Proposed features: Ai Gauntlet and Ai Grid Gauntlet

The Forge Gauntlet facility can be made to test new deck designs using two new AI vs AI features:
Features
1) Ai Gauntlet
Two AIs play against each other, where the first AI has a single deck, versus the second AI that has many decks (the Gauntlet).
The benefit of this approach is to be able to run a deck against many other decks, gather win percentages compared to other types. This helps identify deck weaknesses and strengths using mass numbers of games.
For example, play 100 matches against each type of deck in the specified Gauntlet list, completely autonomously (no clicking a button to close an informational dialog popup), and then produce a table of stats at the end.
2) Ai Grid Gauntlet
Two AIs play against each other, where the first AI has many decks (the Gauntlet), versus the second AI that has many decks (the Gauntlet).
The benefit of this approach is to be able to run many decks against many other kinds of decks to gather stats on the specified meta game. Using this analysis ability allows Gauntlet lists of the current meta game to be analyzed to see trends and changes in the meta game.
For example, play 100 matches of each deck in the specified Gauntlet list against each deck in the specified Gauntlet list, and produce a table of stats at the end.
Details
A couple of key capabilities are necessary to enable these features.
- completely autonomously (no clicking a button to close an informational dialog popup)
- run full speed (100+x) or possibly without UI animations, just status of the turns in a text log perhaps
- Gauntlet list versus Gauntlet list menu UI
- Stats grid tables to show results
Depending on how fast Forge can run through a match, I can see running millions of games to generate meta game statistics.
By having regular collection process of decks from sources like TCGPlayer, it would be possible to apply win characteristics to these decks and how they perform in the current meta game. A bulk import facility would be nice too.
This would be an incredible tool for deck builders and deck analysts.
Thoughts? Suggestions? Implementation ideas?
Features
1) Ai Gauntlet
Two AIs play against each other, where the first AI has a single deck, versus the second AI that has many decks (the Gauntlet).
The benefit of this approach is to be able to run a deck against many other decks, gather win percentages compared to other types. This helps identify deck weaknesses and strengths using mass numbers of games.
For example, play 100 matches against each type of deck in the specified Gauntlet list, completely autonomously (no clicking a button to close an informational dialog popup), and then produce a table of stats at the end.
2) Ai Grid Gauntlet
Two AIs play against each other, where the first AI has many decks (the Gauntlet), versus the second AI that has many decks (the Gauntlet).
The benefit of this approach is to be able to run many decks against many other kinds of decks to gather stats on the specified meta game. Using this analysis ability allows Gauntlet lists of the current meta game to be analyzed to see trends and changes in the meta game.
For example, play 100 matches of each deck in the specified Gauntlet list against each deck in the specified Gauntlet list, and produce a table of stats at the end.
Details
A couple of key capabilities are necessary to enable these features.
- completely autonomously (no clicking a button to close an informational dialog popup)
- run full speed (100+x) or possibly without UI animations, just status of the turns in a text log perhaps
- Gauntlet list versus Gauntlet list menu UI
- Stats grid tables to show results
Depending on how fast Forge can run through a match, I can see running millions of games to generate meta game statistics.
By having regular collection process of decks from sources like TCGPlayer, it would be possible to apply win characteristics to these decks and how they perform in the current meta game. A bulk import facility would be nice too.
This would be an incredible tool for deck builders and deck analysts.
Thoughts? Suggestions? Implementation ideas?