TvP: Proxy Builds

Proxy is ubiquitous in Terran versus Protoss now, and its implications to the balance of the match up are debated by the community recently. This article is a follow up to my previous post on this topic.

Background

What is proxy?

Let’s be clear what proxy means in the context of Starcraft. According to Liquipedia, “A proxy building is a building built outside your own base including your natural base, generally in the proximity of your opponent’s base. … Proxy structures do not necessarily need to be in closer proximity to the opponent, any structure built outside one’s base is a proxy structure.”

Why proxy?

There are two main advantages to proxy.

One is to reduce unit travel distance to opponent’s base. The closer the proxy production buildings are to opponent’s base, the shorter time the units have to travel to commence an attack. The advantage of reduced travel time is multiplied when multiple units are produced from the proxy building. This is best illustrated with the classic proxy Barracks Marine Bunker rush. The vod below is arguably the most celebrated proxy Barracks game in Starcraft II.

The other advantage is to hide information from the opponent. Starcraft is a game of imperfect information, and it is always good to minimise the amount of information opponent has, all else equal. Proxy is a great way to reduce the information opponent can gather about what you are planning, because the proxy buildings are less likely to be scouted.

A player’s base is the most informative location, because that is the starting point and one cannot “hide” resource gathering. You cannot reasonably take two gas geysers that are not in your main base to fake opponent that you do not mine gas. Hence, players always try to gather information by looking at what the opponents have in their bases. When the production buildings are not built in the main base, the opponent does not know what and where they are. As a result, the opponent cannot partial out any of the possible builds you can do with the resource you gathered. This advantage is at the core of the TvP proxy builds.

Lead up to current proxy metagame

Proxy builds were conventionally used sparingly and they create an element of surprise to catch the opponent by surprise. However, proxy builds in fact are now the norm. The lead up to this proxy metagame is the understanding that Terran are usually behind when the game progresses to the mid game. For a long time, Reaper expand is the “standard” opening in TvP, and this can branch out to many different variations. All variations eventually converge to a point when Terran army move out with Stim upgrade, which is the trigger of shifting from early game to mid game in the TvP match up. As discussed in my previous post, Protoss are capable to control the early game to corner for an early third Nexus, and this allows Protoss to match Terran’s power spike in the mid game move out. In sum, the “standard” TvP script is not in favor of Terran as of current understanding. The proxy builds mess up the script to attenuate this asymmetry.

Types of proxy builds

The proxy builds can be understood better systematically after they are broken down into the following categories:

  • Standard
  • All-in
  • Fake

Standard

The standard type is essentially a quick tech to 1-1-1 with proxy building(s) then expand. The game looks like a typical TvP set up once the expansion is active. In other words, the standard proxy builds simply reshuffle the build sequence. I will discuss about the standard proxy builds in details later. Below vods are some examples.

All-in

There are several all-in proxy variations. Just to name a few: proxy Marauder Cyclone, proxy 1-1-1 Marine Tank Liberator push, and proxy Reactor Cyclone. Recently, proxy Reactor Cyclone is the most common, but I hasten to add that it still can transition to a macro game. The first vod below is an all-in example, and the second vod is a macro transition example.

Fake

The idea to fake proxy is not new. In this context, fake proxy builds are standard proxy builds with the buildings built near your own base. This can force an overreaction from Protoss. Below vods are some examples.

Breaking down standard proxy

Let’s move to the main discussion, which focuses on the standard proxy builds. As mentioned earlier, the standard proxy builds are early tech 1-1-1 (or Reactor Cyclone) into expansions with proxy production buildings.

Build order

1-1-1 offers a great flexible set of structures for a wide range of variations, so there is no “standard” build order. A good way to understand the build(s) is to break it down into different building blocks. All the builds start with the basic opening block.

Basic opening

14 – Supply Depot
15 – Barracks (proxy)
16 – Refinery
17 – Refinery
@100% Barracks – Reaper* and Orbital Command
20 – Supply Depot
@100 gas – Factory*

The number of Reaper usually ranges from one to two. Reaper can be replaced with Marine then Reactor for the Reactor Cyclone variation.

The Factory can be built in base or proxy. If it is proxy, you can build it next to Barracks (usually for Reactor Cyclone) or in another location.

Example: Widow Mine drop

@100% Reaper – Reaper (@100% – Send Barracks back to base)
@100% Factory – Reactor and Starport
@400 mineral – Command Centre (@100% – Orbital Command); then Supply Depot
@100% Reactor – 2x Widow Mine
@100% Starport – Medivac; then Supply Depot
@100% 2x Widow Mine – 2x Widow Mine

Load the Widow Mines into the Medivac for a drop. You have many choices in your follow up, but that choice should be something that pins Protoss units in their base. This is because Terran have nothing at home. If you transition straightaway to a standard macro game (e.g., get Stim and build more Barracks), you may just die to a simple attack.

