T

tronbot

Settings | Report Duplicate

0

I Use This!

Inactive

Commits : Listings

Analyzed 12 minutes ago. based on code collected about 3 hours ago.

Commit Message	Contributor	Files Modified	Lines Added	Lines Removed	Code Location	Date
Apr 20, 2023 — Apr 20, 2024 Showing page 2 of 3 Search / Filter on:
O(1) radix priority queue dijkstra's implementation. MUCH FASTER.	Andy Sloane	More...				about 14 years ago
O(n log n) dijkstra's using std::set.	Andy Sloane	More...				about 14 years ago
reweighting a little bit	Andy Sloane	More...				about 14 years ago
adjusted heuristics based on analysis of 11691 top-100-player games	Andy Sloane	More...				about 14 years ago
some junk to examine existing games; trying to come up with better heuristics	Andy Sloane	More...				about 14 years ago
a bunch of cleanup as a result of some experimentation; nothing really new but evaluate is slightly faster	Andy Sloane	More...				about 14 years ago
iterative deepening max-area search for endgame	Andy Sloane	More...				about 14 years ago
more reweighting	Andy Sloane	More...				about 14 years ago
adjust heuristic weights: anti-draw was way too strong	Andy Sloane	More...				about 14 years ago
Anti-draw strategy added	Andy Sloane	More...				about 14 years ago
floodfill tiebreaker: count nodes+edges, not just edges	Andy Sloane	More...				about 14 years ago
do floodfill prediction on both sides; otherwise we run away in perfectly balanced games	Andy Sloane	More...				about 14 years ago
whoops: dijkstra needs to leave all inaccessible squares at INT_MAX	Andy Sloane	More...				about 14 years ago
fix bug in dijkstra where it would leave garbage behind	Andy Sloane	More...				about 14 years ago
fix bug in dijkstra's; potential articulation thing not such a good idea?	Andy Sloane	More...				about 14 years ago
the tiebreaker was a bit too strong.	Andy Sloane	More...				about 14 years ago
oops, verbose	Andy Sloane	More...				about 14 years ago
potential articulation points in our controlled territory are now a negative tiebreaker in the heuristic	Andy Sloane	More...				about 14 years ago
no don't. draw penalties are bad against good players.	Andy Sloane	More...				about 14 years ago
draw penalty -> -500 for now	Andy Sloane	More...				about 14 years ago
avoid potential articulation points in greedy floodfill	Andy Sloane	More...				about 14 years ago
the O(1) connected component update code was totally wrong.	Andy Sloane	More...				about 14 years ago
use an upper bound for our opponent's floodfill abilities while using an accurate estimate for our own	Andy Sloane	More...				about 14 years ago
new evaluation function. faster and smarter!	Andy Sloane	More...				about 14 years ago
connected component checking is now O(mn) for m components instead of O(n^2)	Andy Sloane	More...				about 14 years ago
- make killer heuristic work correctly across deepenings - spend 3 seconds searching for the first move	Andy Sloane	More...				about 14 years ago
killer heuristic buffer wasn't big enough. OOPS!	Andy Sloane	More...				about 14 years ago
switch setlinebuf to just using fflush?	Andy Sloane	More...				about 14 years ago
move "10% over time" warning into #ifdef VERBOSE	Andy Sloane	More...				about 14 years ago
bugfix: timeout leaves board in bad state (but since it's a timeout it probably doesn't matter?)	Andy Sloane	More...				about 14 years ago

←
1
2
3
→