Last christmas I had a correspondence with Letabot about axiom build orders for the 3 terran matchups. I suggested 1 build order for each matchup to act as the starting point for 'the perfect build order'.
For the opening I chose rax expo ->1 rax ebay acad
For the overall style, we are doing a 3 rax midgame, and then going into a mech switch. And we are doing a 3 tank opening rather than a speed vult style.
The idea here was to do a quick first pass and then slowly add more nuance and detail. So I wrote the guide and then asked Letabot for questions. He is ok with me posting this publicly.
1 rax expo, ebay, academy, 3 rax, fac, star, science, 3rd cc inbase, 3 fac, armoury, 4th cc (at new main), 5 fac, 2nd armoury, 7 fac, 5th cc (at new nat), 9 fac (these 2 in new main), 11 fac (these 2 in main), 12 fac (this one in new main)
This is going up against Zerg doing: 3 hat spire 4th hat (at 3rd base) Den Queens Hive Evo Defiler mound Nydus canal 5th hat (at main) 2nd evo 6th hat (4th base) 7th hat (at 3rd base) 8th hat (at 3rd base)
The main deviation comes from scouting a 2 hat opener. Instead of going 1 rax ebay academy, we instead go 2 rax then academy then ebay. This gives earlier scanners and earlier mass. If its 2 hat lurker we can make factory after only these 2 rax. If 2 hat muta, we can go 3-5 rax.
Some little tips: -Get 9 marines then 4 medics. -Start making the 2nd and 3rd barracks, then start the factory. When the barracks finish is when you make turrets. This is vs 3 hat muta. Make 4 turrets at the nat, 2 turrets at the barracks, and 3 turrets in the main near the main cc. Make 2nd gas after the turrets. -Get +1 infantry weapons then +1 infantry armour -Roughly 6 medics, 24 marines, 3 tanks, 1 vessel
-Now we are on 3 facs, make vults with mines then speed from all 3 facs. Make 12 vults. Place mines at the zerg intended 5th. -Now we are up to 5 facs. Make a second machine shop.
-Going up to 7 factories now and 3rd and 4th machine shop. This is when we begin tank production in full earnest. Get Charon booster goliath range, and a handful of goliaths.
When on 5 fac, have 2 shops. When on 7 fac, have 4 shops. When 2 new facs at new main, they are both for tanks only. When on 9 facs in the main, make a 5th machine shop. When 3rd fac at new main, it is also tank only.
-Stay on 5 base for a tick. The strategy attempts to contain zerg on 4 base, and deny any attempt to take a 5th/6th. If the Zerg is able to, Terran is still fine, and is still ‘winning’. Make a couple of turrets at the neutral bases, and take a 6th base.
-Feel free to mix in 4 valks into the lategame army. Position them to defend vs a drop on the new main/nat. Bring them out if a muta or guardian switch is revealed.
-Aiming for around 68 scvs. No more than 70, no less than 60.
Above Ive concentrated on very rigid targets, and focused on buildings and production rounds. I havent talked about army movements or anything else.
Lets do a quick update on current styles and a bit of history too. At the moment Flash is quite often happy doing rax expo->2 rax academy openings even vs 3 hat zerg openings rather than rax expo->ebay academy. This is absolutely fine. It gets 2 medics then 2 firebats. It forces zerg to make some sunkens before muta. It makes T safer vs massling attacks, as T can make firebats earlier. And you can still defend the mutas perfectly adequately and put a lot of pressure on zerg as he tries to take his 3rd base too. I think the ebay style is slightly stronger as you can go head to head with mutaling just a tiny bit better, and hit a little bit harder when zerg is setting up his 3rd base. Zergs have gotten really good at holding this pressure now though, getting just the right amount of lings and picking off small squads of mnm before they link up, and by using lurker egg blocks (start morphing 1 at the top of the ramp slightly before morphing the other 2 directly on the ramp) (or 2 above and 1 on the ramp). So maybe Flash is thinking its better to just force the early sunkens and continue from there. Next, Terrans were previously getting 2 turrets at the barracks and 2 turrets in the main forming a perimeter to keep muta ‘out’. Nowadays its more common that Terran makes the turrets much closer to the main cc so that if the mutas commit to attacking the scv then the turrets will be shooting and for Terran to get a 3rd turret at that location as well. Zergs were finding they could park muta in the corner of the terran main and get in some decent damage, with mnm finding it difficult to navigate the clutter of scvs mining and depots in the way. So yeah that’s a little thing that has changed fairly ‘recently’. (to be honest things like this go in and out of ‘fashion’ to an extent as well. Eg once zergs think that terran is always getting 3 turrets he will stop attacking there so much and then later in the timeline terrans will start cutting down on turrets again etc).
