# Description of research

## How we conduct research

In our research, we stick to the rules of writing:

• hypotheses;
• data used for analysis;
• segments of financial market;
• time intervals and timeframes;
• methods of detecting trade signals;
• a conclusion.

In our research, we calculate the following indicators:

• correlation coefficient;
• momentum;
• share of profitable positions;
• volatility.

Further, we will provide a complete description of the indicators.

## Rules for writing research

### Hypothesis

So, a hypothesis is a problem statement for research. By formulating a hypothesis, we decide on what exactly we want to test, whether it is the interconnection between financial instruments or the impact of an event on the market

We prefer testing hypotheses that are related to statements of economic theory, market "gurus" or generally accepted principles.

We don't apply combinatorics, don't select events and don't use neural networks and various black-boxes. We analyze how each economic event affects the market.

### Data used for analysis

There are many factors that influence changes in share prices, exchange rates and other instruments. A large sample of data helps us to achieve high level of accuracy in research. We are trying to include various segments of financial markets and historical data for several decades in research to test hypotheses.

We tend to conduct research with a sample of more than 300 cases. A research can be based on more than 100,000 cases.

### Market segments

As market segments, we analyze U.S. stocks, Russian stocks, indices, currency pairs, energy commodities, metals and other goods. There are more than 80 instruments in 5 segments.

We use the most widespread instruments of the segment, for instance, we consider only blue chips, DOW30 stocks among all U.S. stocks. They are instruments with a high trading volume, and consequently they are less influenced by random events, financial interventions, etc

### Time intervals and timeframes

When conducting research on economic events, for instance, on the release of unemployment data, we examine the data during the time intervals from one day to three weeks. We have not selected any methods for intraday trading. And there is no point in examining one-month interval or a longer time interval, as economic data is updated every month.

While examining technical analysis signals, we take H1 and D1 timeframes. The dynamics of rates within timeframes less than H1 are influenced by an excessive number of random factors, while for W1 timeframes, there is a small sample of events.

### Conclusion

In this section we present the results obtained from research and review them.

When examining technical analysis signals, the signal should have an effect on most segments, otherwise, we come to the conclusion that there's no influence.

We use the phrase "the influence has been identified" to demonstrate it. We assume that an economic event is a signal indicating changes in the market, rather than infer cause-and-effect relationships.

## Indicators in research

### Correlation coefficient

Correlation is a statistical relationship between variables. The correlation is determined by calculating the Pearson linear correlation coefficient and the correlation coefficient.

The correlation coefficient is calculated according to the following equation:

$\dpi{120}&space;\large&space;r_{XY}&space;=&space;\frac{\sum(X&space;-&space;\bar&space;X)(Y&space;-&space;\bar&space;Y&space;)}{\sqrt{\sum(X&space;-&space;\bar&space;X)^2&space;\sum(Y&space;-&space;\bar&space;Y)^2}}$
$\dpi{120}&space;\large&space;\bar&space;X&space;=&space;\frac{1}{n}&space;\sum\limits_{i=1}^n&space;X_t,&space;\bar&space;Y&space;=&space;\frac{1}{n}&space;\sum\limits_{t=1}^n&space;Y_t$

We calculate the correlation coefficient by comparing the following values:

• rates of financial instruments and economic indicators;
• changes in economic indicators and in rates of economic instruments;
• changes in rates of one type of economic instruments and another type of economic instruments.

Correlation is calculated both with and without a time shift. For example, we can compare the change in oil prices today with the change in the EURUSD rate tomorrow or the day after tomorrow by calculating the correlation between oil prices and the EURUSD rate.

The correlation with time shift makes it possible to predict the dynamics of rates and make a profit.

To evaluate the obtained values of the correlation coefficient in research, we use the Chaddock scale.

Correlation coefficient Interpretation
less than 0.1 no correlation
from 0.1 to 0.3 low correlation
from 0.3 to 0.5 moderate correlation
from 0.5 to 0.7 remarkable correlation
from 0.7 to 0.9 strong correlation
from 0.9 to 0.99 very strong correlation
from 0.99 to 1 functional correlation

#### Note

• A correlation coefficient is sensitive to outliers. One random value can significantly distort the coefficient. The sample of data must be large, so that random fluctuations could cancel each other out.
• A correlation doesn't mean there is a cause-and-effect relationship between the values.
• The absence of a linear correlation doesn't mean there is no relationship at all. There may be a non-linear correlation.

### Momentum

Various events, such as formation of a head and shoulders pattern in technical analysis or seasonal effects, may affect rates of financial instruments. A linear correlation is not suitable for assessing correlation between such events and changes in rates, that's why we introduced our own indicator – momentum.

Momentum allows us to assess the market reaction and its movements.

Momentum is the rate of the average change in the price of a financial instrument over a period of time after a certain market event.

Momentum is calculated according to the formula:

$\dpi{120}&space;\large&space;\textit&space;I&space;=&space;\frac{1}{n}&space;\times&space;\sum&space;\textit&space;P$

where:

P – the relative change in the rate of a financial instrument over a certain period of time, expressed as percentage.

P = (the price of the instrument at the end of the period – the price of the instrument at the beginning of the period) / the price of the instrument at the beginning of the period * 100%
• for sell events
P = (the price of the instrument at the beginning of the period – the price of the instrument at the end of the period) / the price of the instrument at the beginning of the period * 100%

n — the number of events;

A positive value of momentum indicates the profitability of the event, while a negative value indicates its unprofitability. It is not enough to find out if the value of momentum is positive or negative, as it also should be statistically significant.

A significant momentum is a value greater than the required minimum value.

Minimum values of momentum set by the MarketCheese team:

• 0.15% for the H1 timeframe with a sample of 3000 cases;
• 0.3% for the D1 timeframe with a sample of 300 cases.

If there is a significant momentum and a sufficient sample size, we can assert that the influence of the event on the market has been identified.

#### Note

• The minimum values of momentum are axiomatically set by our team. The volatility of rates that differ depending on the segments of financial instruments and holding period are not considered.

### Share of profitable positions

Share of profitable positions, or SPP, is another indicator set by the MarketCheese team that assesses the impact of events.

SPP refers to the share of profitable trading positions out of the total number of positions and is calculated according to the formula:

SPP is a kind of a qualitative indicator used to evaluate effectiveness. The higher SPP is, the more often you will make profitable trades.

#### Note

• Unlike momentum, SPP does not consider the amount of profit and loss. If SPP is more than 50%, the strategy may still be unprofitable, as the average unprofitable position is larger than the average profitable one.

### Volatility

Evaluating volatility of financial instruments is important as instruments with higher volatility are considered to be more risky compared to indicators of less volatile instruments. At the same time, high volatility makes it possible to generate extra income.

We apply an arithmetic method when calculating the volatility:

$\dpi{120}&space;\large&space;V(\%)&space;=&space;\frac{2}{n}&space;\times&space;\sum&space;\frac{(max-min)}{(max+min)}&space;\times&space;100%$

Where max – the maximum value of the instrument over a period of time, min – the minimum value, n – the number of periods.

We evaluate volatility in our research to identify the influence of an event on the market. If the volatility does not change, the event may not affect significantly the change in rates of financial instruments.

In our research, we rely on the following values when evaluating volatility:

Volatility Interpretation
less than 2% low
from 2 to 5% medium
from 5% high

#### Note

• The values of volatility stated above were set on the basis of a large sample of historical rates of financial instruments in various market segments, such as Forex, stock and commodity markets, cryptocurrencies.

Subscribe to our newsletter and stay up to date with all the news!