diff --git a/README.md b/README.md
index 783f996..a2a8b99 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,31 @@
+# Installation
+
+Assumes a clean install of Ubuntu 18.04 LTS
+
+- Install deps
+
+```bash
+sudo apt update
+sudo apt install python3-dev python3-pip
+sudo pip3 install -U virtualenv  # system-wide install
+```
+
+- Create virtualenv
+
+```bash
+virtualenv --system-site-packages -p python3 ./p4e_env
+```
+
+- Source environment and install deps
+
+```bash
+pip3 install --upgrade tensorflow(-gpu)
+pip3 install keras-rl
+pip3 install numpy
+pip3 install 'gym[box2d]'
+pip3 install 'gym[classic_control]'
+```
+
 # Framework agnostic example scripts
 
 
diff --git a/evaluate_algorithm.py b/evaluate_algorithm.py
index 1ec2562..1c1de9d 100644
--- a/evaluate_algorithm.py
+++ b/evaluate_algorithm.py
@@ -18,4 +18,4 @@
 
     evaluation_results = pipeline.run(100)
     print(evaluation_results.records) # the raw curves
-    print(evaluation_results.get_performance()) # the summary
\ No newline at end of file
+    print(evaluation_results.get_performance()) # the summary
diff --git a/prototype4evaluation/Pipeline.py b/prototype4evaluation/Pipeline.py
index 5ba4352..410fe25 100644
--- a/prototype4evaluation/Pipeline.py
+++ b/prototype4evaluation/Pipeline.py
@@ -12,6 +12,7 @@ def __init__(self,
         self.algorithm_arguments = algorithm_arguments
         self.evaluation_constructor = evaluation_constructor
         self.evaluation_arguments = evaluation_arguments
+        self.repeats = repeats
 
     def _run_parallel(self):
         raise NotImplementedError('Future work.')
@@ -19,8 +20,8 @@ def _run_parallel(self):
     def run(self, training_steps):
         evaluation = self.evaluation_constructor(**self.evaluation_arguments)
 
-        for rank in range(3):
-            print('Replicate {}/{}'.format(rank+1, 3))
+        for rank in range(self.repeats):
+            print('Repeat {}/{}'.format(rank+1, 3))
             algorithm = self.algorithm_constructor(evaluation.environment_details,
                                                    **self.algorithm_arguments)