Next, the speed vult first style! This is a fun style where instead of getting 3 siege tanks with siege mode, you get 3 vults with speed first, and potentially run straight past the lurkers holding the ramp into the zerg 3rd and score drone kills or draw the lurker shots away from the ramp so the mnm can advance up and kill the lurkers. The move is sometimes attempted with just 2 vults, sometimes with 4. Terrans will often continue making vults after this move, around 5-8 then a tank or two then back to vults. Now firstly I want to say this is not a new build (I mean specifically skipping tanks and getting vults first). Midas was using it in 2010. Yes 2010. But it didn’t become a standard option until the last couple of years. You keep getting speed vults and get mines and continue into the standard mech switch, which has been standard since 2011. The idea is that Zerg have gotten really good at getting to darkswarm to hold the 3 tank attack. So at the worst, T simply skips to the next stage, and at the best he manages to sneak the vults into the 3rd base and get an advantage by killing drones or distracting the lurkers so mnm can engage. This style works particularly well on ramps that have inverted main/nat setups (low ground mains and high nats). Indeed this was where Midas used the style, on aztec. Sea was using this style on Jade quite a bit etc. Zergs are now using a simple trick that helps prevent the vults getting in though. They keep 1 lurker unburrowed sitting blocking the ramp. Another option is just to keep some lings at the top of the ramp to block the runby. Its certainly an added worry for the zerg to look out for atleast. The reason siege tanks are good at this timing traditionally (in the most stating the obvious way) is because they can engage lurkers safely and can attack sunkens safely. If zerg goes for the +1 armour 12 lurkers (mass lair) style to kill the 3 tank attack then siege tanks are more vital. If zerg goes for the standard faster hive darkswarm style to hold off the 3 tank attack then vulture style is more potent. So if Terran can reliably scout which method the zerg is aiming for then he could reactively choose whether to do vulture style or tanks style. However I don’t think its possible to accurately reliably scout. You can scan the timing of the zerg buildings but he could potentially switch after your decision point and this would be early enough for him to do damage that could be fatal. Some Terrans might believe that through mnm control they can hold off lair heavy style without needing tanks. And infact by not having any tanks they can simply retreat from the position easily rather than ‘having to commit’ to defending the tanks somewhere in the centre of the map. Meanwhile the vults can just run around and cause havoc as the zerg is concentrating on using his whole army as one. And these Terran might say that if the zerg lurkers manage to push them all they way back into the nat, then this is the point he can start making tanks reactively. Well yes there is a point to be made there. I worry that without tanks you wont have enough muscle to really deal with lurkers when it properly comes to it though. And I worry that if you don’t get tanks then you never really threaten the zerg sunkens, so the drive for him to tech up to hive and darkswarm potentially is reduced. If zerg somehow knew you wouldn’t get tanks, he might even be able to make sunkens to defend his 3rd base for the extreme example. Obviously that’s not real but what I mean is, the mnm escorts the tanks and the tanks break down the sunkens, which forces zerg to tech to darkswarm. Its all well and good skipping a stage to get ahead but it isnt sustainable if looking for a standard axiom build order. So those are the reasons Ive recommended 3 tank style rather than the currently popular speed vult style. Just to be clear though, I think the speed vult style is viable, and is very powerful, and should be used sometimes. Lastly, Battlecruisers are a minor thing now by the way, if you didn’t already know. This is thanks to Piano. He is one of the few current day terrans who doesn’t use mech switch as his standard tvz style. He tried it for a while but has since gone back to using 7 rax 2 stars, and gets some vessels, then dropships, then 2 battlecruisers to tank scourge shots, which helps dropships to land their cargo. While we are touching the subject, Mong sometimes uses a build where he keeps making mnm and goes up to 5 rax and gets higher (sometimes all) infantry upgrades and gets dropships. Then does a later mech switch if the drops get shutdown. He also experimented with getting a squad of wraith at this late game stage, particularly if zerg is using queens. But the wraith often get plagued and are inefficient. Lastly on this point, Flash still quite happily goes 7 rax and 2 stars if he is going into the lategame with a big advantage, particularly if he has killed the zerg 3rd base or delayed it for a long time with just mnm. He was historically slow to adopt the mech switch too (see below).
