calculate_rolling_metrics#

API documentation for tradeexecutor.visual.grid_search_advanced.calculate_rolling_metrics Python function.

calculate_rolling_metrics(grid_search_result, visualised_parameters, fixed_parameters, sample_freq='MS', lookback=Timedelta('90 days 00:00:00'), benchmarked_metric=BenchmarkMetric.sharpe)[source]#

Calculate rolling metrics for grid search.

We can have two parameters e.g. - N: size of traded basket - M: number of different pick sizes

For each N: - Calc rolling sharpe using last 3 months of returns (do in pandas) - This will give you an M-sized array or returns

For each quarter: - Look back 3 months and plot - yaxis: sharpe ratios - x-axis: array of Ns

Example output if using a single visualised parameter:

                0.50      0.75      0.99
2021-06-01       NaN       NaN       NaN
2021-07-01 -7.565988 -5.788797 -7.554848
2021-08-01 -3.924643 -1.919256 -3.914840
2021-09-01 -1.807489 -1.050918 -1.798897
2021-10-01 -1.849303 -1.604062 -1.841385
2021-11-01 -3.792905 -3.924210 -3.784793
2021-12-01 -4.156751 -4.186192 -4.148683

Example of 2d heatmap output:

                0.50                0.75                0.99
                   a         b         a         b         a         b
2021-06-01       NaN       NaN       NaN       NaN       NaN       NaN
2021-07-01 -7.565988 -7.565988 -5.788797 -5.788797 -7.554848 -7.554848
2021-08-01 -3.924643 -3.924643 -1.919256 -1.919256 -3.914840 -3.914840
2021-09-01 -1.807489 -1.807489 -1.050918 -1.050918 -1.798897 -1.798897
2021-10-01 -1.849303 -1.849303 -1.604062 -1.604062 -1.841385 -1.841385
2021-11-01 -3.792905 -3.792905 -3.924210 -3.924210 -3.784793 -3.784793
2021-12-01 -4.156751 -4.156751 -4.186192 -4.186192 -4.148683 -4.148683

Parma visualised_parameters:

Single parameter name for a line chart, two parameter name tuple for a heatmap.

Parameters:

sample_freq (pandas._libs.tslibs.offsets.DateOffset | str) – What is the frequency of calculating rolling value
lookback – For trailing sharpe, how far look back
grid_search_result (list[tradeexecutor.backtest.grid_search.GridSearchResult]) –
visualised_parameters (str | tuple[str, str]) –
fixed_parameters (dict) –

Returns:

DataFrame where

Index is timestamp, by step
Each column is value of visualisation parameter
Each row value is the visualised metric for that parameter and that timestamp

The first row contains NaNs as it cannot be calculated due to lack of data.

Return type:

DataFrame