From b7c229431a78f857829eb04d01f3061768bf4977 Mon Sep 17 00:00:00 2001
From: nikosbosse <nikosbosse@gmail.com>
Date: Sat, 18 Jul 2026 14:44:30 +0200
Subject: [PATCH 1/4] Consolidate score() input preparation into a single
 helper

Add internal prepare_forecast_for_scoring() that performs the input
preparation steps previously duplicated across all eight score()
methods: cleaning the forecast, determining the forecast unit,
validating the metrics and converting to a plain data.table.

Validating the metrics inside the helper also locks down a subtle
ordering constraint: the default `metrics = get_metrics(forecast)`
promise must be forced while `forecast` still carries its forecast
class, i.e. before the rebind to a plain data.table.

Closes #941

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 NEWS.md                                |  1 +
 R/class-forecast-binary.R              |  6 ++--
 R/class-forecast-multivariate-point.R  |  8 ++----
 R/class-forecast-multivariate-sample.R |  8 +++---
 R/class-forecast-nominal.R             |  8 +++---
 R/class-forecast-ordinal.R             |  8 +++---
 R/class-forecast-point.R               |  6 ++--
 R/class-forecast-quantile.R            |  8 +++---
 R/class-forecast-sample.R              |  8 +++---
 R/score.R                              | 32 ++++++++++++++++++++++
 man/prepare_forecast_for_scoring.Rd    | 38 ++++++++++++++++++++++++++
 11 files changed, 100 insertions(+), 31 deletions(-)
 create mode 100644 man/prepare_forecast_for_scoring.Rd

diff --git a/NEWS.md b/NEWS.md
index e1ee239eb..71e4d5873 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,6 @@
 # scoringutils (development version)
 
+- Added an internal helper `prepare_forecast_for_scoring()` that consolidates the input preparation steps (cleaning the forecast, determining the forecast unit, validating the metrics and converting to a plain `data.table`) previously duplicated across all `score()` methods (#941).
 - Added `filter_scores()` and `impute_missing_scores()` for handling missing forecasts before summarisation. `filter_scores()` removes target combinations with insufficient model coverage, while `impute_missing_scores()` fills in missing scores using configurable strategies (worst, mean, NA, or reference model). Both use a strategy function pattern for extensibility. See `vignette("handling-missing-forecasts")` for details (#1122).
 - Added `plot_discrimination()` to visualise the discrimination ability of binary forecasts by plotting the distribution of predicted probabilities, stratified by the observed outcome. The function requires a `forecast_binary` object (created with `as_forecast_binary()`) (#942).
 - Fixed `summarise_scores()` producing a data.table with duplicate column names when the input `scores` object had no score columns (e.g. because every metric in `score()` warned and returned nothing). `summarise_scores()` now matches metric columns by exact name rather than regex partial match, and errors with a clear message when there is nothing to summarise (#1179).
diff --git a/R/class-forecast-binary.R b/R/class-forecast-binary.R
index 5c19c3df0..639e79713 100644
--- a/R/class-forecast-binary.R
+++ b/R/class-forecast-binary.R
@@ -103,9 +103,9 @@ is_forecast_binary <- function(x) {
 #' @rdname score
 #' @export
 score.forecast_binary <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
 
   scores <- apply_metrics(
     forecast, metrics,
diff --git a/R/class-forecast-multivariate-point.R b/R/class-forecast-multivariate-point.R
index 217249738..581f1e338 100644
--- a/R/class-forecast-multivariate-point.R
+++ b/R/class-forecast-multivariate-point.R
@@ -107,11 +107,9 @@ is_forecast_multivariate_point <- function(x) {
 score.forecast_multivariate_point <- function(
   forecast, metrics = get_metrics(forecast), ...
 ) {
-  forecast <- clean_forecast(
-    forecast, copy = TRUE, na.omit = TRUE
-  )
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
 
   observed <- forecast$observed
   predicted <- matrix(forecast$predicted, ncol = 1)
diff --git a/R/class-forecast-multivariate-sample.R b/R/class-forecast-multivariate-sample.R
index b44da448a..65907b57c 100644
--- a/R/class-forecast-multivariate-sample.R
+++ b/R/class-forecast-multivariate-sample.R
@@ -144,10 +144,10 @@ is_forecast_multivariate_sample <- function(x) {
 #' @rdname score
 #' @export
 score.forecast_multivariate_sample <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  forecast_unit <- get_forecast_unit(forecast)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
+  forecast_unit <- prep$forecast_unit
 
