Make the bot build a sensible deck #1

New Issue

Open

opened 2019-06-26 12:12:29 +00:00 by hamcha · 0 comments

hamcha commented

2019-06-26 12:12:29 +00:00

Owner

The bots in drafting sessions should pick cards according to something and not just at random.

This is the current idea for long-term bot AI:

Pick the first 2-color rare you find, those colors will be the basis for your deck
Pick cards of the chosen 2 colors, or a third if you can't find anything.
Make sure you have enough entry cards. While in the future this might also mean color-fixing cards, for now let's just focus on 0-color req friends in the chosen colors.
Pick problems that are compatible with the colors, falling back on generic if can't find anything
Pick events and resources in color or colorless if an arbitrary friend quota was met

To make drafting session interestings, these factors should not be taken into consideration like a plain algorithm (pick rare -> pick entry -> pick friends -> pick rest) but should be shaken up and randomized a little.

For example, don't let the bot pick too many entry friends in a row, the "entry needed" quota could rise as time goes on, so we avoid bots stealing all the entry friends early on and then ignoring them.

We could also make them prioritize rarity over needs. If a good rare is in color, why not think about entry friends later?

The bots in drafting sessions should pick cards according to something and not just at random. This is the current idea for long-term bot AI: 1. Pick the first 2-color rare you find, those colors will be the basis for your deck 2. Pick cards of the chosen 2 colors, or a third if you can't find anything. 3. Make sure you have enough entry cards. While in the future this might also mean color-fixing cards, for now let's just focus on 0-color req friends in the chosen colors. 4. Pick problems that are compatible with the colors, falling back on generic if can't find anything 5. Pick events and resources in color or colorless if an arbitrary friend quota was met To make drafting session interestings, these factors should not be taken into consideration like a plain algorithm (pick rare -> pick entry -> pick friends -> pick rest) but should be shaken up and randomized a little. For example, don't let the bot pick too many entry friends in a row, the "entry needed" quota could rise as time goes on, so we avoid bots stealing all the entry friends early on and then ignoring them. We could also make them prioritize rarity over needs. If a good rare is in color, why not think about entry friends later?