It is not meaningful for me to provide further build order notation from here on out for this variation as it really is free style. The first vod below is a Widow Mine drop into Cloak Banshee, and the second vod below is a mass Widow Mine drop.

Example: Hellion drop

@100% Reaper – Send Barracks back to base
@100% Factory – Reactor and Starport
@400 mineral – Command Centre (@100% – Orbital Command); then Supply Depot
@100% Reactor – 2x Hellion
@100% Starport – Medivac; then Supply Depot
@100% 2x Hellion – 2x Hellion

Load the Hellions into the Medivac for a drop. You can do the same thing by changing Widow Mine to Hellion. The example uses one Reaper and it is a fake proxy. Thus, the build order notation is different.

Example: Reactor Cyclone then Hellion drop

*Get a Marine rather than a Reaper at the start, and the Factor is built next to the Barracks.

@100% Marine – Reactor
@100% Reactor – Swap Factory onto the Reactor and produce 2x Cyclone; send Barracks home; Supply Depot
@100 gas – Starport (proxy)
@100% 2x Cyclone – 2x Hellion; Supply Depot
@100% Starport – Medivac
@100% 2x Hellion – 2x Hellion

Drop 4x Hellion in Protoss main, and attack with the 2x Cyclone at the natural. See vod below.

Proxy versus non-proxy

If the standard proxy builds are essentially early tech into expansion, then why don’t Terran pro players just go for the same early tech builds without proxy?

The challenge Protoss face is not the quick tech per se, but it is the layers of information gap on what Terran are doing. Let’s go through the build from the beginning.

The build starts with a proxy Barracks (usually near to opponent’s base). The Reaper arrives at Protoss’ base earlier than a Reaper travelling from Terran’s base, and it has the potential to kill some Probes. A typical reaction to a proxy Reaper opening is to get a Zealot, because this complicates Reaper’s micro and directly reduces the damage Protoss take. More recently, it appears that Protoss players skip the Zealot and rely on Probe micro. They take two gas and use Chronoboost to get an Adept instead. This usually follows up with investment on more Gateway units or tech. Whatever the option Protoss pick, the mining of Protoss is disrupted and the Nexus is delayed. Thus, the proxy itself forces a reaction. If the location of the proxy Barracks is not scouted, Protoss do not know whether Terran build more units, put down a add-on, or send it back home. This creates information gap in Protoss’ decision making.

Next, a proxy Factory adds another layer of information gap. Since Protoss’ Probe scout in Terran base can see two Refinery taken, they know it is likely a tech follow up (i.e., Factory). But what is that Factory building? Does it have an add-on? From Protoss’ perspective, it can be Reactor Cyclone, which requires an immediate response from Protoss. For example, as a reaction, Protoss build Shield Battery and get Immortal (see image below). But what if it is a Widow Mine drop instead of Reactor Cyclone? These immediate investments become “bad” moves. More importantly, even if Protoss do not see the anticipated Cyclones, they still do not know what the follow up can be until they scouted the buildings. A possible proxy Starport increases the uncertainty further.

In the image below, Neeb does not know whether TY is having another pair of Cyclone on the way. In Neeb’s mind, TY may be pulling Scvs for an all-in with four Cyclones. TY has previously shown that proxy Reactor Cyclone all-in is in his build arsenal (see the vods that I have already linked earlier). As you can see in the production tab in the image below, TY is following it up with a Hellion drop, and that requires Neeb to position some units in the main. Protoss do not have such a high degree of uncertainty if Terran were to build the buildings in base.

Fundamentally, it is about deduction. A player cannot conclude that the opponent does not do X, because s/he did not see X. The thinking process is whether the opponent can do X with the things they have already done. I won’t go into depth about scouting and deduction here, but you can read about it in this article. In sum, when Protoss know that there are proxy buildings out on the map, they have a hard time knowing what to prepare for as they cannot cross out the different possibilities.

In fact, the issue with information gap in the TvP match up dates way back, but the role simply has swapped now. Terran have been dealing with the uncertainty of Protoss proxy tech even before Legacy of the Void. For example, in Heart of the Swarm, Terran can scout Protoss’ base and know that Protoss have proxy something early on. From Terran’s perspective, Protoss could have proxy multiple Gateways, Stargate, Robotic Facility, or Twilight Council. Terran have to roll the dice and prepare blindly if they cannot get confirmation. In the vod below, HerO proxy Twilight Council for Blink, but INnoVation blindly reacts to the possibility that there is a Stargate and builds Engineering Bay and Missile Turrets. This issue is more serious before Legacy of the Void, because the opportunity cost of any decision in build order with six starting workers is drastically bigger than with twelve.