Lets talk about the mech switch in general briefly too. It was showcased in 2010 and quickly its power was evident. There were problems in TvZ for terran. Zergs had become adept at holding 4 bases vs the incredible pressure put on by mnm, vessels and drops. They were holding the nats with lurkers under darkswarm, they were holding off the drops with scourge ling, darkswarm lurker, and they were holding off the vessels by making more defiler in time, and eventually plague the vessels, spore colony, a muta, or just more scourge. Once past this stage, Terran was finding the ultralisk followup with lings and darkswarm and plague was proving to be just a bit too strong. MnM with vessels were not getting the job done. It was back to the drawing board. Flash starting massing tanks in the lategame. But others took it further. What can stop lurkers under darkswarm? What hurts ultralisks? What isnt weak to plague? The answers are spider mines and tanks. The goal was clear, to get these units out quickly after the 3 tank 1 vessel timing, but the trick was working out how to make the transition. How far can Terran push it, how much econ can he safely claim, how many corners can be cut. It took an amount of time to work out the kinks but by 2011 the mech switch was very common and in full swing. Many of the best terrans were using it regularly and most of the lower terrans too. And every terran had used it in some televised games. Sea was one of the main terrans to adopt it quickly and had great success with it. Id say he kind of lead the way with it. It continued to be used through 2012, it was absolutely the standard style. Funnily enough Flash was the exception here. He was the terran who pioneered getting mass tank lategame but this was always with constant mnm up to that point. He was the driving force behind different number of barracks eg 5 rax mnm midgames. But he was so successful at TvZ that he literally had no need to adopt a different style. He was very slow to adopt the mech switch style compared to every other terran. And that’s pretty unusual for Flash. In all 3 terran matchups Flash would happily innovate and lead and adopt new styles immediately. He was crushing everyone in all 3 matchups and didn’t technically need to adopt latest styles in order to do so, he could have won by just playing any old style he wanted really. So that’s what makes it an interesting little thing. Don’t get me wrong, Flash totally innovated and lead the way with many many TvZ things all on his own. And other times he would see someone do something once and instantly take it and improve it, as all good players should. And that’s not taking away anything from Flash. Usually naturally in all games the onus is on the weaker players to find new innovations. But Flash would innovate anyway, partly because he is very good player, partly because he seemed to have an insatiable thirst to explore the game, and partly (sadly) because he thought the fans were putting pressure on him to use more aggressive strats. But with the mech switch he chose to be late to the party and why not, he was winning without it. But Flash was using mech switch regularly by 2012 (likely earlier) like everyone else. By the way, Upmagic and Midas don’t get enough love for their innovations.
On a different note; AI is not bound by usual human constraints of keyboard controls. These are things like: -12 units in a hotkey -10 hotkeys -f1, f2, f3 locations -human hand size, fingers reaching keys
The above things dictate strategy to some extent. AI are not bound by these things which opens up additional options. The optimal AI strategy is likely vastly different from a humans strategy. Eg, an AI could open with 1 vulture vs Zerg and reliably depend upon the knowledge that it will kill infinite zerglings. A human can (and sometimes does) open in this way but its much riskier.
