The Galcon Prize: mid-contest results
“The Galcon Prize is a once-in-a-lifetime chance to improve the Galcon rankings system! Galcon is a winner-takes-all multi-player game for 2-4 players. Entries are to be written in python. The prize is an 8GB iPod Touch. Entries must be uploaded by Oct 5.”
I’ve updated the dataset with way more data … so you’ll want to download that.
Here are the mid-contest results:
1. rafsoaken 452.44 0:20 [-407,836] 2. codepenguin 452.42 0:20 [431,1392] 3. mewo2 452.36 0:19 [167,1601] 4. azbuky 451.46 0:16 [54,903] 5. fl0yd13 450.88 0:14 [103,7400] 6. paulatreides 449.94 0:16 [0,24990] 7. elo 449.93 0:18 [415,1352] 8. adamzap 446.13 0:18 [0,7] 9. simple 445.15 0:13 [-9922,6624] 10. mithrandir 440.20 0:28 [-33,1019] 11. percent 438.49 0:14 [0,100] 12. kester 424.54 0:13 [1,502] 13. rounds 387.33 0:12 [1,10710] 14. noop 317.46 0:11 [0,0] 15. bad 224.82 0:12 [-3803,0]
So .. It looks like the results are REALLY close .. The scoring is basically a percentage of how often the algorithm guesses correctly. So all the top entries are in the 45% correct zone. This is a nice improvement over my current algorithm which is at 43% correct.
Also, be sure to to keep uploading your entries throughout the week, I’ll try and post an update every day or so with the current standings so you can see how you are doing. I may be testing your entries against NEW data that you will not have access to, in order to test how well your algorithm handles data it hasn’t previously trained against. So watching for updates on the results will be quite valuable.
Also, rafsoaken sent me a histogram viewer function, which I’ve added to gprize.py. Run it with -histogram to see a visual of how your numbers are spreading. I’ll probably be using that to help me decide if an entry will really “work” for Galcon.