Cleaning up thousands of build order openings

Posted on by
Reply

At Spawning Tool, we’re all about labeling StarCraft replays with build orders. With tens of millions of unit build events tracked, build order labels are a good way to simplify and make that data more accessible. Spawning Tool collects build order labels from various sources, including:

  1. machine learning – given many examples and counter-examples of Bio Mine builds, which unlabeled builds look the closest?
  2. programmatic logic – if there are 2 Hatcheries built before the first Spawning Pool, it’s a “3 Hatch before Pool”
  3. community-approved suggestions – the yes/no you sometimes see on replays
  4. community-contributed labels – anything else the community punches in

Our certainty about the labels follows the list above as well: we are relatively uncertain of the labels from machine learning but highly certain on those contributed by hand. The gray area, however, has always been those in the middle: how certain do we have to be to add the build orders straight into the system versus just posing it as a suggestion?

After going through the data, we determined that openings are relatively standard, and have shuffled most of those to be automatically, programmatically labeled for you. By cleaning this up, we added ~9900 new build order tags. We also removed ~5000 suggestions that were relatively obvious. Currently, the list of programmatic build order labeled are:

  • Protoss: Forge Fast Expand, Nexus First, 1 Gate MSC Expand, 1 Gate Expand
  • Terran: Proxy 2 Rax, 8/8/8 Proxy Reaper, 14 CC, 15 CC, Reaper Expand, 1 Rax Expand, 1/1/1
  • Zerg: 3 Hatch Before Pool, X Pool X Hatch, X Hatch X Pool, X Pool (10 and below)

Overall, we hope that this greatly increases the quality and accessibility of our build order data. However, the data can always be better, and we would appreciate suggestions for more build orders that we can add into the system. Our hope is to come up with a broad taxonomy for build orders to label all of them and understand the relationship between them.

And even if you don’t have any new ideas, we would appreciate any help approving existing suggestions. We still have over 10,000 undecided suggestions that all of you should feel welcome to adjudicate on.

What can we do to make the build order labels more useful?

View Results

Loading ... Loading ...

LotV Featured Replay of the Week: Geiko’s PvZ 4 Gate Adept

Posted on by
Reply

View this replay on Spawning Tool

Part of learning a new meta-game is learning new cheese strategies, and the 4 Gate Adept build popularized by Geiko is a pretty strong Protoss cheese, so here is our first Legacy of the Void (LotV) Protoss build order.

The basic gist of the strategy is to abuse the mobility and tankiness of Adepts. With the Shield Upgrade from the Twilight Council, Adepts have 90 Health and 140 Shields (and 1 armor). Although they have relatively low DPS because of a slow attack speed, their bonus damage against light units allows them to 2-shot workers, Combat Shield-less Marines, and Zerglings.

Above all that, they also have the Psionic Transfer ability to create a shade to bypass static defenses and attack in 2 locations at once. You can also cancel the shade so that Adepts can effectively threaten the main and natural bases simultaneously, forcing the defender to split up or be left completely defenseless.

The build itself is relatively straightforward, and Geiko provides a good explanation in the TL thread: open 2 Gate to harass. Get your Twilight Council for the Shield Upgrade and 2 more Gateways and place a proxy Pylon to start wailing on your opponent. With the Twilight Council, you also have the flexibility to mix in DTs, just to make their lives even more difficult.

This particular game against ViBE shows how strong this build is against Zerg. Going down the list of potential defenses,

  1. Zerglings die in 2 shots
  2. Queens aren’t numerous enough
  3. Spine Crawlers can be bypassed
  4. Roaches aren’t fast enough and actually have less HP/Shields than an upgraded Adept

Although ViBE defends well against the initial 2 Adept poke and even the first warp-in, the next wave of Mass Adepts (still off of 1 base) is too much to handle, even with burrowed Roaches.

