如何从R语言backtest包的results数组中提取收益并转为DataFrame？

Nice question! I’ve messed around with the backtest package a fair bit, so let’s walk through how to pull those decile returns into a usable DataFrame step by step.

Step 1: Understand the results Array Structure #

First, let’s confirm the layout of that 5-dimensional results array. If you ran your backtest like this:

bt <- backtest(your_strategy, your_data, ...)

Then bt$results is the array in question. To see exactly what each dimension represents, run:

dimnames(bt$results)

Per the package docs, the dimensions are typically (strategies, symbols, dates, statistics, portfolios). The decile returns will live in the statistics dimension, with names like "decile1.return", "decile2.return", ..., "decile10.return".

Step 2: Extract the Decile Return Subset #

First, let’s isolate just the decile return entries from the array. We can use grep() to match the statistic names:

# Grab the indices for decile return statistics
decile_stat_indices <- grep("decile.*return", dimnames(bt$results)[[4]])

# Extract the subset (adjust indices if you have multiple strategies/symbols/portfolios)
# Here we're taking the first strategy, first symbol, first portfolio
decile_subset <- bt$results[1, 1, , decile_stat_indices, 1]

Step 3: Convert to a DataFrame (Base R Approach) #

Now let’s turn this 3D slice (dates x deciles) into a clean DataFrame with dates as rows and deciles as columns:

# Extract date labels and convert to Date type
dates <- as.Date(dimnames(decile_subset)[[1]])

# Extract decile labels
decile_names <- dimnames(decile_subset)[[2]]

# Transpose the subset to get dates as rows, deciles as columns
decile_data_matrix <- t(decile_subset)

# Convert to DataFrame and add the date column
decile_df <- as.data.frame(decile_data_matrix)
decile_df$date <- dates

# Reorder columns to put date first (optional but cleaner)
decile_df <- decile_df[, c("date", decile_names)]

Step 4: Tidyverse Alternative (Cleaner for Complex Cases) #

If you use the tidyverse ecosystem, this becomes even more straightforward, especially if you have multiple strategies/symbols/portfolios to handle:

library(tidyverse)

decile_df <- as.data.table(bt$results) %>%
  # Filter to only decile return statistics
  filter(str_detect(statistic, "decile.*return")) %>%
  # Keep only relevant columns
  select(date, statistic, value) %>%
  # Reshape to wide format (one column per decile)
  pivot_wider(names_from = statistic, values_from = value) %>%
  # Convert date to proper Date type
  mutate(date = as.Date(date))

Quick Validation & Notes #

• Double-check that the output matches what summary(bt) prints by running head(decile_df) and comparing to the summary output.
• If you have multiple strategies, symbols, or portfolios, adjust the indices in the base R approach (e.g., bt$results[2, , , decile_stat_indices, 1] for the second strategy) or add grouping in the tidyverse code (e.g., group_by(strategy, symbol) before pivoting).

内容的提问来源于stack exchange，提问作者Niccola Tartaglia