In Legacy of the Void, Terran still struggle to know what proxy Protoss are hiding. For a period of time, Protoss have two main proxy builds in their arsenal, and Terran have to roll their dice what the proxy is:

I even developed a framework on how to react to the possibility that Protoss can proxy either of these tech choices.

Executing the build

On paper, it appears that Terran simply have to pick a variation, and then it is on Protoss to roll the dice and pray. Contrary to this popular belief, the skill requirement to play proxy builds is extremely high. The Terran player needs to have incredible understanding of the game and is able to free style on the fly. “Just play like Maru” is a fitting description.

Here are some examples of how Terran react and change to a different variation.

Example 1

In the vod above, Maru attempted to go for a Marauder and Cyclone follow up after the initial Reaper (see image below).

However, once Zest’s Probe scouted the Factory, Maru cancelled the Cyclone and put down a Starport for Widow Mine drop. Zest can tell that it is a Cyclone, because any other Factory unit produced without an add-on is far inferior to Cyclone in that situation. Then, Maru intentionally showed his Marauder by attacking with it, while he cancelled his second Marauder and Concussive Shell upgrade. The Marauder is to sell that he was doing a Marauder and Cyclone attack. Typically, you do not show your first Marauder, and you only attack when the second Marauder is out. This is because Concussive Shell completes at around the same time as the second Marauder. If you show the Marauder before Concussive Shell completes, you basically tell Protoss to keep their units safe from Concussive Shell. With the information that Maru had a Marauder and was producing a Factory unit without an add-on, Zest prepare for an attack at the front door of the natural. His attention will not be on his main base (for further reading), and hence, Maru’s tech switch to Widow Mine drop is a beautiful reaction. Look at the screen shots below.

Example 2

In the vod below, TY was doing a Reactor Cyclone all-in with Scvs.

Stats found the proxy Factory and destroyed the Reactor just before the two Cyclones were out. This is the worst possible scenario for this build. TY had committed enough to the two Cyclones that he did not have other things. The attack had no chance to work with one Cyclone. Stats then could make more units to counter attack, because TY would not be able to pin Protoss in base. The game could easily end with a counter attack.

TY reacted by continuing to make a second Cyclone and built a Starport in base. The Starport was for a Medivac and subsequent Cloak Banshees. The Medivac was for the two Cyclone. This is probably the best option to force Protoss to have the units in his base. The subsequent Cloak Banshee does not only serve the same purpose, but it also has the potential to help TY come back from the early game deficit. This can then transition to the typical Cloak Banshee TvP build.

These two examples do not do justice in showcasing how brilliant are Maru and TY. Countless decisions are made on the fly to change things up based on what Protoss know and what the gambles they make. But I hope these examples are sufficient in helping you appreciate the skill these top Terran players have demonstrated with these proxy builds.

Addressing common misunderstandings

After reading the discussion others have on this topic on different platforms, it appears to me that many have misunderstood the rationale behind these proxy builds. I want to address some of the common ones here. I cite some sources merely as examples of the prevalent perspectives, and the citations are not meant to target specific people.

This is a single strong build.

Some generalise a specific variation as the proxy build (example). This line of thinking overlooks the fact that the essence of proxy builds is the range of variations with the basic 1-1-1 set up. There is no real “build order” beyond the basic notation at the start. Of course, Terran would plan a specific variation at the start as there are some differences in details for the different options, but the build order changes according to the in-game interactions as demonstrated in the above discussion.

The lack of a definitive build order makes it extremely tough for many players to adopt this play style. This also explains why casual players often opt for 2-1-1 over 1-1-1 even though 2-1-1 has many weaknesses. 2-1-1 provides a relatively easy framework for players to execute a build order.

Replace it with a non-proxy tech build.

After I had published the last post on this topic, quite a number of readers have asked me whether the double Refinery Reaper tech expand builds work (example). They are referring to the builds that are common in TvT:

When you compare these TvT build orders with the standard proxy ones, they are quite comparable. You take two Refinery early to get a Factory immediately after Barracks, and then you expand later when you have 400 mineral. Hence, it is reasonable to suggest the common TvT builds should be equally variable in TvP.