-StoicLoofah

Geiko’s PvZ 4 Gate Adept

  • 13 Pylon
  • 14 Gateway
  • 15 Assimilator
  • 18 Gateway
  • 19 Cybernetics Core, Assimilator
  • 20 Warp Gate, Pylon at your ramp
  • 22 Adept x2 (send them to harass)
  • 27 Twilight Council
  • 27 Gateway x2 (to wall off), Shield Upgrade
  • 28 Shield Upgrade
  • 29 Proxy Pylon
  • 27 Dark Shrine
  • Then build Adepts

Featured Replay of the Week: TaeJa vs PartinG MindGames

Posted on by
Reply

View this replay on Spawning Tool

https://www.youtube.com/watch?v=w0svEpeFf4w&feature=youtu.be&t=1h1m19s

This week I had a very hard time deciding which replay to feature on the front page. I knew I wanted it to be one of the games of PartinG vs TaeJa, but I didn’t know which one. There was the double cheese game that ended in a draw, then there was the game after on the same map in which TaeJa cheesed again, and there was this game. I chose this game because TaeJa finally realizes that he needs to mix it up to beat PartinG. In a standard macro game, TaeJa was stomped by PartinG.

In this game, TaeJa opens with a gas first. When one sees this opening, he can expect to see a 1/1/1 hyper-aggressive and possible allin from the Terran player. TaeJa decides he wants to mix in some mind games on this map though. When PartinG’s probe slips past the marine and gets into the base, he sees that the factory is moments from being finished. TaeJa immediately pulls all of his SCVs from gas and pools up to 400 minerals after the probe is dealt with. He soon after puts down a 3rd Command Center, being extremely greedy.

PartinG’s reaction to the potential aggression was to constantly chrono out stalkers and the warp gate tech, delaying his own economy. This puts TaeJa ahead because of his 2 extra command centers on the way. TaeJa also moves his hellion out onto the map to trick PartinG even more. The build that TaeJa showed made PartinG almost positive that it was a 2 hellion, 6 marine, 1 mine, 1 medivac attack at about 6 minutes. When TaeJa moves his hellion across the map, he makes PartinG think for even longer that the build he scouted was accurate information.

TaeJa takes advantage of this trick and goes EXTREMELY greedy. He stays on one gas for very long just so he can get his extra barracks out sooner, then takes 3 gases in progression to tech up as fast as possible. TaeJa also takes his third at the 9:30 mark.

Based on the stats and replays we have on Spawning Tool, it shows that PartinG only ever loses games when they go past the 16 minute mark. TaeJa must have been checking up on our valuable statistics 🙂 , because he wanted to get into the mid/late game as fast as possible!

If people say that INnoVation is a robot, I say that TaeJa is a piece of clay. In game 3 on Iron Fortress, in multple cases, his medivac drops were sniped. This ultimately was too costly and was one of the main reasons why he lost that game. In this game, TaeJa, when sending out drops, would drop one marine out to scout ahead and to the third positions the find PartinG’s army. TaeJa also goes double starport this game in order to counter PartinG’s large colossus count that overwhelmed him in game 3.

The multipronged aggression/multitasking from both players throughout the game was truly impressive. PartinG sends out a warp prism with DTs and attacks mutiple bases with it. TaeJa also defends this agression while killing PartinG’s third base!

Towards the end of this game, you can see why PartinG is known as one of the best players in the world: he just refuses to die. Even when TaeJa was a base up and microing his heart out, PartinG kept on taking good trades with his units and holding the Terran forces back.

TaeJa eventually overwhelms the Protoss army and wins despite PartinG’s efforts.
This game was extremely close. It is surprising that TaeJa, even when he is half retired, can still keep up and take games off of one of, if not the best, Protoss in the world.

-MorN_sc

TaeJa vs PartinG Mind Games

  • 9  0:54  Supply Depot
  • 11  1:26  Refinery
  • 13  1:50  Barracks
  • 16  2:56  Orbital Command
  • 16  2:58  Factory
  • 17  3:14  Supply Depot
  • 21 3:59   Pulls out of gas after being scouted
  • 21  4:01  Widow Mine
  • 24  4:19  Barracks Reactor
  • 24  4:27  Command Center
  • 24  4:34  Supply Depot
  • 28  5:20  Command Center
  • 29  5:46  Bunker
  • 33  6:07  Orbital Command
  • 33  6:17  Barracks
  • 34  6:20  Barracks
  • 37  6:40  Engineering Bay

Announcing our newest (first) sponsorship: Team Gravity’s Fight Night!

