Dreamhack Valencia is coming up this weekend, and in preparation for that, I fired up Spawning Tool for some analysis. Check out a few observations I made in a TL post http://www.teamliquid.net/forum/viewmessage.php?topic_id=421974
Because Dreamhack releases replays after their events, we actually have a ton of data available for each of these players. I went through the players seeded into Group Stage #2 and checked their performance at Dreamhack Summer. Here’s what I found:
1. Lucifron beat Sonder 2-0 using almost the exact same opening in both games. He opened with 2 Reapers into 3 CC and Hellions. The builds only start to diverge around 7:00 (ref)
2. Harstem went 4-4 (50%) in PvZ. He cannon rushed in 3 of his 4 wins, and in none of his losses (ref)
3. TheStC only played TvT in 1 series, going mech in both games against Morrow in by far his longest games (24 and 31 minutes) (ref)
4. Hyun went 7-1 (82.5%) in ZvT (ignore the issues parsing gameheart games). The games lasted between 9 and 19 minutes, with his 1 loss coming against TheStC in the longest game. His 3 fastest wins, unsurprisingly, came off of Roach-Zergling all-ins (ref)
5. StarNaN went 0-6 in PvZ. Granted, it was against Hyun, Life, and Pig, but YugiOh isn’t going to be much easier (ref)
6. Stephano went 6-2 (75%) in ZvZ, losing only to Hyun. He doesn’t build very many Roaches but does build a lot of Zerglings(ref)
7. Tefel went 8-1 (89%) in ZvP, losing only to Harstem’s cannon rush. He went Roach/Hydra in ALL of these games (ref)
8. NightEnD went 3-1 (75%) in PvZ, using a Gateway Expand build in all of them (ref)
I thought those were pretty interesting, but they also show off some of what you can do with the research tool. You might have noticed that this looks a lot different from the old research tool. After getting some feedback about the valuable parameters and developing an approachable interface, I scaled back the page a bit. The fancier bits about searching over the build orders are still around in advanced search, though I’ll need to revisit that code, too, probably.
Oddly, my work recently has been at 2 extremes. There was a lot of parsing stuff I needed to go through that basically just result in more accurate data and low-level extraction of data and tags. On the other extreme, I worked with Julie to get a splash of purple and revised interfaces. The result is that the actual core functionality of the site (somewhere between those 2 ends) didn’t change much, so not too many new features to hawk. Despite that, the site is in much better shape than before.
Anyways, enjoy Dreamhack. I think we’re on the cusp of some really exciting developments with Spawning Tool as we get more replays and hit the core functionality.