While this is logically sound, it shows a lack of perspective taking. As mentioned earlier, Protoss’ difficulties to deal with these proxy builds stem from the information gap created by proxy. If Terran were to build the production buildings in base (i.e., non-proxy), Protoss can tailor their defences specifically to the information they gathered. Thus, non-proxy version tech builds are not comparable to those used in the current proxy metagame.

This is low risk high benefit for Terran.

If executing proxy builds without reliance on a specific build order involves little risk, then it means the act of proxy itself that is overpowered. If proxy is overpowered, why players only proxy so much now? If proxy is overpowered, why Terran players do not do it as often in TvT and TvZ? Since these phenomenons that should occur when proxy itself is indeed overpowered do not occur, the argument that proxy involves little risk is unlikely to be correct.

This misunderstanding could perhaps be attributed to the association of proxy with all-in. All-in in general is an all or nothing attack, so the all-in player should lose the game when it fails. Indeed, as mentioned earlier, the proxy builds can be all-in, but it is only one of the many variations. However, the “standard” proxy builds discussed here are in fact closer to “standard” macro than all-in. The main goal of the discussed proxy builds is to mess up the typical TvP early to mid macro game script that many players are familiar with. Thus, when an attack appears to “fail”, it does not mean Terran’s plan has failed (example).

Proxy actually involves a high degree of risk. When scouted early by the opponent, Terran may not even complete the building and this can easily end the game. In the example mentioned earlier, TY’s Reactor Cyclone cannot be executed because the Reactor was found and destroyed by Stats. In the below vod, Zest found Maru’s proxy Barracks, and Maru had no choice but to cancel it. Cancelling your first Barracks is a huge deficit, and the fact that opponent knows you just started your Barracks again made it a lot worse.

Despite proxy’s advantage to create information gap in opponent’s decision making, it has other downsides even if it is not scouted. No race benefits more than Terran to have their production buildings close to each other. The ability to switch add-on is paramount to Terran’s production. One production cycle is wasted when a Factory builds its own Reactor instead of using the one prepared by a Barracks. The time required for the proxy building to travel back to base equates to three to four production cycles. Thus, assuming the proxy is not interrupted like in the vod above, Terran is trading unit production for ambiguity on the battle field. In other words, this is actually a huge gambit!

Then, why do Terran consistently gamble big? This leads to the next discussion.

Implications for the match up

There is an unspoken truth in decision making in strategy games. When you are winning, you reduce uncertainty on the playing field. When you are losing, you take risk.

Many of us apply this logic in Starcraft. The most celebrated example is “when behind, Dark Shrine”. This is because, making Dark Templars involves high risk and high return. In the last GSL final (see vod below), after Maru lost his Command Centre at the natural (i.e., losing), he pulled Scvs for a do or die push (i.e., big risk).

These examples only apply the logic based on the situations in individual games, and the decisions to take risk are not planned before the games. In contrast, top Terran players now plan before the game to go for high risk moves like proxy right from the start, and this suggests these Terran players believe they are at a disadvantage in the match up to begin with. Beliefs do not necessarily equal to facts. I will let you make your own judgement as to whether these top Terran players’ assessment of the match up is accurate.


A shout out to sundaysushi for being a Code-A supporter on Patreon. Thank you very much. It helps to sustain the site.

Advertisements

5 thoughts on “TvP: Proxy Builds

  1. Great post like any time Max! I am a bit concerned because of the upcoming revamp/testing changes and all their implications, so I will not going to train these so hard like other Meta BOs.

  2. “Contrary to this popular belief, the skill requirement to play proxy builds is extremely high. The Terran player needs to have incredible understanding of the game and is able to free style on the fly. “Just play like Maru” is a fitting description.” – So true!

    “when behind, Dark Shrine”. This is because, making High Templars involves high risk and high return.
    I guess you mean Dark Templars.

    Besides that, great article! Thank you once more!
    I think you may have touched on almost every aspect of the current TvP meta game. I am by far not as good as any of the korean terran players, but what I experience in my ladder games is that Protoss now try to play more proxy as well. Especially the 1 Gate at home, 1 Gate proxy is very common.
    I think Zest played it at least once in the last GSL finals. So the build was out there before and not an immediate reaction to the current meta game. But still, looking at the upcomming Blizzcon I fear a lot of very dicey TvPs and the supposed balance changes will not do anything to counteract against it. But then again, who knows what Maru, TY and so on will come up with.

    Keep up the good work :)

    1. Thanks. I edited the Dark Templar part. Win rate wise, TvP is at a decent spot. So if Blizzard probably won’t do anything, It is about the way the match up is played (aka proxy). There is little you do about proxy itself for balance.

What do you think?

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s