Posted on by
Reply

teamgravitysfightnightbanner

This announcement comes a few days late for those who either frequent reddit or watch Fight Night, but there aren’t enough ways to say how excited we are to be sponsoring Team Gravity and their Sunday night series, “Team Gravity’s Fight Night”.

If you have been keeping up with the data on Spawning Tool, this announcement should be no surprise: Team Gravity has been phenomenal in regularly sharing the replays from Fight Night every week. They believe, like Spawning Tool, in having replay data openly available for the community and students of the game. Both through the players and events they support, Team Gravity embodies the spirit and values that we believe are necessary to maintain a vibrant StarCraft 2 scene in North America, and we are proud to be sponsoring them.

What is Spawning Tool? Spawning Tool is a website where you can easily browse pro replays by matchup, build orders, maps, or players. It makes it easy to find, learn, and evaluate new build orders and pull together interesting stats about players and upcoming matches.

What is Team Gravity’s Fight Night? Fight Night is a weekly $100 show match series consisting of two preliminary BO5, and a main event as a BO7. You can see the full list of past participants on liquipedia, but Fight Night #19 featuring TaeJa v PartinG gives you a sense of the caliber of players participating. The events are also sponsored and streamed exclusively on hitbox.tv

You can follow us at @spawningtool or Team Gravity at @TeamGrav for the latest news on what we’re up to. We all hope you are watching Fight Night on upcoming Sunday evenings!

Featured Replay of the Week: Maru’s Proxy Marauder into Bio Tank

Posted on by
2

View this replay on Spawning Tool

Game 2. Overgrowth. Grand Finals. What do you expect Maru to do? Proxy of course. Maru goes for a proxy marauder. He places it right below Dream’s base, because he expects dream’s reaper to go scout to the his third, then go to Maru’s base (which would then make his barracks unscouted). These type of attempted mindgames only take place in matches where the players know each other very well.

After dream scouts the proxy marauder, he makes the choice to send his reaper to Maru’s base instead of using it to defend. This choice would let him get scv kills, while still defending his own base by pulling scvs.

Maru ends up getting a surround on the reaper which potentially saves this game for him. He had no units on his own side of the map, so this reaper could have been game ending.

After the multitasking dies down, maru goes 3cc! In units, he has a hellion, reaper, and a marauder to defend against any aggression that Dream would throw at him. This is the definition of greed, which is not usually Maru’s play style. He is usually the one multitasking the opponent and constantly applying pressure.

One of the more impressive things in this game is how Maru defended against Dream’s medivac drop at ~ 7:00. His micro is top notch. The fact that he defended this aggression so well means that he can get away with his 3cc. Dream’s followup banshee also gets deflected by a hidden widow mine!

The game evens up after Dream gets 8 scvs in a sneaky double medivac drop as Maru moves his army out onto the map. In order to kill just a little more scvs, he does not attack with his marines until he is on top of the mineral line.

Maru actually forgets combat shields and does not have it researched until ~17:00. Dream did not capitalize on this though.

At around the 18 minute mark, dream sneaks about 30 marines into Maru’s natural and ends up killing a command center. This forces Maru into a counter attack which kills dreams 3rd command center.

These guys are trading blow for blow in the mid-game. They both managed to get maxed by the 16 minute mark, but end up at 100 supply each at the 19 minute mark.

Maru who still has an army advantage and the air advantage closes in and forces Dream into a contain. With great siege tank placement he manages to stop the mining at the natural and forces dream to try something crazy.

Dream ends up trying for a surround by dropping marines behind Maru’s army then attacking with both halves. It almost works, because he has a slight upgrade advantage, but Maru’s medivac count is what saves him. Who would have thought this all started with a scouted proxy marauder? Well, I guess Maru would have ;P

-Morn_sc

TvT Maru’s Proxy Marauder into Bio Tank

  • 10 1:00 Supply Depot (send your proxy SCV at the same time)
  • 11 1:38 Barracks
  • 13 1:54 Refinery
  • 16 2:43 Barracks Tech Lab
  • 16 2:55 Orbital Command
  • 16 3:03 Supply Depot
  • 16 3:08 Marauder
  • 21 3:50 Reaper (defensive)
  • 21 4:29 Factory
  • 21 4:57 Command Center
  • 23 5:33 Starport
  • 24 5:49 Command Center
  • 25 5:56 Supply Depot
  • 25 6:04 Engineering Bay
  • 26 6:21 Hellion
  • 26 6:26 Viking
  • 26 6:29 Engineering Bay

Nathanias reviving the iEchoic build in LotV TvZ?

