The risk of China’s military conflict and the United States is close to 50%21 октября, 2021
My new post is dedicated to a very interesting event —
Opened free access to the visual interface of the new
The system was developed by 10 years with a team of 30 specialists of the COOLABAH CAPITAL INVESTMENTS investment fund.
This is not an academic study. Here by telling the words
$ 7 billion investments under his management depend on the accuracy of predictions of this investment fund. That is why it is not only curious, but also quite intriguing news.
In addition, because Access to the system is open, everyone can see and analyze how the chances of war between couples of countries (Russia and USA, Russia and China, India and China, Germany and Russia, Australia and China …) Over the past 160 years, and what is their forecast For the coming year, 5 and 10 years.
No less interesting context of this news.
For, not knowing it, you can miss the assessment of the importance and prospects of the new prediction system. Therefore, I decided to write in more detail and about this context that concerns the following.
If the context of the news is not important to you, immediately go to the last section of the post «Skura on the horse predictions».
Inventor of the Applied Information Economy
«There are two main interpretations of the word» intangible «. Firstly, intangible, according to tradition, they consider things that are not literally not material (bodily, tangible), but, nevertheless, measurable. Good examples of objects that cannot be touched, but can be measured, it is time for the time, the budget, the ownership of the patent, etc. Today there has already developed a whole industry assessment of intangible assets, such as copyright and trademark. But over time, the word «intangible» began to consume and in the meaning » Measurement is direct or indirect. «And so I argue that objects, intangible words in this sense, do not exist at all.»
Explaining why the immeasurable intangible is just an illusion, Hubbard writes that people consider things in immeasurable for three reasons arising from erroneous ideas about different aspects of quantitative assessment:
Essence of measurement
Essence of the concept of measurement
The value of such a interpretation of the measurement essence is that it should not eliminate the uncertainty completely. It is quite enough to reduce uncertainty, because its effect can repeatedly exceed the measurement costs.
«Nothing prevents the progress of knowledge as the blurry of terminology
That is why, according to Douglas Hubbard, «giving the definition of its terms, so it is important to understand what solutions we carry out our measurements.» Therefore, then we will use the following definitions.
Indicator of uncertainty
For example, we believe that there are 40% the likelihood that during the year between the United States and China may occur an armed clash with losses of the United States, at least dozens of life servicemen and losses in billions of dollars.
We have decided on two aspects of a quantitative assessment of three.
✔️ with the essence of measuring the risk of military conflicts:
We do not expect absolutely accurate predictions, but strive to reduce the uncertainty of the occurrence of various militarized interstate disputes and the risks of subsequent losses by military living force and technology, as well as losses among the population and infrastructure facilities.
✔️ with definitions of uncertainty and risks, as well as with their indicators.
Now it’s about the definition and selection of the measurement method.
The reader should not confuse that until recently the accuracy of the prediction methods of military conflict was extremely low. First, as already mentioned above, even from not the most accurate measurements (probability estimates, etc.) may be benefit. And secondly, now in this business changes with speed, commensurate with the growth rate of the laptop disk.
So here. Wishing to get acquainted with the prehistory of the question I can recommend the review work of 2017 «
Readers who want to further get acquainted with conflict prediction systems makes sense to appeal to the bibliography of the collection published in December 2020
The essential lack of most existing systems is that machine learning models used by them do not give
Suppose, based on the set of historical data, we trained the prognostic model to predict the likelihood of a big war between China and the United States. Now suppose that the model predicted the likelihood of such a 5% war. If we lived in a multivelenic, we would check this prediction, observing the relationship between China and the United States in several parallel universes. And if when observed in 100 universes, the war between China and the USA happened in five, we would say that the predicted probability of 5% reflects the real probability.
But alas, we have no multivers. And check the predicted probability, watching the next 100 wars between China and the United States in our only reality, we, too, alas, can not. Therefore, the only way for us is to ensure that the predicted number reflects the real probability is
The challenge of calibration is to learn
✔️ first set the desired probability with which an unknown value will fall into the interval proposed by you,
✔️ And then, on the basis of some conclusions, to offer the upper and lower boundary of the confidence interval (for example, 90%, in which an unknown value will be inside it in nine cases out of ten).
The exact answer to this question, most likely, knows quite few people, however, any reader is quite capable of approximately evaluating this amount.
One way to show the inaccuracy of determining the value is to express it in the form of an interval of possible values. In statistics, the interval in which, with some probability, a correct answer may be called a confidence interval (Confidence Interval); The 90 percent confidence interval is the range of values containing the correct with a probability of 90%.
Suppose we strive to give a response that will be accurate with a probability of 90%. Here are options. Suppose someone involves the range of wings springs from 9 to 10 meters. But it seems fairly narrow, and if there is no big special specials in airplanes, then the probability of misachering will be quite large. In other words, giving such a narrow range of ratings, man shows
Another extreme case can be a range from 9 to 100 meters
In addition to the problem of calibration, almost all existing systems of predicting wars have
We end up get acquainted with the context of the news on the launch of a new system of predicting paramilitary interstate disputes (Militarized Interstate Disputes — MID). Now look at the new system.
It was designed for 10 years with a team of 30 investment fund specialists.
Like bankers with insurers, the investment fund has a detailed database on the history of unpredictable «black swans» — militant interstate disputes over the past 160 years. Investment analysts perfectly operate in advanced quantitative methods, including the latest methods of machine learning to predict the empirical probability of various types of military conflicts (from logistic regression to gradient decisions trees).
In addition (and this is fundamentally important), used in the CGI system approach:
✔️ is based on the calibration of probabilistic forecasts;
✔️ takes into account the temporary prediction horizon;
✔️ calculates the risks not clicked (for all types of military conflict), but depending on the severity of the conflict.
The value of calibration was said in the previous section.
Temporary forecasting horizons are: 1, 5 and 10 years.
4 types (severity) of conflicts are predicted: the threat to the use of force («threat»), the actual use of force («power»), military collision or raid («attack»), a long series of military operations («War»).
Military conflict correlates use an extensive set of indicators from various international databases.
When forecasting, more than 200 correlates are taken into account, the detailed description of which can be found in the brochure and technical description of the prediction system (they can be downloaded from its open site.
To assess the work of the prediction system, it is best to play with it yourself. It remains only to give several explanatory examples.
Example 1. The probability forecast of China’s military collision and Taiwan on the horizon is 10 years old.
Example 2. The probability forecast of China’s military clash and the United States on the horizon is 10 years old.
Example 3. Forecast of the probability of military collision of Russia and the United States on the horizon for 10 years.
Example 4. Matrix forecasts of the probability of military clashes between eight countries (China, India, Russia, Taiwan, United Kingdom, USA, Japan, Germany) on the horizon 10 years.
Example 5. Matrix of forecasts of the likelihood of a full-scale war between pairs of eight countries (China, India, Russia, Taiwan, United Kingdom, USA, Japan, Germany) on the horizon 10 years.
The greatest probability of a full-scale war in the next 10 years in China with India is 21.9%.
Russia has the greatest probability of a full-scale war in the next 10 years, it is projected with China 9.2%, and from the US only 2.4%.
Giving any other comments see no sense. Christopher Joy put his skin on the con. We will look further to the results.
predictions # war