   # transpose the forecasts that belong to the same forecast unit
   f_transposed <- forecast[, .(
diff --git a/R/class-forecast-nominal.R b/R/class-forecast-nominal.R
index 187d0bdc0..57e1abdb0 100644
--- a/R/class-forecast-nominal.R
+++ b/R/class-forecast-nominal.R
@@ -134,10 +134,10 @@ is_forecast_nominal <- function(x) {
 #' @rdname score
 #' @export
 score.forecast_nominal <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  forecast_unit <- get_forecast_unit(forecast)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
+  forecast_unit <- prep$forecast_unit
 
   # transpose the forecasts that belong to the same forecast unit
   # make sure the labels and predictions are ordered in the same way
diff --git a/R/class-forecast-ordinal.R b/R/class-forecast-ordinal.R
index 4af5ab897..fde36039d 100644
--- a/R/class-forecast-ordinal.R
+++ b/R/class-forecast-ordinal.R
@@ -140,10 +140,10 @@ is_forecast_ordinal <- function(x) {
 #' @rdname score
 #' @export
 score.forecast_ordinal <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  forecast_unit <- get_forecast_unit(forecast)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
+  forecast_unit <- prep$forecast_unit
 
   # transpose the forecasts that belong to the same forecast unit
   # make sure the labels and predictions are ordered in the same way
diff --git a/R/class-forecast-point.R b/R/class-forecast-point.R
index 25e0a97dd..8ca0cde25 100644
--- a/R/class-forecast-point.R
+++ b/R/class-forecast-point.R
@@ -80,9 +80,9 @@ is_forecast_point <- function(x) {
 #' @rdname score
 #' @export
 score.forecast_point <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
 
   scores <- apply_metrics(
     forecast, metrics,
diff --git a/R/class-forecast-quantile.R b/R/class-forecast-quantile.R
index 7e1ac6246..43279253e 100644
--- a/R/class-forecast-quantile.R
+++ b/R/class-forecast-quantile.R
@@ -133,10 +133,10 @@ as_forecast_point.forecast_quantile <- function(data, ...) {
 #' @rdname score
 #' @export
 score.forecast_quantile <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  forecast_unit <- get_forecast_unit(forecast)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
+  forecast_unit <- prep$forecast_unit
 
   # transpose the forecasts that belong to the same forecast unit
   # make sure the quantiles and predictions are ordered in the same way
diff --git a/R/class-forecast-sample.R b/R/class-forecast-sample.R
index 27c1f653f..9a0bb9d1c 100644
--- a/R/class-forecast-sample.R
+++ b/R/class-forecast-sample.R
@@ -129,10 +129,10 @@ as_forecast_quantile.forecast_sample <- function(
 #' @rdname score
 #' @export
 score.forecast_sample <- function(forecast, metrics = get_metrics(forecast), ...) {
-  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
-  forecast_unit <- get_forecast_unit(forecast)
-  metrics <- validate_metrics(metrics)
-  forecast <- as.data.table(forecast)
+  prep <- prepare_forecast_for_scoring(forecast, metrics)
+  forecast <- prep$forecast
+  metrics <- prep$metrics
+  forecast_unit <- prep$forecast_unit
 
   # transpose the forecasts that belong to the same forecast unit
   f_transposed <- forecast[, .(predicted = list(predicted),
diff --git a/R/score.R b/R/score.R
index 037604fab..56e1ff91a 100644
--- a/R/score.R
+++ b/R/score.R
@@ -132,6 +132,38 @@ score.default <- function(forecast, metrics, ...) {
 }
 
