diff --git a/src/hepstats/hypotests/calculators/frequentist_calculator.py b/src/hepstats/hypotests/calculators/frequentist_calculator.py
index a892e6a7..7416829e 100644
--- a/src/hepstats/hypotests/calculators/frequentist_calculator.py
+++ b/src/hepstats/hypotests/calculators/frequentist_calculator.py
@@ -83,7 +83,7 @@ def qnull(
             >>> poialt = POI(mean, 1.2)
             >>> q = calc.qnull(poinull, poialt)
         """
-        toysresults = self.get_toys_null(poinull, poialt, qtilde)
+        toysresults = self.get_toys_null(poinull, poinull, qtilde)
         ret = {}
 
         for p in poinull:
diff --git a/src/hepstats/hypotests/hypotests_object.py b/src/hepstats/hypotests/hypotests_object.py
index d922954c..76091683 100644
--- a/src/hepstats/hypotests/hypotests_object.py
+++ b/src/hepstats/hypotests/hypotests_object.py
@@ -201,7 +201,6 @@ def sampler(self):
 
         >>> sampler = calc.sampler()
         """
-        self.set_params_to_bestfit()
         nevents = []
         for m, d in zip(self.loss.model, self.loss.data):
             nevents_data = get_nevents(d)
@@ -251,8 +250,7 @@ def toys_loss(self, parameter_name: str):
         Example with `zfit`:
             >>> loss = calc.toys_loss(zfit.Parameter("mean"))
         """
-        if parameter_name not in self._toys_loss:
-            parameter = self.get_parameter(parameter_name)
-            sampler = self.sampler()
-            self._toys_loss[parameter.name] = self.lossbuilder(self.model, sampler)
+        parameter = self.get_parameter(parameter_name)
+        sampler = self.sampler()
+        self._toys_loss[parameter.name] = self.lossbuilder(self.model, sampler)
         return self._toys_loss[parameter_name]
diff --git a/src/hepstats/hypotests/toyutils.py b/src/hepstats/hypotests/toyutils.py
index 5ad1fe22..868c2ea7 100644
--- a/src/hepstats/hypotests/toyutils.py
+++ b/src/hepstats/hypotests/toyutils.py
@@ -232,11 +232,18 @@ def generate_and_fit_toys(  # TODO PROFILE THIS
             poieval: POI values to evaluate the loss function
         """
 
-        self.set_params_to_bestfit()
-
         minimizer = self.minimizer
         param = poigen.parameter
 
+        # generate toys from best fit at profile point
+        param.set_value(poigen.value)
+        is_poi_floating = param.floating
+        param.floating = False
+        profile_minimum = minimizer.minimize(loss=self.loss)
+        if not profile_minimum.valid:
+            print(poigen, "PROBLEM: profile fit not valid, toys won't be either")
+        param.floating = is_poi_floating
+
         toys_loss = self.toys_loss(poigen.name)
         sampler = toys_loss.data