Two big pieces of testing news this time! First, I got a fast workstation. I can run more games against more opponents in less time. Previous test tournaments ran on a laptop that is slower than the contest machines. The workstation is faster than the contest machines. Timeouts and crashes fell from one every few hundred games to one every few thousand games. Results didn’t appear to be significantly different, but they’re definitely a trifle different. Some bots benefit (a little) from the extra thinking time. Oddshrimp isn’t one of them... yet.
Second, Fredrik Persson was kind enough to send me #3 Slin as a test opponent. With the top three contest bots, it’s clear where oddshrimp4.2 stands. Without Slin, beating iouri would not mean as much.
Some of the opponent names here still do not match the correct names on the official site. Sorry.
500 games each against old versions. Notice the declining draw count.
bot | win rate | wins | losses | draws |
oddshrimp4.1 | 57.8% | 257 | 179 | 64 |
oddshrimp3.4 | 66.6% | 311 | 145 | 44 |
oddshrimp3.3 | 65.6% | 306 | 150 | 44 |
oddshrimp3.2 | 66.5% | 317 | 152 | 31 |
oddshrimp3.1 | 68.8% | 331 | 143 | 26 |
oddshrimp2.4 | 76.7% | 373 | 106 | 21 |
400 games each against 24 opponents, 9600 games total, all on randomly-generated maps. Oddshrimp4.2 beats all except #1 bocsimacko. Compared the the previous version, it scored better against most opponents, worse only against the aggressive medrimonia and Mistmanov, and the draw rate is down.
opponent | rank | win rate | wins | losses | draws |
bocsimacko | 1 | 34.6% | 134 | 257 | 9 |
iori | 2 | 54.0% | 203 | 171 | 26 |
Slin | 3 | 53.6% | 210 | 181 | 9 |
GreenTea | 8 | 60.6% | 231 | 146 | 23 |
dmj111 | 12 | 57.0% | 225 | 169 | 6 |
davidjliu | 13 | 65.1% | 250 | 129 | 21 |
wagstaff | 17 | 67.9% | 263 | 120 | 17 |
medrimonia | 18 | 56.2% | 221 | 171 | 8 |
smloh | 19 | 81.2% | 321 | 71 | 8 |
CorwinAlex | 26 | 65.4% | 256 | 133 | 11 |
Neverstu | 28 | 75.6% | 300 | 95 | 5 |
Manwe | 31 | 71.4% | 282 | 111 | 7 |
rsergio | 33 | 71.9% | 281 | 106 | 13 |
barabanus | 41 | 77.8% | 307 | 85 | 8 |
mogron | 46 | 83.1% | 332 | 67 | 1 |
deccan | 47 | 85.4% | 337 | 54 | 9 |
rebelxt | 53 | 85.8% | 336 | 50 | 14 |
malazan | 54 | 77.5% | 304 | 84 | 12 |
fglider | 61 | 73.6% | 294 | 105 | 1 |
eAshoka | 65-ish | 84.1% | 333 | 60 | 7 |
Ice_Harley | 65 | 87.4% | 345 | 46 | 9 |
Mistmanov | 77 | 73.9% | 291 | 100 | 9 |
E323 | 78 | 83.9% | 331 | 60 | 9 |
FlagCapper | 91 | 85.1% | 333 | 52 | 15 |
Slin is a great opponent. It beats oddshrimp4.1 and is nearly as tough as iouri. But as you see above, it can’t keep up with oddshrimp4.2
Oddshrimp4.2 and iouri are so close that I had to run an extremely long match to convince myself that oddshrimp is stronger. I ran an equally long match against Slin.
opponent | rank | win rate | wins | losses | draws |
iouri | 2 | 51.1% | 9785 | 9356 | 859 |
Slin | 3 | 55.3% | 10875 | 8744 | 381 |
April 2014