(Note: This thread is only for discussing non-expanding builds. FE is always more economical if it can be executed without safety or hatch-blocking concerns.)
A Second Look at Zerg Openings
I'm sure many of you remember my previous threads testing different Zerg opening builds and analyzing the 11overpool 18hatch in particular. Those threads I believe did a lot of good work, but at the time we were all pretty much groping in the dark. It was difficult for people to see the forest for the trees and actually apply the data meaningfully. Eventually I accepted the criticisms of a vocal minority and abandoned my efforts.
However, after recieving quite a bit of positive and curious feedback from people on battlenet, streams, and forums, I have decided to once again turn my attention to an analysis of Zerg openings. With each attempt, I have learned a little more and refined the process and the data to be more meaningful. I have been asked several times given this data why the pros choose the openings they do. I always assumed that the pros remember what worked in broodwar. Back then, delaying spawning pool production would always result in increased economy. However, with the introduction of queens and the spawn larvae mechanic, is this still the best decision? At what timing and ordering of production can we maximize both mineral and larvae production in a way that is viable strategically? This is the question I have been attempting to answer.
In order to prevent any confusion regarding gas timings, I decided to incorporate gas into every build order. I used the fairly standard ZvZ opening, 14gas/14pool in order to have a set time for gas. Therefore, every build gets an extractor at 1:40 and continues to mine gas throughout the test. I recognize that in most situations, getting gas this quickly is not necessary, but I did this primarily to decrease some of the bickering that was previously present regarding safety and viability, and to negate the argument that somehow gas timings would affect one build to a significant degree relative to another. After incorporating gas into each build, I came up with identical ordering of results as before, hopefully negating such arguments.
(Note: DET stands for double extractor trick.)
(Excel Spreadsheet Stats)
(Typical 1Base Larve Production Rate)
What does it all mean?
Note that the build which mined the most total resources, 14pool, also resulted in the fewest final drones. The build with the worst mining rate, 10pool, resulted in the highest total larvae count. There is a clear inverse relationship between minerals mined and larvae produced in the short run. (Please do not confuse this with minerals OWNED... We are NOT talking about spending or keeping resources when we say minerals mined.) This is common sense when you consider that any initial investment will sacrifice in the short run for gains in the long run. It is clear that as Zerg players, we are faced with a constant decision of trade-off between resource production and larvae production.
From watching the pro-Zergs play, it appears they tend to favor more early resources instead of more early larvae, whether knowingly or unknowingly. Are they correct in this assessment? Won't greater larvae production eventually result in the greater economy in the long run? These are very complicated questions with many variables and unknown factors, but if we ever hope to answer them, we have to start somewhere.
For some, this data will be a bit confusing, therefore I have offered a straightforward interpretation of the data and what it potentially means.
1) Which pool-first build gives the most larvae?
(Graph of Differences in Drones Produced with 14Pool as a baseline)
This is difficult to determine using specific moments in time, because the spawn larvae mechanic is constantly fluctuating larvae rates. In order to reach an estimate of which build yields the most total larvae, we must analyze a simple moving average (SMA) of the differences in drones produced for each build. That data can be seen here:
(Simple Moving Average of difference in drones produced with 14Pool as a baseline)
Therefore, we can see that the 11overpool build will yield the highest average larvae counts of any other pool-first build tested. Eventually larvae production will stabilize for every build, and additional larvae production will only be possible with the addition of hatcheries. But unlike resource count, the faster pool builds will have a permanent larvae advantage for the duration of the game without taking into account expansion timings.
2) Which pool-first build gives the most resources?
(Graph of Differences in Minerals Mined with 10Pool as a baseline)
Typically, the longer you delay your production of a spawning pool, the faster you will be able to gather minerals, but only in the short run! Note from the graphs that the mineral advantage of the 14 pool begins to decline at the 4:20 mark. If you look at the other graph, you will see this is the exact moment when the faster pool builds catch up and surpass in drone count. Therefore, once you pass the 4:20 mark, the mineral advantage begins to slowly decline and become overtaken by the faster pool builds due to their faster rate of drone production. Which build will yield greatest long-run economy is complicated by expansion timings and perhaps can be covered again later.
