What are Targets?

Rally-Based Targets

Rally-based targets quantify the position of candles in rallies. A rally is a price movement connecting high and low pivots in the price. An upward rally is a rally that starts with a low pivot and ends with a high pivot. A downward rally is a rally that begins with a high pivot and ends with a low one. This class of target finds pivots in the price and assigns numeric values or labels to candles based on their relative position concerning the nearest high and low pivots.

Rally-Based Rating

Rally-based Rating (RBR) assigns a rating (a number between -100 and +100) to data points based on their price position in the market rallies. A market rally refers to the data points that are between a high and low pivot point. A high pivot point marks the end of an upward trend and the beginning of a downward one, where its price is higher than both preceding and subsequent N prices. The inputs of this target are:

• Pivot Window Length: The number of candles considered to find pivots in the past and future; smaller values make the target consider shorter-term rallies.

• Target Name: The name given to the column that stores the result

When the market is in a downtrend, the RBR value will be negative. The closer it is to the high pivot point of rallies, the more negative the RBR value will be. Conversely, in an upward trend, the RBR value will be positive and will have a higher value as it gets closer to the high pivot point.

Reward/Risk Targets

Reward/Risk targets consider N candles in the future and measure the potential reward and risk of opening a position at a candle. We have Reward/Risk Rating and Reward/Risk Classifying that return numeric values and labels, respectively.

Reward/Risk Rating

Reward/Risk Rating measures the maximum run-up and drawdown in the forward-looking window with a defined window size. The reward to risk ratio is defined as (maximum run-up/maximum drawdown). inputs of this target are:

• Look Ahead Window Length: The size of the window in which drawdown and run-up are calculated for candles

• target Name: The name given to the output column

Gain-Based Rating

Gain-Based Rating (GBR) checks candles in a forward-looking window and returns window attributes as a rating. These attributes include maximum run-up, maximum drawdown, average price change percentage, and price change percentage at the end of the window. Inputs of this target are:

• Look Ahead Window Length: The size of the forward-looking window

• Measurement Method: The metric used to model the price behavior in the window; could be max, min, average, and end of window. When max is used, the maximum run-up in the window is reported as the target value. Using min gives you maximum drawdowns. The average and end of window return the average rate of change ( sum of candle rate of changes divided by the number of candles in the window) and the rate of change when only the price of data of the first and last candles in the windows are considered, respectively.

• Source: If shadow is selected, the target uses the upper shadow for the max measurement method and the lower shadow for the min measurement method. If close is selected, the close price is always used, regardless of the selected measurement method.

• Target Name: The name given to the column that stores the result

Stoploss-Takeprofit Classifying

This target labels candles based on future price movements within a forward-looking window of a specified length. If the price increases by at least the Take Profit Percentage, the candle is labeled TP (Take Profit). If the price decreases by at least the Stop Loss Percentage, the candle is labeled SL (Stop Loss). Any candle that does not meet either condition is labeled NA (No Action).

Note: If both the take-profit and stop-loss levels are reached within the same window, the label is determined by whichever level is hit first.

The inputs for this target are:

• Look Ahead Window Length: The number of candles considered when evaluating whether the take-profit or stop-loss level is reached.

• Trade Type: The trade direction for which the target is calculated. This can be either Long or Short.

• Stop Loss Percentage: The stop-loss threshold, expressed as a percentage.

• Take Profit Percentage: The take-profit threshold, expressed as a percentage.

• Trading Fee Percentage: The percentage deducted to account for trading fees. A candle is labeled TP only if the realized profit, after subtracting this value, exceeds the Take Profit Percentage.

• Labels: The labels assigned to candles, which are tp, sl, and na by default.

• Target Name: The name assigned to the output column.

The figure below illustrates the behavior of this target. For example, when the Take Profit Percentage is set to 3% and the Stop Loss Percentage is set to 1%, the labels for the first and second windows are SL and TP, respectively.

Positive Momentum Targets

Positive Momentum Rating

Positive draws a horizontal line from the candle to the end of a forward-looking window of size N. It then calculates the area between candles above the line and the line and the area between candles below the line and line. In the figure below, These areas are shown in green and red, respectively. If G is the green area and R is the red area, the Positive Momentum Rating is G/(G+R)

The inputs of this target are:

• Look Ahead Window Length: The size of the forward-looking window

• Target Name: The name given to the output column

Normalized Stoploss-Takeprofit Based Classifying

This target labels candles based on future price movements within a forward-looking window of a specified length, using volatility-adjusted thresholds. Instead of fixed percentages, the price levels are normalized using the Average True Range (ATR).

