The original version of this submission used a more complex 5-checker trap to defeat the Hard AI, but now I found a way to break it with the simple and infamous 3-checker trap. Please replace the BK2 attached to this submission with the following 90-frame improvement

Objectives

  • Uses hardest difficulty
  • Genre: Board

Difficulty

"Hard Game" is the hardest difficulty this game offers and was selected here.
Unlike the "Normal Game", the AI in the "Hard Game" will block any win in 1. It will also block the player from pulling off the infamous horizontal "3-checker trap"... but not unless we're clever in how we go about it. You see, the AI will block any 3-checker trap from forming on the bottom row just fine. It's when you try to set up the trap on anything higher than the first row where something—I'm not exactly sure what—begins to break down in the AI's decision-making. If you proceed to set up the trap on the second row by placing the two checkers on the edge of the trap first, the AI will not notice a threat and may proceed to not block the key middle space, which will result in the AI's immediate loss.
This specific version of the 3-checker trap can be set up in a minimum of 8 half-moves (i.e., 8 ply), meaning we want to be taking our turn on that 8th half-move. Therefore, letting the computer go first is optimal for this TAS.

RNG

If I didn't miss anything, I also exhaustively tested every possible alternate RNG scenario that could have possibly led to a different, faster outcome than this TAS. This included:
  • Testing different start frames of the game.
  • Testing different start frames for each of the computer's turns.
  • Every possible alternate move which would still set up the fastest 3-checker trap on the second row, including variants where the whole setup is shifted one column left or right.

nymx: Claiming for judging. --Usually I would take on four at a max, as requested by our senior staff, but I need to pair this one up with the other that winslinator submitted of this game. Seems that an interesting situation has developed.
nymx: Ok...as with the "Normal" submission, this one was also discussed with our judges. This TAS would basically qualify for "Full Completion"...meaning that among the 3 difficulty selections, "hard" presents all that the game has to offer, in terms of play-ability.
Accepting as "Full Completion" for publication.

despoa: Processing...
nymx: Replacing with an improved movie file.


TASVideoAgent
They/Them
Moderator
Joined: 8/3/2004
Posts: 15566
Location: 127.0.0.1
This topic is for the purpose of discussing #8713: Winslinator's GBA Connect Four "Hard Game" in 00:31.23
Joined: 8/16/2022
Posts: 7
Please replace the BK2 attached to this submission with the following 90-frame improvement: User movie #638348164960181448
Wins alt account
Post subject: Movie published
TASVideoAgent
They/Them
Moderator
Joined: 8/3/2004
Posts: 15566
Location: 127.0.0.1
This movie has been published. The posts before this message apply to the submission, and posts after this message apply to the published movie. ---- [5668] GBA Connect Four "Hard Game" by Winslinator in 00:31.23
Alyosha
He/Him
Editor, Emulator Coder, Expert player (3810)
Joined: 11/30/2014
Posts: 2829
Location: US
https://tasvideos.org/UserFiles/Info/638358560032196485 Here is a better strategy that wins with only 9 checkers. Basically, the computer will not let you win in 1 move if you have 3 in a row and an opportunity for 4 in a row the next move, but it will give you a chance to win on your next move if the opportunity doesn't already exist. It's possible this can also be done with only 8 checkers with a red start, I haven't checked yet, this is the first case I could get working. EDIT: Actually that's not possible since then red would get the connect 4 first, so 9 should checkers should be the minimum.
Alyosha
He/Him
Editor, Emulator Coder, Expert player (3810)
Joined: 11/30/2014
Posts: 2829
Location: US
https://tasvideos.org/UserFiles/Info/638361257271922150 Here's the current best i can get with 9 checkers. Assuming no one gets faster manip I guess I'll submit this one after I try a little more.