1 / 3

Challenges of Playing Bridge Well

Challenges of Playing Bridge Well. http://www.ny-bridge.com/allevy/Montreal/indexMenton.html Bridge requires a lot of real-time decision making in uncertain environments.

maxima
Download Presentation

Challenges of Playing Bridge Well

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Challenges of Playing Bridge Well • http://www.ny-bridge.com/allevy/Montreal/indexMenton.html • Bridge requires a lot of real-time decision making in uncertain environments. • The distribution of the opponents’ and partner’s cards is initially unknown and follows the laws of probability. However, as bids are made and cards are played these initial probabilities change  quite difficult to automate (belief networks). • Defenders have to collaborate and employ complex signals for this purpose. • There is a huge amount of domain specific knowledge on how to bid and how to play the hand  difficult to computerize • Declarer play requires planning! • Determining the best plan to play a hand is very difficult; determining the best linr of play with all hands visible is easy  Deep Finessehttp://www.deepfinesse.com/ does the job. • Players false card or do others things to mislead the declarer (and frequently also partner).

  2. Thoughts Concerning the King+Rook_vs_Rook Problem • The task is to solve the problem at hand • Hill-Climbing, Randomized Hill Climbing, Backtracking, Best-first Search,… all seem to be suitable search strategies for the problem • Find “good” evaluation functions” (e.g. f(pos)=king_close_w_king(pos)*a + king_close_to_rook(pos)*b + c*1/Freedom_black_king(pos)) for one of the available search strategies • Rapid prototyping is strongly recommended; also don’t make things too complicated at the beginning • Try to learn from sample runs • Idea: break the problem into sub-problems and solve each sub-problem separately. • Idea: Reduce the search space by disallowing certain moves • Idea: Use look-ahead when you use backtracking or hill climbing • You are allowed to take advantage of the face that “you know in advance how the black king moves”. It is okay if you implement a game-style system but you should be aware of the fact that you are trying to solve a more complicated problem which makes it much harder to get your system running. • “Complex” search strategies are frequently hard to debug/improve.

  3. Thoughts Concerning the King+Rook_vs_Rook Problem (cont.) • Instead of looking to all possible positions sample some of those to reduce the complexity of the search • Checking for duplicate states might or might not help (this answer is dependent on the employed search strategy) • Disjunctive goals in search create problems • Taking advantage of symmetry is complicated, but might help! • You will receive a file consisting of 10 training cases no later than September 19, 2001. There will be a demo where your program will be run for 10 test cases that are somewhat similar to the ten training test cases. • Test cases will be sorted by difficulty; if you program cannot solve all the test problems this is not the end of the world; however, if it is not able to solve the “simple” cases you will not get a good grade for project. • The primary goal is to checkmate the black king; however, programs that find shorter solutions will get higher scores. • The project will be graded evaluating your report, the quality of the designed program, and the programs performance for the training and test set. Program performance for test case will count at least 55% for the overall grade.

More Related