3) Which builds should we not use?
The following builds should be avoided: 10pool, 12pool-DET, and 12overpool-DET.
The 10pool and 12pool-DET should generally NOT be used because they do not produce sufficient additional larvae to justify their reduced resource count compared with the 11overpool. Notice that the 11overpool, 10pool, and 12pool-DET builds all end up with very similar larvae counts. However, the 11overpool build can give more than 100 extra minerals in comparison, and is therefore the preferred option. Likewise, the 12overpool-DET build should generally NOT be used. It gives you up to 35 minerals more than the 11overpool, but results in up to 2 fewer larvae, which is not a good trade-off.
4) Which builds should we use?
The build you should use is of course dependent upon the map, matchup, and your goal or play style. Here are some factors to consider:
In ZvZ, assuming you aren't quickly expanding, I think 11overpool is clearly best. An increased larvae count is a significant advantage in such an aggressive matchup. The extra minerals of the later pool builds generally aren't needed in this matchup if you are playing pool first, since you will typically not be fast-expanding, and you will be gathering more than enough in order to constantly produce lings and banelings. Getting an earlier pool will also give you the flexibility of going all-in if you scout a player going hatch-first against you, and also give you greater safety in defending against 6/7pool cheese. Due to all these factors, I would recommend 11overpool with gas at 1:40 instead of the more common 14gas/14pool opening in ZvZ.
In ZvP I would lean towards the 14pool build, because it is fairly common to expand at some point in the 20's supply, and also to get a roach warren and/or spines, all of which require a higher mineral count advantage as opposed to larvae.
In ZvT I almost always favor opening hatch first, but if you are going for an aggressive baneling bust or roach rush, 11overpool should always be best.
Real-world application and comparison
In order to test my theory of the relative advantages of 11overpool vs 14pool in speedling/baneling builds, I set up an experiment. I played repeated tests according to the following general guidelines:
@1:40 - Geyser
@16 drones - pump 100% zerglings/overlords
@100 gas - Metabolic Boost
@50-100 gas - Baneling Nest
Continue pumping speedlings and spending 100% of remaining gas on banelings
In attempting to maximize the use of the 11overpool's larvae count while putting all available resources to their best use, I believe I have developed perhaps the most efficient baneling build possible for Zerg. I am fairly confident that no other build can produce a greater amount of banelings and speedlings in a similar timeframe, but I am open to any challengers.
In adapting this build to a fast baneling bust, it also surpassed the current Liquipedia entry for baneling rush, which is very similar but has some small obvious errors in the ordering and timing.
Best Results @ 6:00:
15 drones, 27.04 Speedlings, 7 banelings, 5 overlords
11Overpool - encRoach ZvZ build
15 drones, 33.42 Speedlings, 7 banelings, 5 overlords
Build Orders for encRoach Baneling Openings:
+ Show Spoiler +
ZvZ, Speedling heavy or most aggressive:
10 Extractor Trick
11 Spawning Pool
11 Geyser (immediately after drone)
@100 gas - Speed, remove 1 worker from gas
18+ Mass speedlings
@75 gas - Baneling nest
-Continue pumping speedlings and spend 100% of gas on banelings.
ZvT, baneling heavy, or fastest bust, 5:20 target:
Same as above, EXCEPT:
@100 gas, do not remove worker from gas
@50 gas, baneling nest
Cut overlord at 26
This test confirmed that in ZvZ or a ZvT baneling bust, there is a slight advantage to going 11overpool when compared to 14pool due to the larvae advantage. This of course is a subjective assessment depending upon a players style and average expansion timing. As was expected, there were more minerals left unspent by the 14pool build. Therefore, with a differing build, it would be possible to divert minerals into gas for more banelings. I will not attempt to compare the relative worth of a baneling to speedlings here since that is a highly subjective and situational assessment. However, it must be noted that this added flexibility of additional banelings must be contrasted with the 11overpool's flexibility of creating a higher number of either drones or speedlings if your opponent goes for play which negates banelings such as roach openings.
