diff --git a/DESCRIPTION b/DESCRIPTION
index 4d81a3a..11d67f1 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -19,7 +19,7 @@ URL: https://github.com/tidymodels/finetune,
 BugReports: https://github.com/tidymodels/finetune/issues
 Depends: 
     R (>= 3.5),
-    tune (>= 1.1.2.9011)
+    tune (>= 1.1.2.9020)
 Imports: 
     cli,
     dials (>= 0.1.0),
diff --git a/NEWS.md b/NEWS.md
index 55f5ce5..6fe8a90 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -10,12 +10,11 @@
 
 * Enabling the `verbose_elim` control option for `tune_race_anova()` will now additionally introduce a message confirming that the function is evaluating against the burn-in resamples.
 
-* Updates based on the new version of tune, primarily for survival analysis models. 
+* Updates based on the new version of tune, primarily for survival analysis models (#104). 
 
 ## Breaking Change
 
-* `show_best.tune_race()` gains an `eval_time` argument for censored regression models. This breaks passing `n` by position (#104).
-
+* Ellipses (...) are now used consistently in the package to require optional arguments to be named. `collect_predictions()`, `collect_metrics()` and `show_best()` methods previously had ellipses at the end of the function signature  that have been moved to follow the last argument without a default value. Optional arguments previously passed by position will now error informatively prompting them to be named (#105).
 
 # finetune 1.1.0
 
diff --git a/R/racing_helpers.R b/R/racing_helpers.R
index 07f92c2..5f47a9b 100644
--- a/R/racing_helpers.R
+++ b/R/racing_helpers.R
@@ -665,12 +665,13 @@ randomize_resamples <- function(x) {
 #' @name collect_predictions
 collect_predictions.tune_race <-
   function(x,
+           ...,
            summarize = FALSE,
            parameters = NULL,
-           all_configs = FALSE,
-           ...) {
+           all_configs = FALSE) {
+    rlang::check_dots_empty()
     x <- dplyr::select(x, -.order)
-    res <- NextMethod(summarize = summarize, parameters = parameters)
+    res <- collect_predictions(x, summarize = summarize, parameters = parameters)
     if (!all_configs) {
       final_configs <- subset_finished_race(x)
       res <- dplyr::inner_join(res, final_configs, by = ".config")
@@ -681,10 +682,11 @@ collect_predictions.tune_race <-
 #' @inheritParams tune::collect_metrics
 #' @export
 #' @rdname collect_predictions
-collect_metrics.tune_race <- function(x, summarize = TRUE, all_configs = FALSE, ...) {
+collect_metrics.tune_race <- function(x, ..., summarize = TRUE, type = c("long", "wide"), all_configs = FALSE) {
+  rlang::check_dots_empty()
   x <- dplyr::select(x, -.order)
   final_configs <- subset_finished_race(x)
-  res <- NextMethod(summarize = summarize, ...)
+  res <- collect_metrics(x, summarize = summarize, type = type)
   if (!all_configs) {
     final_configs <- subset_finished_race(x)
     res <- dplyr::inner_join(res, final_configs, by = ".config")
@@ -704,7 +706,13 @@ collect_metrics.tune_race <- function(x, summarize = TRUE, all_configs = FALSE,
 #' resampled). Comparing performance metrics for configurations averaged with
 #' different resamples is likely to lead to inappropriate results.
 #' @export
-show_best.tune_race <- function(x, metric = NULL, eval_time = NULL, n = 5, ...) {
+show_best.tune_race <- function(x,
+                                ...,
+                                metric = NULL,
+                                eval_time = NULL,
+                                n = 5,
+                                call = rlang::current_env()) {
+  rlang::check_dots_empty()
   if (!is.null(metric)) {
     # What was used to judge the race and how are they being sorted now?
     metrics <- tune::.get_tune_metrics(x)
@@ -722,7 +730,7 @@ show_best.tune_race <- function(x, metric = NULL, eval_time = NULL, n = 5, ...)
   x <- dplyr::select(x, -.order)
   final_configs <- subset_finished_race(x)
 
-  res <- NextMethod(metric = metric, eval_time = eval_time, n = Inf, ...)
+  res <- NextMethod(metric = metric, eval_time = eval_time, n = Inf, call = call)
   res$.ranked <- 1:nrow(res)
   res <- dplyr::inner_join(res, final_configs, by = ".config")
   res$.ranked <- NULL
diff --git a/R/tune_race_anova.R b/R/tune_race_anova.R
index 86480b2..a979930 100644
--- a/R/tune_race_anova.R
+++ b/R/tune_race_anova.R
@@ -303,8 +303,8 @@ tune_race_anova_workflow <-
 
     opt_metric_time <- tune::first_eval_time(
       metrics,
-      opt_metric_name,
-      eval_time,
+      metric = opt_metric_name,
+      eval_time = eval_time,
       call = call
     )
 
diff --git a/R/tune_race_win_loss.R b/R/tune_race_win_loss.R
index 1c67d9d..8f4c0e8 100644
--- a/R/tune_race_win_loss.R
+++ b/R/tune_race_win_loss.R
@@ -283,8 +283,8 @@ tune_race_win_loss_workflow <-
 
     opt_metric_time <- tune::first_eval_time(
       metrics,
-      opt_metric_name,
-      eval_time,
+      metric = opt_metric_name,
+      eval_time = eval_time,
       call = call
     )
 