The base threshold is calculated by multiplying the ATR Multiplier by the current ATR value. For a Long trade:

• The Stop Loss distance is equal to .

• The Take Profit distance is dynamically scaled using the Reward to Risk ratio: .

If the price hits the take-profit level first within the window, the candle is labeled tp. If it hits the stop-loss level first, it is labeled sl. Any candle that does not meet either condition within the window is labeled na (No Action).

Note: If both levels are reached within the same window, the label is determined by whichever level is hit first chronologically.

The inputs for this target are:

• Look Ahead Window Length: The number of candles considered when evaluating whether the take-profit or stop-loss level is reached (e.g., 1000).

• Trade Type: The trade direction for which the target is calculated. This can be either long or short.

• Atr Multiplier: The multiplier applied to the ATR value to establish the baseline risk distance (e.g., 2.5).

• Reward to Risk: The ratio used to scale the take-profit distance (reward) relative to the stop-loss distance (risk) (e.g., 1).

• Trading Fee Percentage: The percentage deducted to account for trading fees. A candle is labeled tp only if the realized profit, after subtracting this value, satisfies the take-profit condition.

• Labels: The labels assigned to the candles, which are tp, sl, and na by default.

• Target Name: The name assigned to the output column (e.g., nstc_long_2p5_1).

The system dynamically shifts these boundaries for every candle depending on market volatility. For example, if the Atr Multiplier is set to 2 and the Reward to Risk ratio is set to 3, a long trade will trigger a tp label if the price increases by at least , and an sl label if the price drops by .

Trade-Based Rating

This returns attributes of the price before a user-defined exit signal occurs. For each candle. this target defines a window starting at the candle itself and ending at the next exit signal. The target returns the attributs of this window and the price itself based on the Measurment Method selected by the user. These attributes are:

• Number of time steps to the next exit signal

• Return made by opening a position at candle and closing at next exit signal

• Average price before next exit signal

• Standard deviation of prices before next exit signal

• Max return made by opening a position at candle before next exit signal

• Min return made by opening a position at candle before next exit signal

The inputs of this target are:

• Timeframe: The timeframe in which the exit signal is defined

• Exit Signal: The signal against which candles are evaluated

• Measurement Method: The attribute of the window or price returned by the target

• Target Name: The name given to the output column

Triple Barrier Method

The Triple Barrier Method (TBM) is an adaptation of a labeling technique first introduced in Advances in Financial Machine Learning by M. L. de Prado. This method assigns labels to price candles using three barriers: two horizontal barriers and one vertical barrier.

If the price touches the upper barrier first, the candle is labeled tp, indicating that the price has increased and reached the take-profit level. If the price touches the lower barrier first, the candle is labeled sl, indicating that the price has decreased and reached the stop-loss level. If neither the upper nor the lower barrier is touched before the end of the forward-looking window, the vertical barrier is reached, and the candle is labeled na.

The positions of the upper and lower barriers are calculated dynamically based on the standard deviation of the price’s percentage changes over a backward-looking window. This results in adaptive take-profit and stop-loss levels for each candle, reflecting recent market volatility.

The inputs for this labeling method are:

• Backward-Looking Window Size: The number of previous candles used to compute the standard deviation of the price’s percentage changes.

• Forward-Looking Window Size: The number of future candles during which the price is monitored to determine whether it hits the upper or lower barrier. If neither barrier is reached, the candle is labeled na.

• Take-Profit Coefficient: The upper barrier is defined as

  
  

  current price + (Take-Profit Coefficient × standard deviation of price percentage changes in the backward-looking window).

• Stop-Loss Coefficient: The lower barrier is defined as

  
  

  current price + (Stop-Loss Coefficient × standard deviation of price percentage changes in the backward-looking window).

• Min Move (%): The minimum price change required to label a candle as buy, or sell

• Labels: The labels assigned to the candles, which are tp, sl, and na by default.

• Target Name: The name given to the output column

The figure below illustrates how the target works. The green lines represent the upper barriers across three different windows, while the red lines indicate the corresponding lower barriers. The dashed and solid black lines mark the beginning and end of each window, respectively. Specifically, the dashed line denotes the first candle, and the solid line represents the vertical barrier. The number of candles between these two lines corresponds to the Forward-Looking Window Size.

The positions of the upper and lower barriers depend on the Take-Profit and Stop-Loss coefficients, respectively. Higher values of these coefficients make the barriers harder to reach. Note that the positions of the barriers are determined not only by these coefficients, but also by the standard deviation of the price’s percentage changes over the most recent Backward-Looking Window Size number of candles.

To read more about this method, please refer to Advances in Financial Machine Learning.