The Malmo mission/world is launched in step 1. In step 2, MALMO_HPROC is the history processor configuration. You can refer to step 6 of the previous recipe for the sample configuration. Once the agent is subjected to evaluation, you should see the results as shown here:

For every mission evaluation, we calculate the reward score. A positive reward score indicates that the agent has reached the target. At the end, we calculated the average reward score of the agent. 

In the preceding screenshot, we can see that the agent has reached the target. This is the ideal target position, no matter how the agent decides to move across the block. After the training session, the agent will form a final policy, which the agent can use to reach the target without falling into lava. The evaluation process will ensure that the agent is trained enough to play the Malmo game on its own.