Posted on by
1

So far, the metagame in the Legacy of the Void beta has been hard to decipher while players are still figuring out the game. Openings are definitely inspired by existing builds, but exact timings and supply counts don’t match. Compositions from then on out are a mish-mash of all units with multiple pivots in-between. It has been really inspiring to see how players are using new units.

As I was going through LotV replays on Spawning Tool, however, I noticed that Nathanias has done new things with old units in reviving a version of the old iEchoic build from Wings of Liberty TvT. The exact build is not too important, but looking at the build order, the important thing is that he starts with Hellions and Banshees, then transitions into Hellions and Battlecruisers. That’s basically the idea.

The build worked in TvT with the basic assumption that there are 3 units that Terran have to counter air: Marines, Thors, and Vikings. At the time, Thors didn’t have the High Impact Payload, so with enough armor on Battlecruisers, neither Thors nor Marines could really handle them, except in very large numbers. Large numbers of Marines get roasted by blue flame Hellions, and Vikings can be countered by more Vikings, and hence, air superiority wins. And all of this can be achieved on a few Factories and Starports by moving add-ons back and forth.

In modern TvZ, Zerg has Queens, Hydralisks, Mutalisks, Corruptors, and Viper’s Parasitic Bomb. To be totally honest, I have no idea why this build should work. I guess standard ZvT is Muta-Ling-Bane. The only meaningful change as part of LotV that I can think of is the Tactical Jump, which would seem to help a lot.

Sorry for not providing more insight on this build, but I don’t have a VOD link, and I can’t watch the replays. Even so, hopefully there are more of you out there who are just as excited as I was to see this build come back. Let me know how this works for you in the beta!

Download the replay

LotV Zerg v Terran by Stephano (featuring Vipers and Hydras)

Posted on by
Reply

This morning, I tuned into Stephano streaming Legacy of the Void. Simply, I was blown away. I have watched various streamers over the past 2 weeks, and most of them are still working out through new units and strategies. Even if he hasn’t solidified his builds, Stephano played with the decisiveness in engagements and ease of micro that made him an instant success in Wings of Liberty, and it was a tremendous pleasure to watch. I watched him play 3 ZvT and 1 ZvZ games, so let’s dive into some analysis about what he did.

Stephano (Hydra/Viper) v. Semper (Mech with Banshees) – Link

Continue reading →

Apollo shows us HotS builds in LotV (and how to adjust build order timings)

Posted on by
Reply

Most streamers are trying new, fancy things in LotV. Openings are adapting for the economy changes. New units are everywhere. Not Apollo. Apollo knows HotS builds, and he’s using HotS strategies. And they seem to work, more or less. It might not be the most innovative StarCraft play, but this is really helpful for us because it helps us compare timings from old HotS builds versus LotV builds.

We have some writeups from his past tutorials, though the most useful one is probably his Terran in HotS. Here’s an example build he used in TvT. It is a very standard Reaper Expand into Starport

Apollo’s HotS TvT

  • 1:00 10 Supply Depot
  • 1:37 12 Barracks (at ramp)
  • 1:50 12 Refinery (delay next SCV slightly)
  • 2:45 15 Orbital Command, Supply Depot (at ramp), Reaper (scout)
  • 3:40 18 Reactor (Marine x2 when ready)
  • 4:00 19 Command Center (in-base)
  • 4:12 20 Factory
  • 4:36 21 Marine x2
  • 24 Marine x2
  • 5:15 27 Starport, Hellion x2, Swap, Orbital Command

 

Compare this to a build he just used on stream against ForGG.

