Twitch Plays Pokémon beats Pokémon
by Ian Cutress on March 1, 2014 4:27 AM EST
A departure from the normal tech news, but for those not keeping tabs on the phenomenon Twitch Plays Pokemon, it is an interesting social experiment via the Twitch platform. Users are encouraged to help enter commands into the chat box to control around Red from the first generation of the Pokémon games. At the peak, an army of 100,000 users were issuing commands to the in-game character where to go. A multitude of hilariousness has come forth due to this method, as well as frustration as users with the intent of causing havoc also enter the mix (the use of the start button was adjusted, for example).
Aside from issues such as taking six hours to navigate a ledge and accidentally releasing 12 Pokémon in a single day, the extraordinary happened: millions of users at millions of keyboards managed to beat the final sequence. After 22 attempts at the Elite Four, that combination of users powered though. Not only this, but the fear of accidentally starting a new game after the end sequence came to naught.
The final moment was captured in all its glory:
Many thanks to James Croft for the image
The task took 16 days 7 hours 45 minutes, or 391 hours in total. This is compared to my individual personal best of 13 hours, to put it into perspective. The final team of Pokémon were as follows:
Omastar, Level 51
Pidgeot, Level 69
Venomoth, Level 39
Zapdos, Level 81
Lapras, Level 30
Nidoking, Level 53
For a brief update of the adventure, this Google document was maintained throughout the journey.
No news as to the next quest in the ‘Twitch Plays’ series, however the channel states that the next journey should start around midday Sunday GMT.
Facebook
Twitter
RSS
12 Comments
View All Comments
inighthawki - Sunday, March 2, 2014 - link
I completely disagree. this experiment, at best, showed that pokemon can be beaten by seemingly random input with the occasional requirement of democracy to handle the cases where it can't.With the sheer number of people providing input, the bias/intentions of the users, and the lag during the stream, the input into the game is statistically random. It shows absolutely nothing for collaborative effort.
mkozakewich - Tuesday, March 4, 2014 - link
Not even. The Random-Number-Generator version didn't even get to the first town after a few months. There was another version with some bias introduced in order to guide it toward its goals, but even then they needed someone to take over for a few spots.The kind of randomness from this system is like the randomness of growing slime moulds or something. They've got a direction they're going, and they can make it there in very little time. The biggest problem was the 20-second lag, which caused most commands to overshoot. If every command was instant, we'd actually have an interestingly-high accuracy. I was looking out for trolls when we were in democracy mode, but I saw exactly what I'd expect from a bunch of unintelligent but well-meaning players.
The lag was always the biggest issue with control, because the actions people inputted would only be executed 20-30 seconds later, and so they'd overshoot; and when they tried to correct they'd correct too far back. They didn't learn very well, either, and it remains a problem even as we play through the second generation.