Interesting gaming experiment - Twitch Plays Pokemon

I guess I'll get the ball rolling!

I've checked in on it, off and on, for almost a day now. While I love the idea of it, the implementation has its flaws. For the majority of the day the character was stuck in the same area where it needed to walk along a 2 by 12 ish path without pushing down more then twice, and push up at a precise time at the end to even be on track to progress forward.

Whats became apparent was the flaws in the script and the 20+ second twitch delay. I believe the script is designed to press all of the commands sent to it by chat, and whatever one happens to be pressed first at the right frame to register is basically pure luck. Combine that with needing to predict what action might be needed 20s from now basically adds a further random element. This setup is barely better then determining input every frame by something like "inputButton = rand() % 8 + 1;". Even excluding obvious sabotage commands it's to close to pure luck to hope that the proper input will be given at the proper time.

It's not a very god example of a hive mind. Input should be determined by a majority. And any mind, hive or singular, should not fair to well with 20+ second lag from thought to action without something to help compensate. It basically needs to buffer all input messages for 20+ seconds, then select the majority. Progress would be slow, but not as slow as randomly walking around a single area for a day.

I understand the guy who made this didn't intend for it to get this popular. And I understand not wanting to create a script that's to refined as doing so could contaminate results. But upon further reflection the experiment has little to provide outside of a stress test on twitchs chat servers and "Are 20,000 people issuing different commands with a 20s+ delay barely better then random input?".

I think before this experiment became popular then it could of been beaten, but now...... not a chance. :P UNLESS somehow the majority of the decisions, for what to do/select next, were considered instead of implementing everyone's command; maybe take into account the majority vote/command every 5 seconds? .... it does implement everyone's command right? I could be wrong.

I would be in for progress but just watching the sabotage now gives me a good laugh. :lol: From the quick glimpses of the character profile/stats (forgot its proper name in the game), it looks like they have made it halfway through the game by already collecting four badges. Of course that progression most likely occurred before this experiment became popular. :|

"CoryandStuff" wrote:
I think before this experiment became popular then it could of been beaten, but now...... not a chance. :P UNLESS somehow the majority of the decisions, for what to do/select next, were considered instead of implementing everyone's command; maybe take into account the majority vote/command every 5 seconds? .... it does implement everyone's command right? I could be wrong.

Thats my understanding based of the following exert from the twitch page.

Why are buttons showing in the input feed not being registered in-game or why buttons that are being pressed not showing up in the input feed?

With the exception of start button throttling this is not intended behavior. The cause of buttons not being pressed in-game is usually because Pokemon doesn't queue button presses, this leads to a lot of button presses technically being registered by the game but ignored (for example: while walking from one space to another no inputs have any effect). Delayed/erroneous button presses are usually due to overload or other bugs with my software, I'm working on it and it has gotten better but it is still not perfect.

He's basically saying all input commands are sent to the game, but while an action is in progress inputs are ignored. It may sound better in theory to some as everyone's input gets sent to the game, so everyone feels like they're making a direct impact, but it's not. Majority buffering would have individuals making a greater contribution. Just think about a hundred people saying "right", but it registering one of the 3 people saying "start" for no good reason. For that most people aren't contributing.