 
+#' @title Prepare a forecast object for scoring
+#'
+#' @description
+#' This function performs the input preparation steps shared by all
+#' [score()] methods: it validates and cleans the forecast object (removing
+#' rows with missing values), determines the forecast unit, validates the
+#' metrics, and converts the forecast to a plain `data.table`.
+#'
+#' The metrics are validated within this function so that the default
+#' `metrics = get_metrics(forecast)` in the [score()] methods is evaluated
+#' while `forecast` is still a forecast object (method dispatch for
+#' [get_metrics()] would fail after the conversion to a plain `data.table`).
+#'
+#' @inheritParams score
+#' @returns A list with three elements: `forecast` (the cleaned forecast as
+#'   a plain `data.table`), `metrics` (the validated list of metrics) and
+#'   `forecast_unit` (a character vector with the columns that define the
+#'   unit of a single forecast).
+#' @importFrom data.table as.data.table
+#' @keywords internal
+prepare_forecast_for_scoring <- function(forecast, metrics) {
+  forecast <- clean_forecast(forecast, copy = TRUE, na.omit = TRUE)
+  forecast_unit <- get_forecast_unit(forecast)
+  metrics <- validate_metrics(metrics)
+  list(
+    forecast = as.data.table(forecast),
+    metrics = metrics,
+    forecast_unit = forecast_unit
+  )
+}
+
+
 #' @title Apply a list of functions to a data table of forecasts
 #' @description
 #' This helper function applies scoring rules (stored as a list of
diff --git a/man/prepare_forecast_for_scoring.Rd b/man/prepare_forecast_for_scoring.Rd
new file mode 100644
index 000000000..1220a0eda
--- /dev/null
+++ b/man/prepare_forecast_for_scoring.Rd
@@ -0,0 +1,38 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/score.R
+\name{prepare_forecast_for_scoring}
+\alias{prepare_forecast_for_scoring}
+\title{Prepare a forecast object for scoring}
+\usage{
+prepare_forecast_for_scoring(forecast, metrics)
+}
+\arguments{
+\item{forecast}{A forecast object (a validated data.table with predicted and
+observed values).}
+
+\item{metrics}{A named list of scoring functions. Each element should be a
+function reference, not a function call. For example, use
+\code{list("crps" = crps_sample)} rather than \code{list("crps" = crps_sample())}.
+Names will be used as column names in the output. See \code{\link[=get_metrics]{get_metrics()}} for
+more information on the default metrics used. See the \emph{Customising metrics}
+section below for information on how to pass custom arguments to scoring
+functions.}
+}
+\value{
+A list with three elements: \code{forecast} (the cleaned forecast as
+a plain \code{data.table}), \code{metrics} (the validated list of metrics) and
+\code{forecast_unit} (a character vector with the columns that define the
+unit of a single forecast).
+}
+\description{
+This function performs the input preparation steps shared by all
+\code{\link[=score]{score()}} methods: it validates and cleans the forecast object (removing
+rows with missing values), determines the forecast unit, validates the
+metrics, and converts the forecast to a plain \code{data.table}.
+
+The metrics are validated within this function so that the default
+\code{metrics = get_metrics(forecast)} in the \code{\link[=score]{score()}} methods is evaluated
+while \code{forecast} is still a forecast object (method dispatch for
+\code{\link[=get_metrics]{get_metrics()}} would fail after the conversion to a plain \code{data.table}).
+}
+\keyword{internal}

From c875c18cad001de94232d91c945f5242181a9dc6 Mon Sep 17 00:00:00 2001
From: nikosbosse <nikosbosse@gmail.com>
Date: Sat, 18 Jul 2026 14:49:20 +0200
Subject: [PATCH 2/4] Add unit tests for prepare_forecast_for_scoring()

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 tests/testthat/test-score.R | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/tests/testthat/test-score.R b/tests/testthat/test-score.R
index 44cd7b65f..145c49000 100644
--- a/tests/testthat/test-score.R
+++ b/tests/testthat/test-score.R
@@ -108,6 +108,42 @@ test_that("function produces output for a nominal format case", {
 })
 
 
+# =============================================================================
+# prepare_forecast_for_scoring() # nolint: commented_code_linter
+# =============================================================================
+
+test_that("prepare_forecast_for_scoring() prepares a forecast for scoring", {
+  prep <- prepare_forecast_for_scoring(
+    example_quantile, get_metrics(example_quantile)
+  )
+  expect_named(prep, c("forecast", "metrics", "forecast_unit"))
+
+  # the forecast is returned as a plain data.table with NA rows removed
+  expect_s3_class(prep$forecast, c("data.table", "data.frame"), exact = TRUE)
+  expect_identical(nrow(prep$forecast), nrow(na.omit(example_quantile)))
+
+  # the input is not modified by reference
+  expect_s3_class(example_quantile, "forecast_quantile")
+
+  expect_identical(prep$metrics, get_metrics(example_quantile))
+  expect_identical(prep$forecast_unit, get_forecast_unit(example_quantile))
+})
+
+test_that("prepare_forecast_for_scoring() validates its inputs", {
+  expect_error(
+    prepare_forecast_for_scoring(data.frame(x = 1), list(mean = mean)),
+    "forecast object"
+  )
+  expect_warning(
+    prepare_forecast_for_scoring(
+      example_binary,
+      list(brier_score = brier_score, invalid = "not a function")
+    ),
+    "not a valid function"
+  )
+})
+
+
 # =============================================================================
 # apply_metrics() # nolint: commented_code_linter
 # =============================================================================

From 79a6477b754d77c0d9c3ab3cc16808473ded60a1 Mon Sep 17 00:00:00 2001
From: nikosbosse <nikosbosse@gmail.com>
Date: Sat, 18 Jul 2026 14:56:15 +0200
Subject: [PATCH 3/4] Address adversarial review findings

