Is the 'first boss attempt' phenomenon know to occur amongst NN playing games, or is this learning trajectory unique to human players?[D]

I'm curious about wether this unusual learning trajectory observed in humans has also been observed in artificial neural nets. A well known phenomenon in the 'dark souls' video game series is that ones first attempt at a boss is often much better than subsequent attempts. Boss hp at time of death by attempt might go something like: 35%, 55%, 85%, 87%, 75%, 54%, , 60%, 43%, 27%, 38%, 12%, 0%. This sounds very anecdotal, but its know to the community …

games human learning machinelearning playing