (negative). They are corn, wheat, and soybean oil, for example.  In general, the accuracy of the forecast ranges between 87% and 92%, concerning the evolution of the price of the continuous future.  

On the other hand, when the static series presents positive net buying coverage for at least 12 weeks in a row, there is difficulty in predicting the price trend. They are called “P” (positive). This is the case with soybeans and dairy products, to name just a few examples.  In general, the accuracy of the forecast ranges between 72% and 78%, concerning the evolution of the price of the continuous future.  

The other very important point that arises, analyzing the trend of the static numerical series of each species, with the evolution of the prices of continuous futures, is the notion of the so-called “latency time”.  This means, for example, that a hedge buying peak takes a while to materialize in prices. In some cases, up to six and twelve months, depending on the species.  

However. I realized that further analysis is needed to refine the algorithm.  That is where the dynamic numerical series is born, the one that predicts the price, based on the first numerical static series.  

The basic idea is to analyze the maximum values ​​of buying hedges, and the maximum values ​​of selling hedges of each species, in the static series, to understand how far a net hedge can rise or fall, whether buying or selling.  The analysis period covers the year 2015. 

That is, I built each static series, with the data from that year, and from there I built each dynamic series, of each species, from the end of January 2021.  I used the simple harmonic oscillator mathematical model, which takes into account the fastest movements, when the net coverage is in the middle, of the highest bid and highest selling values.  Here an improved latency time arises, compared to that obtained with the static series. 

The latency times arising from the dynamic series range from one week to a maximum of six weeks, which speeds up the rotation of the invested capital.  This is in short, what is expressed in the algorithm. It is simple to understand. But it aspires no more than to give an impartial opinion, in mathematical language, of the price trends of each of the species.

----------------------------------------

 Source: Commodities Futures Trading Commission (CFTC) 

  Tags: Algoritms, Forcast Prices, COT 

• Commodity Name: Name of commodity as CFTC published it CFTC 
• Contract Number: Number of contract as CFTC published it 
• Algoritmic Number :It is designed with a main focus on Commercials, being the Non-Commercial and Non-reportable ones of lower weight. If it is negative, it means that there are an abundance of merchandises, and if it is positive, there are beginning to be shortages. 
• Algoritmic Number Date Processing: Date of last CFTC publishing.
• Latency Time In Weeks: It is the number of weeks, in which the algorithmic number of net coverage used, to return to the same value. Consequently, it is the period covered by the analysis of the algorithmic series. The longer the latency time, the more reliable the forecast will be, since the variations in the coverage are small, week by week, and thus it is greater the amount of algorithmic numbers analyzed. 
• Latency time vis of the price forecast, in weeks: It is the number of weeks, before the trend observed in the numerical series of the price forecast, materializes in the future price. It is calculated, based on the maximum correlation between this series of forecasts prices, against the series of future prices. It is recalculated periodically, every six months, since it is a complex process, and it always needs a lot of new data processed, to have a good degree of reliability. 
• Peak Value Buyer Coverage: It is the maximum value of the algorithm, measured since January 2015. 
• Peak Value Buyer Coverage Date Date: it happened Peak Value Seller Coverage It is the minimum value of the algorithm, measured since January 2015. 
• Peak Value Seller Coverage Date: Date it happened 
• Forecast Up / Down: It is an index of the forecast of rises and falls in future prices, in percentage values, until the next publication of the CFTC. Normally this factor anticipates, between 1 and 10 weeks in advance, with an accuracy of at least 75%, future price movements. Observe carefully, the trend of the numerical series of the price forecast, for each commodity,  in order to be able to anticipate the movement of the future price, of each one of them. 
• Type of Forecast: If it is "N" it means negative, that is to say that on average there is never a shortage of merchandise. The opposite is "P". In this case, it is convenient to open future contracts, taking opposite protection in options. The forecasts are better when a commodity is an "N" type, and has a latency value greater than 54 weeks, since it has slight variations from week to week in its coverage. In the case of "P" type, it is convenient to open future contracts, taking opposite protection with options contracts.  

----------------------------------------

 Source: Alondra Union Books

  Tags: Glossary, Forecast

This information is a commodity futures price forecast, similar to a weather forecast.

The importance of this is that we can trust how the conditions of temperature, humidity, and clarity of the day will be, within the next seven days.

Here we try to have a clear trend of future prices, within the next four to twelve weeks in advance.

What you have to do is something simple: you have to look for the commodity that interests us to know the forecast, in the menu, and after we have entered the information, carefully observe the cash price curve, and the forecast curve, which anticipates its trend, in general terms, within the latency time, which is mentioned in the heading.

Let's take the example of Soybeans. It can be seen that the spot price (first chart) is showing a strong downturn trend, from mid-March up to May 2022. This trend was already anticipated in the forecast chart (second one) from October 2021, due to the strong net selling trend. 

Now we are seeing that the downward trend has changed to an upward trend, since May 2022. This is corroborated, since a spot rise in prices is observed, since June 2022, although without sufficient force. It is expected that by November, spot prices will be consolidated.

It is important to explain that the price forecast charts show three curves. Maximum forecast values, minimum forecast values, and mean standard deviation.

In the first case, the maximum value means that of the four monthly forecast values, the largest of them is taken. There are four, because on each Friday of the month, the CFTC publishes a coverage value (buyer and seller) of futures and options on futures, for each commodity, from which the forecast comes.

In the second case, it is the minimum forecast value processed, among the four coverage data, published by the CFTC, in that month.

Finally, the mean standard deviation tells us about the divergence or convergence between the first two curves, indicating turbulence or stabilization of the trend, respectively. 

----------------------------------------

 Source: AlondraUnion Books 

  Tags: Spot, Forecast, Explanation, Algorithm