- Document precisely why metrics are validated inside
  prepare_forecast_for_scoring() and that the lazy default is evaluated
  on the original (pre-cleaning) forecast object.
- Describe the metrics argument locally instead of inheriting it from
  score(), removing a dangling reference to the "Customising metrics"
  section that only exists on the score() help page.
- Reword the NEWS item: determining the forecast unit was previously
  only done by the methods that need it, not duplicated across all.
- Slim the helper tests down to one focused test and make the
  not-modified-by-reference check a real before/after comparison.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 NEWS.md                             |  2 +-
 R/score.R                           | 14 ++++++++++----
 man/prepare_forecast_for_scoring.Rd | 21 ++++++++++-----------
 tests/testthat/test-score.R         | 21 +++------------------
 4 files changed, 24 insertions(+), 34 deletions(-)

diff --git a/NEWS.md b/NEWS.md
index 71e4d5873..8f0314cd1 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,6 +1,6 @@
 # scoringutils (development version)
 
-- Added an internal helper `prepare_forecast_for_scoring()` that consolidates the input preparation steps (cleaning the forecast, determining the forecast unit, validating the metrics and converting to a plain `data.table`) previously duplicated across all `score()` methods (#941).
+- Added an internal helper `prepare_forecast_for_scoring()` that consolidates the input preparation steps previously duplicated across the `score()` methods: cleaning the forecast, validating the metrics and converting to a plain `data.table`, plus determining the forecast unit for the methods that need it (#941).
 - Added `filter_scores()` and `impute_missing_scores()` for handling missing forecasts before summarisation. `filter_scores()` removes target combinations with insufficient model coverage, while `impute_missing_scores()` fills in missing scores using configurable strategies (worst, mean, NA, or reference model). Both use a strategy function pattern for extensibility. See `vignette("handling-missing-forecasts")` for details (#1122).
 - Added `plot_discrimination()` to visualise the discrimination ability of binary forecasts by plotting the distribution of predicted probabilities, stratified by the observed outcome. The function requires a `forecast_binary` object (created with `as_forecast_binary()`) (#942).
 - Fixed `summarise_scores()` producing a data.table with duplicate column names when the input `scores` object had no score columns (e.g. because every metric in `score()` warned and returned nothing). `summarise_scores()` now matches metric columns by exact name rather than regex partial match, and errors with a clear message when there is nothing to summarise (#1179).
diff --git a/R/score.R b/R/score.R
index 56e1ff91a..fffbc8328 100644
--- a/R/score.R
+++ b/R/score.R
@@ -140,12 +140,18 @@ score.default <- function(forecast, metrics, ...) {
 #' rows with missing values), determines the forecast unit, validates the
 #' metrics, and converts the forecast to a plain `data.table`.
 #'
-#' The metrics are validated within this function so that the default
-#' `metrics = get_metrics(forecast)` in the [score()] methods is evaluated
-#' while `forecast` is still a forecast object (method dispatch for
-#' [get_metrics()] would fail after the conversion to a plain `data.table`).
+#' The metrics are validated within this function so that the lazy default
+#' `metrics = get_metrics(forecast)` of the [score()] methods is forced
+#' before those methods rebind `forecast` to a plain `data.table` (there is
+#' no `get_metrics.default()`, so forcing the default after that rebind
+#' would fail). The default is thereby evaluated on the original forecast
+#' object passed to [score()], i.e. before rows with missing values are
+#' removed. This makes no difference for the built-in [get_metrics()]
+#' methods, which do not inspect the data.
 #'
 #' @inheritParams score
+#' @param metrics A named list of scoring functions. See [score()] for
+#'   details.
 #' @returns A list with three elements: `forecast` (the cleaned forecast as
 #'   a plain `data.table`), `metrics` (the validated list of metrics) and
 #'   `forecast_unit` (a character vector with the columns that define the
diff --git a/man/prepare_forecast_for_scoring.Rd b/man/prepare_forecast_for_scoring.Rd
index 1220a0eda..1f27f85e0 100644
--- a/man/prepare_forecast_for_scoring.Rd
+++ b/man/prepare_forecast_for_scoring.Rd
@@ -10,13 +10,8 @@ prepare_forecast_for_scoring(forecast, metrics)
 \item{forecast}{A forecast object (a validated data.table with predicted and
 observed values).}
 