A look at the in-base hatch
The in-base macro hatch is an interesting concept. It is difficult to say how many situations exist where you would prefer a second hatchery in your main as opposed to at a natural expansion, but let's not suppose they do not exist. How efficient would such a build be?
I conducted two tests. The first test includes gas at the 1:40 mark to provide a fair comparison with the other builds. In the second test, I eliminated gas altogether in order to test maximum drone count. What I found was very interesting, but it must be noted they they underscore the production capacity of such builds. 6:00 was too short a time to compare the power of double hatch/double queen to single.
3160 Minerals Mined, 37.82 Drones
3600 Minerals Mined, 41.71 Drones
3154 Minerals Mined, 35.24 Drones
Both of the in-base hatch builds surpassed the pool-first builds in workers, and were either better or comparable in minerals. At first glance it would appear that even an in-base hatch is always better than pool first, but we must consider other factors before jumping to such a conclusion. The in-base hatch builds were only so efficient because they continually produced drones and thus were able to actually take advantage of nearly all the additional production capacity. In a real-game, this would typically not be possible due to the necessity of acquiring an army and upgrades, therefore leaving many larvae incapable of being spent at all. I decided to dig a little deeper to determine when, if ever, an in-base hatch was the best option.
Obviously the sole advantage to an in-base hatch as opposed to an expansion is the greater safety afforded by the building location/terrain. Clearly this safety is an asset, since saturating quickly is both what this build is capable of and requires in order to put its high production capacity to actual use. I envision such a build being most useful on maps such as Backwater Gulch, where your natural + main are practically impossible to defend against a 4gate. So here is the general theme such a build would take:
1) Take advantage of terrain to safely and quickly saturate your main. This can most easily be accomplished by utilizing the two fast queens to block your ramp and the proxy hatch creep near the ramp for spine crawlers.
2) Take advantage of this quickly saturated main and very high production capacity to pump tons of speedlings for map control or aggression.
3) Take advantage of this map control in order to expand and once again quickly saturate you natural. Repeat.
Here is a test of maximum production capacity of zerglings beyond the six minute mark to illustrate the points at which each build has a comparable production advantage. It could be suggested that the in-base hatch is to vulnerable to a more standard build before the 6:00 mark, but let's not forget that we can very quickly get two queens out to block the ramp, and also remember the defender's advantage as far as distance travel is concerned.
I took this general strategy to the master ladder for a few games for fun and testing and had surprisingly good success with it. Of course there are always weaknesses to any build, and much of the execution has to be ironed out, but it is clearly something for us to consider seriously.
EDIT: ZvZ replay removed at player request.
I'm not sure how much we can draw from the replays. Obviously the wins don't say too much, considering my opponents played imperfectly, but I find the production capacity at least to be pretty amazing, and the possibilities in ZvZ given the fast double-queen ramp block to be potentially very effective in trying to increase saturation in a matchup that has become so hyper-aggressive. I will continue experimenting with the in-base hatch in the few situations where it could perhaps be warranted.
Note, these are very simplistic generalizations of the conclusions I reached for people who don't want to read through all the data:
Never 10 Pool, and never double-extractor trick.
Start going 11Overpool for speedling/baneling play in ZvZ or cheese ZvT.
Keep going 14 Pool in ZvP.
Consider an in-base 14 hatch on maps with a difficult to defend natural in ZvZ or ZvP, using queens to block the ramp and spines to defend while you quickly saturate and use high production capacity for mass speedling map control.