diff --git a/spacy/_ml.pyx b/spacy/_ml.pyx
index 5bfa05573..706396fd1 100644
--- a/spacy/_ml.pyx
+++ b/spacy/_ml.pyx
@@ -29,22 +29,25 @@ cdef int arg_max(const weight_t* scores, const int n_classes) nogil:
 cdef int arg_max_if_true(const weight_t* scores, const int* is_valid,
                          const int n_classes) nogil:
     cdef int i
-    cdef int best = 0
-    cdef weight_t mode = -900000
+    cdef int best = -1
+    cdef weight_t mode = 0
     for i in range(n_classes):
-        if is_valid[i] and scores[i] > mode:
+        if is_valid[i] and (best == -1 or scores[i] > mode):
             mode = scores[i]
             best = i
     return best
 
 
+class ValidationError(Exception):
+    pass
+
 cdef int arg_max_if_zero(const weight_t* scores, const int* costs,
                          const int n_classes) nogil:
     cdef int i
-    cdef int best = 0
-    cdef weight_t mode = -900000
+    cdef int best = -1
+    cdef weight_t mode = 0
     for i in range(n_classes):
-        if costs[i] == 0 and scores[i] > mode:
+        if costs[i] == 0 and (best == -1 or scores[i] > mode):
             mode = scores[i]
             best = i
     return best
@@ -63,13 +66,18 @@ cdef class Model:
             self._model.load(self.model_loc, freq_thresh=0)
 
     def predict(self, Example eg):
+        assert self.n_classes == eg.c.nr_class
         memset(eg.c.scores, 0, sizeof(weight_t) * eg.c.nr_class)
         self.set_scores(eg.c.scores, eg.c.atoms)
         eg.c.guess = arg_max_if_true(eg.c.scores, eg.c.is_valid, self.n_classes)
+        if eg.c.guess == -1:
+            raise ValidationError("No valid classes during prediction")
 
     def train(self, Example eg):
         self.predict(eg)
         eg.c.best = arg_max_if_zero(eg.c.scores, eg.c.costs, self.n_classes)
+        if eg.c.best == -1:
+            raise ValidationError("No zero-cost classes during training.")
         eg.c.cost = eg.c.costs[eg.c.guess]
         self.update(eg.c.atoms, eg.c.guess, eg.c.best, eg.c.cost)