diff --git a/Project.toml b/Project.toml
index d51fcb3..e0b949e 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "MLJTuning"
 uuid = "03970b2e-30c4-11ea-3135-d1576263f10f"
 authors = ["Anthony D. Blaom <anthony.blaom@gmail.com>"]
-version = "0.5.2"
+version = "0.5.3"
 
 [deps]
 ComputationalResources = "ed09eef8-17a6-5b46-8889-db040fac31e3"
diff --git a/src/selection_heuristics.jl b/src/selection_heuristics.jl
index eb0f469..dc6ea36 100644
--- a/src/selection_heuristics.jl
+++ b/src/selection_heuristics.jl
@@ -39,9 +39,13 @@ with measures that are neither `:loss` nor `:score` are reset to zero.
 struct NaiveSelection <: SelectionHeuristic
     weights::Union{Nothing, Vector{Real}}
 end
-NaiveSelection(; weights=nothing) =
-    NaiveSelection(weights)
-
+function NaiveSelection(; weights=nothing)
+    if weights isa Vector
+        all(x -> x >= 0, weights) ||
+            error("`weights` must be non-negative. ")
+        end
+    return NaiveSelection(weights)
+end
 
 function best(heuristic::NaiveSelection, history)
     first_entry = history[1]
@@ -49,10 +53,6 @@ function best(heuristic::NaiveSelection, history)
     weights = measure_adjusted_weights(heuristic.weights, measures)
     measurements = [weights'*(h.measurement) for h in history]
     measure = first(history).measure[1]
-    if orientation(measure) == :score
-                               measurements = -measurements
-
-    end
     best_index = argmin(measurements)
     return history[best_index]
 end
diff --git a/test/selection_heuristics.jl b/test/selection_heuristics.jl
index ebad203..2596a27 100644
--- a/test/selection_heuristics.jl
+++ b/test/selection_heuristics.jl
@@ -1,5 +1,39 @@
+using .Models
+
 measures = [accuracy, confmat, misclassification_rate]
 @test MLJTuning.measure_adjusted_weights([2, 3, 4], measures) == [-2, 0, 4]
 @test MLJTuning.measure_adjusted_weights(nothing, measures) == [-1, 0, 0]
 @test_throws(DimensionMismatch,
              MLJTuning.measure_adjusted_weights([2, 3], measures))
+
+@testset "losses/scores get minimized/maximimized" begin
+    bad_model = KNNClassifier(K=100)
+    good_model = KNNClassifier(K=5)
+
+    am = [accuracy, misclassification_rate]
+    ma = [misclassification_rate, accuracy]
+
+    # scores when `weights=nothing`
+    history = [(model=bad_model, measure=am, measurement=[0, 1]),
+               (model=good_model, measure=am, measurement=[1, 0])]
+    @test MLJTuning.best(NaiveSelection(), history).model == good_model
+
+    # losses when `weights=nothing`
+    history = [(model=bad_model, measure=ma, measurement=[1, 0]),
+               (model=good_model, measure=ma, measurement=[0, 1])]
+    @test MLJTuning.best(NaiveSelection(), history).model == good_model
+
+    # mixed case favouring the score:
+    weights = [2, 1]
+    history = [(model=bad_model, measure=am, measurement=[0, 0]),
+               (model=good_model, measure=am, measurement=[1, 1])]
+    heuristic = NaiveSelection(weights=weights)
+    @test MLJTuning.best(heuristic, history).model == good_model
+
+    # mixed case favouring the loss:
+    weights = [1, 2]
+    history = [(model=bad_model, measure=am, measurement=[1, 1]),
+               (model=good_model, measure=am, measurement=[0, 0])]
+    heuristic = NaiveSelection(weights=weights)
+    @test MLJTuning.best(heuristic, history).model == good_model
+end