diff --git a/R/tune_sim_anneal.R b/R/tune_sim_anneal.R
index 4cb6e0b..6e4f190 100644
--- a/R/tune_sim_anneal.R
+++ b/R/tune_sim_anneal.R
@@ -344,8 +344,8 @@ tune_sim_anneal_workflow <-
     eval_time <- tune::check_eval_time_arg(eval_time, metrics, call = call)
     opt_metric_time <- tune::first_eval_time(
         metrics,
-        opt_metric_name,
-        eval_time,
+        metric = opt_metric_name,
+        eval_time = eval_time,
         call = call
       )
 
diff --git a/man/collect_predictions.Rd b/man/collect_predictions.Rd
index 9e4c59d..9d35d4b 100644
--- a/man/collect_predictions.Rd
+++ b/man/collect_predictions.Rd
@@ -8,18 +8,26 @@
 \usage{
 \method{collect_predictions}{tune_race}(
   x,
+  ...,
   summarize = FALSE,
   parameters = NULL,
-  all_configs = FALSE,
-  ...
+  all_configs = FALSE
 )
 
-\method{collect_metrics}{tune_race}(x, summarize = TRUE, all_configs = FALSE, ...)
+\method{collect_metrics}{tune_race}(
+  x,
+  ...,
+  summarize = TRUE,
+  type = c("long", "wide"),
+  all_configs = FALSE
+)
 }
 \arguments{
 \item{x}{The results of \code{\link[tune:tune_grid]{tune_grid()}}, \code{\link[tune:tune_bayes]{tune_bayes()}}, \code{\link[tune:fit_resamples]{fit_resamples()}},
 or \code{\link[tune:last_fit]{last_fit()}}. For \code{\link[tune:collect_predictions]{collect_predictions()}}, the control option \code{save_pred = TRUE} should have been used.}
 
+\item{...}{Not currently used.}
+
 \item{summarize}{A logical; should metrics be summarized over resamples
 (\code{TRUE}) or return the values for each individual resample. Note that, if \code{x}
 is created by \code{\link[tune:last_fit]{last_fit()}}, \code{summarize} has no effect. For the other object
@@ -34,7 +42,11 @@ if \code{tune("my_param")} was used).}
 configurations or just those that made it to the end of the race (the
 default).}
 
-\item{...}{Not currently used.}
+\item{type}{One of \code{"long"} (the default) or \code{"wide"}. When \code{type = "long"},
+output has columns \code{.metric} and one of \code{.estimate} or \code{mean}.
+\code{.estimate}/\code{mean} gives the values for the \code{.metric}. When \code{type = "wide"},
+each metric has its own column and the \code{n} and \code{std_err} columns are removed,
+if they exist.}
 }
 \value{
 A tibble. The column names depend on the results and the mode of the
diff --git a/man/show_best.Rd b/man/show_best.Rd
index 0a52e56..c8fc257 100644
--- a/man/show_best.Rd
+++ b/man/show_best.Rd
@@ -4,11 +4,25 @@
 \alias{show_best.tune_race}
 \title{Investigate best tuning parameters}
 \usage{
-\method{show_best}{tune_race}(x, metric = NULL, eval_time = NULL, n = 5, ...)
+\method{show_best}{tune_race}(
+  x,
+  ...,
+  metric = NULL,
+  eval_time = NULL,
+  n = 5,
+  call = rlang::current_env()
+)
 }
 \arguments{
 \item{x}{The results of \code{\link[tune:tune_grid]{tune_grid()}} or \code{\link[tune:tune_bayes]{tune_bayes()}}.}
 
+\item{...}{For \code{\link[tune:select_by_one_std_err]{select_by_one_std_err()}} and \code{\link[tune:select_by_pct_loss]{select_by_pct_loss()}}, this
+argument is passed directly to \code{\link[dplyr:arrange]{dplyr::arrange()}} so that the user can sort
+the models from \emph{most simple to most complex}. That is, for a parameter \code{p},
+pass the unquoted expression \code{p} if smaller values of \code{p} indicate a simpler
+model, or \code{desc(p)} if larger values indicate a simpler model. At
+least one term is required for these two functions. See the examples below.}
+
 \item{metric}{A character value for the metric that will be used to sort
 the models. (See
 \url{https://yardstick.tidymodels.org/articles/metric-types.html} for
@@ -23,12 +37,7 @@ default will automatically use the first evaluation time used by \code{x}.}
 
 \item{n}{An integer for the maximum number of top results/rows to return.}
 
-\item{...}{For \code{\link[tune:select_by_one_std_err]{select_by_one_std_err()}} and \code{\link[tune:select_by_pct_loss]{select_by_pct_loss()}}, this
-argument is passed directly to \code{\link[dplyr:arrange]{dplyr::arrange()}} so that the user can sort
-the models from \emph{most simple to most complex}. That is, for a parameter \code{p},
-pass the unquoted expression \code{p} if smaller values of \code{p} indicate a simpler
-model, or \code{desc(p)} if larger values indicate a simpler model. At
-least one term is required for these two functions. See the examples below.}
+\item{call}{The call to be shown in errors and warnings.}
 }
 \description{
 \code{\link[=show_best]{show_best()}} displays the top sub-models and their performance estimates.