PyCon 2016 APAC Kakao 'CODING BATTLE Rock-Paper-Scissors' Recap: Day 1 Random Strategy Dominance and Day 2 Abuse Mitigation

The CODING BATTLE Rock-Paper-Scissors event held at the Kakao booth during PyCon 2016 APAC transitioned from an initial cloud-based HTTP communication model to a method where participants submitted their show_me_the_hand function via email for execution on a local server.

Strategy Analysis: The Strength of Randomness

On Day 1, strategies using the random module achieved a complete victory. The winner (k239507) used a simple probability adjustment strategy that only played Scissors and Rock, which was evaluated as the only strategy to beat random. On Day 2, the use of the random module was prohibited, with 164 participants playing a total of 13,366 games. The winner of that day (qkqnrpa) used a simple index-based strategy that played the counter-hand based on the opponent's previous move. Participants in the TOP 5 win rates (62~71%) also employed simple approaches that were intuitive or slightly strategic, rather than complex analysis.

Abuse Attempts and Security Response

Participants attempted abuse by using operator overloading (==, __str__) to check the opponent's hand, or by manipulating the module loader to forcibly replace other participants' functions. In response, the isolated process approach proposed by ganadist was applied. Each participant's module is loaded in an independent process and interacts via pipe communication, thereby fundamentally blocking cross-module intrusion or function theft. Additionally, cases where participants violated the rule prohibiting the use of the random module by pre-loading and training local modules were disqualified.