Apollo’s LotV TvT

  • 14 0:20 Supply Depot
  • 16 0:42 Barracks
  • 16 0:52 Refinery
  • 19 Reaper, Orbital Command (continuous Marines afterwards)
  • 20 Command Center (at natural)
  • 21 Supply Depot
  • 23 2:10 Factory
  • 23 Refinery
  • 30 2:56 Starport (delayed because of pathing, Medivac when it’s done), Tech Lab on Factory
  • 35 3:00 Bunker
  • 35 Cyclone (Tank next)
  • 4:26 Command Center, Barracks x2
  • 4:45 Tank drop starts
  • 5:30 Barracks x3

There are a few differences we have to account for between HotS and LotV builds:

1. Supply counts are about +4 from before. This begins with the Supply Depot going down at 14 instead of 10 and continues forward. This, of course, affects the minerals available, but most builds are timed out based on the tech tree, not minerals, so it works out

2. Times start about 40 seconds earlier. Because of the increase from 6 to 12 starting workers, builds start earlier. Use this as a delta on your builds.

3. The game clock is faster by a factor of ~1.4. Legacy of the Void has blessed us with a game clock that is the same as real-time, so your 15 minute game actually takes 15 minutes, not 11 minutes.

These changes lead to the build adjustment formula for HotS Terran builds to LotV Protoss builds, which I will coin as “Apollo’s First Law of Void”

LotV = (HotS – 40) / 1.4

By the way, Apollo crushed ForGG in that game. My rough analysis is that Cyclones are good in small numbers, but they aren’t a backbone to a Mech army, because mass Marines work pretty well.

Tune into Apollo’s stream if you get a chance. He’s a great player to learn from.

See Legacy of the Void build orders from replays

Posted on by
1

Legacy of the Void (LotV) replays can now be parsed and viewed at http://lotv.spawningtool.com/! Big thanks are in order to Blizzard and Graylin for getting this issued so quickly. Just like LotV itself, the build is unstable and requires a lot of testing, but in principle, we can see what’s going on in the game.

In terms of our knowledge of the game, the biggest obstacle at the moment is that we actually haven’t really converged on any terminology or builds, so it’s a little tricky to know how to label the replays properly. That, however, is part of the excitement of the beta, and I’m looking forward to refining our taxonomy over time.

The biggest obstacle, however, is just sample size! If you have any LotV replays at all, please upload them or email them to us at spawningtool@gmail.com. We want to get as much into the system as possible. So far, the biggest dump just came out of the Lycan LotV Beta Tournament held earlier today. Big props to them for organizing something out of the gate and showing how exciting the game can be.

One more thing: if you see any interesting builds or strategies that you want us to do a writeup of, please let us know where you saw it, or better yet, send us a replay! We are looking to do some analyses of build orders and to really dig into this game.

Keep watching!

The First Night of LotV Beta

Posted on by
Reply

Like many of you, I was anxiously watching Twitter all day to see when the Legacy of the Void (LotV) closed beta would start. Thankfully, I am on the west coast, so it did start at a reasonable time for me, and I dove into watching streams.

Just watching, I haven’t yet developed a great sense of the game, but frankly, that’s really exciting: I admit that I lost interest in Heart of the Swarm as the game settled into a relatively stable metagame. I know I’m diminishing a lot of the refine that happened over time, but I could come back months later and still feel like I knew what the builds and timings would be. LotV is totally new, and that’s really refreshing to learn and see new things.

I watched Pseudorandom, a masters level Zerg player. He played the LotV fan alpha mod, so he has a decent sense of the interactions and timings coming in, and it showed as he played quite a few ZvZs. Many players are trying out early Ravager pressure, and his response to that is to go Mutas: the timing seems tight, but the Corrosive Bile can be dodged, and none of those units can shoot up. I recommend you watch his stream to see how he plays it out.

Hopefully, http://lotv.spawningtool.com will be operational soon so that we can actually see some builds soon. Unfortunately, Blizzard has changed the replay format and haven’t yet released any details on what the changes are so that we can properly parse it. Even so, please upload your replays or let us know if any streamers are releasing replays so we can get them into the system. Once the parser is working, we can go back to analyze those replays, and I hope we get a good set in there when that’s ready.

Enjoy the game, and let me know if there are any particular games you would like a more in-depth analysis of!