My dad's side of the family holds an annual summer Family Reunion vacation for a whole week. In 2022, while on vacation, my dad taught me how to play Cribbage. He then mentioned that my grandma (his mother) was the best player he knew.
Challenge: build a Cribbage-playing bot good enough to beat my grandma.
(It doesn't have to be perfect. It just has to work well enough.)
- Add hand worth recalculation with turn card
- Add pegging section
- Add opponent hand worth calculation (with turn card)
- Add crib worth calculation (with turn card, of course)
- Add opponent crib worth calculation (with turn card)
- Add scorekeeping, with auto-repeat until win or loss
- Design playing card images
- Implement GUI card selection
- Implement GUI worth reporting
- Design pegboard and pegs images
- Implement virtual-pegboard scorekeeping
At this stage, all that might remain would be gradual strategy improvement, spurred by observations during use and possibly community suggestions/contributions.
Update 6/26/2022
Back from family reunion. The bot beat my dad twice and my grandma once! Success! Unfortunately, it then suffered a loss against my uncle Cliff. He managed to uncover a failure case in which the bot decided to throw away a 5 into the opponent's crib (Cliff was the dealer that turn), even though I knew I'd negatively weighted doing that.
It turned out my 15% difference cutoff idea was flawed. If I remember correctly, I think I may have had a Ace, 2, 3, 4, 5, and 9 that round. The bot decided to keep the Ace, 2, 3, and 9 (for a run of 3 and a 15). The better move would have been to keep the 2, 3, 4, 5 (for a run of 4, but at least you're not throwing away a 5 now).
After thinking it through, I realized that any hands worth 4 wouldn't have even been considered for bonus points in that round, because the 15% cutoff was too small. The bot considered the hand worth 5 points to be the only option. I am now removing the percent-difference cutoff system, in favor of a composite-score sort, in an attempt to lend a little more consideration to the bonus points.
Update 9/25/2022
Played against my younger sister Mary for the second time. She's been learning from our dad how to play. She faced off against the bot once before, and lost. This time she got two great hands and one fantastic crib, and beat the bot.
It was her crib, and the bot threw away a 10 and a 4. She threw away a 10 and another 10. The turn card was an Ace. Thats 12 points in a single crib. She got more than enough points from that to go out.
Update 9/9/2023
After quite a long time occupied by other activities, including school, my first Computer Science-related internship, and several other hobby projects (some of which are also still works in progress at this date), I have returned to testing and tinkering with CribEDGE. My thanks go out to the kind folks at Daily Cribbage Hand for providing me with test scenarios and feeback. I have been able to use their advice to inform several small improvements to the hand-selection algorithm.
I have also been pondering methods of implementing a decent pegging algorithm, and I now have some workable ideas forming in my head. I may also be able to enlist my dad in the effort of wrtiting this one out. No idea when we'll actually get around to this, but it'll happen eventually.
Update 6/16/2024
Took a break from working on this for quite a while so I could focus on completing my Bachelor's degree. Just recently returned from our 2024 family reunion. Spent some time discussing strategy with my dad and worked out the contents of these notebook pages. Implementing those ideas is now in progress.