-\item{metrics}{A named list of scoring functions. Each element should be a
-function reference, not a function call. For example, use
-\code{list("crps" = crps_sample)} rather than \code{list("crps" = crps_sample())}.
-Names will be used as column names in the output. See \code{\link[=get_metrics]{get_metrics()}} for
-more information on the default metrics used. See the \emph{Customising metrics}
-section below for information on how to pass custom arguments to scoring
-functions.}
+\item{metrics}{A named list of scoring functions. See \code{\link[=score]{score()}} for
+details.}
 }
 \value{
 A list with three elements: \code{forecast} (the cleaned forecast as
@@ -30,9 +25,13 @@ This function performs the input preparation steps shared by all
 rows with missing values), determines the forecast unit, validates the
 metrics, and converts the forecast to a plain \code{data.table}.
 
-The metrics are validated within this function so that the default
-\code{metrics = get_metrics(forecast)} in the \code{\link[=score]{score()}} methods is evaluated
-while \code{forecast} is still a forecast object (method dispatch for
-\code{\link[=get_metrics]{get_metrics()}} would fail after the conversion to a plain \code{data.table}).
+The metrics are validated within this function so that the lazy default
+\code{metrics = get_metrics(forecast)} of the \code{\link[=score]{score()}} methods is forced
+before those methods rebind \code{forecast} to a plain \code{data.table} (there is
+no \code{get_metrics.default()}, so forcing the default after that rebind
+would fail). The default is thereby evaluated on the original forecast
+object passed to \code{\link[=score]{score()}}, i.e. before rows with missing values are
+removed. This makes no difference for the built-in \code{\link[=get_metrics]{get_metrics()}}
+methods, which do not inspect the data.
 }
 \keyword{internal}
diff --git a/tests/testthat/test-score.R b/tests/testthat/test-score.R
index 145c49000..7e6fc5159 100644
--- a/tests/testthat/test-score.R
+++ b/tests/testthat/test-score.R
@@ -113,9 +113,8 @@ test_that("function produces output for a nominal format case", {
 # =============================================================================
 
 test_that("prepare_forecast_for_scoring() prepares a forecast for scoring", {
-  prep <- prepare_forecast_for_scoring(
-    example_quantile, get_metrics(example_quantile)
-  )
+  input <- data.table::copy(example_quantile)
+  prep <- prepare_forecast_for_scoring(input, get_metrics(input))
   expect_named(prep, c("forecast", "metrics", "forecast_unit"))
 
   # the forecast is returned as a plain data.table with NA rows removed
@@ -123,26 +122,12 @@ test_that("prepare_forecast_for_scoring() prepares a forecast for scoring", {
   expect_identical(nrow(prep$forecast), nrow(na.omit(example_quantile)))
 
   # the input is not modified by reference
-  expect_s3_class(example_quantile, "forecast_quantile")
+  expect_identical(input, example_quantile)
 
   expect_identical(prep$metrics, get_metrics(example_quantile))
   expect_identical(prep$forecast_unit, get_forecast_unit(example_quantile))
 })
 
-test_that("prepare_forecast_for_scoring() validates its inputs", {
-  expect_error(
-    prepare_forecast_for_scoring(data.frame(x = 1), list(mean = mean)),
-    "forecast object"
-  )
-  expect_warning(
-    prepare_forecast_for_scoring(
-      example_binary,
-      list(brier_score = brier_score, invalid = "not a function")
-    ),
-    "not a valid function"
-  )
-})
-
 
 # =============================================================================
 # apply_metrics() # nolint: commented_code_linter

From 97db99d389111bcd7e353078d7bccd9ee364dd39 Mon Sep 17 00:00:00 2001
From: nikosbosse <nikosbosse@gmail.com>
Date: Sat, 18 Jul 2026 14:57:24 +0200
Subject: [PATCH 4/4] Add score() test for the ordinal format

Scoring an ordinal forecast was not exercised anywhere in the test
suite, which codecov surfaced as missing patch coverage on the
refactored score.forecast_ordinal() preamble.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 tests/testthat/test-score.R | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/tests/testthat/test-score.R b/tests/testthat/test-score.R
index 7e6fc5159..5a50f3f13 100644
--- a/tests/testthat/test-score.R
+++ b/tests/testthat/test-score.R
@@ -108,6 +108,12 @@ test_that("function produces output for a nominal format case", {
 })
 
 
+# test ordinal case ------------------------------------------------------------
+test_that("function produces output for an ordinal format case", {
+  expect_no_condition(score(as_forecast_ordinal(na.omit(example_ordinal))))
+})
+
+
 # =============================================================================
 # prepare_forecast_for_scoring() # nolint: commented_code_linter
 # =============================================================================