Similarly, a human often uses 11 muta 1 overlord, or 11 goons 1 observer, or 12 marines, or 24 tanks 24 vultures. Whereas an AI can use 27 ‘squads’ of 1 zealot each effectively. Or 13 mutas perhaps. If the unit design allowed it, an AI could probably beat any human with just one unit potentially. For example, a reaver and shuttle PvT AI might be unstoppable, or a vulture to wraith TvZ AI etc. Or just worker micro and bunker rush/4 pool etc.
The details on certain things are important in many areas, for example: building placement unit positioning recognising and exploiting opportunities
What deviations do you have so far? The bot recognises the deviation and adjusts its build order etc? This area is where you will be able to really make the bot dangerous. Its also very important in tvz because we have 2 hatch or 3 hatch, and we have lurker or muta. There is 2 hat lurker drop, there is mass ling. These are the deviations I would identify and adjust to first. Ofcourse we can just keep going more and more in depth as desired.
Letabot asks me three questions 1. When you push out with the science vessel + tank + M&M army, when is it a good idea to attack the 3rd gas, and when should you attack the main + nat?
2. What can you do against a zerg that sneaks a defiler + lurker/ling around your army and casts a dark swarm over your natural, breaking it and threatening to destroy the main with this trick as well.
3. Why do most pro-gamer not build a single or perhaps several wraiths to harass overlords? The lurker/ling army doesn't have any anti-air, so the zerg would be forced to spend minerals on spore colonies or go for hydralisks, thus reducing their lurker count. Well worth the cost of a single wraith.
1. The easiest way to decide is to simply base it entirely on spawns. If it is cross spawn, then you can attack the zerg 3rd instead. If its not cross spawn, then attack the zerg nat. So basically pressure whichever base is closest. But choosing to push the nat with the 3 tank 1 vessel army is never a bad choice. As soon as darkswarm is cast and lurkers are under it, swing round to pressure the 3rd. He has to use nydus to save his 3rd. Once he has proven he can defend both locations with defiler then we go into containment mode. Position units outside both of his nats. This is when vultures are coming out with mines which strengthen the contain. Zerg will soon be able to take his 4th base but don’t let him out of either natural. Most of the terran army is focused on containing the natural, while the vults with mines are more focused on containing the new bases. Note that the first mnm army (pre tank) is going to move towards whichever base is closest to begin with. If the closest base is the zerg nat, then move towards it until you reach the halfway line, then scan and if there are 3 sunkens then don’t go there, instead pressure the 3rd base.
2. There will be no sneaking defilers around our army. The terran 3 tank 1 vessel army will hit when zerg only has 1 defiler. So the defiler will be trapped there at the zerg nat. Our strategy revolves around threatening the zerg nat and zerg 4th base with mnm tank vessel. In the meantime, we are getting vultures with mines. So these mines control one side of the map to begin with. If zerg attempts to run past the terran army then the defiler lurker ling has to run through the mine field and will die. Terran then just replants more mines. What if things are going badly? Imagine a scenario where zerg has managed to cast a darkswarm in the terran nat, and has burrowed 4 lurkers under it, forcing the nat cc to lift. What terran should do is: Save the scv, send them to the main. Rally buildings into the main so they dont go down in to the nat and suicide. Plant mines out of range of the lurkers so they cant advance any further. If we have any vults in the middle of the map then use them to plant mines to prevent any zerg reinforcements from getting in to the terran nat. Put any tanks into siege mode in the main so they overlook the nat and will kill the lurkers once dark swarm wears off. Irradiate the defiler. Look near the mid of the map for a second defiler that may be coming to join and irradiate him too. So now we are safe. We can have a poke with our main army and pressure the zerg nat, force him to cast a defensive dark swarm there. If he doesnt, we might be able to actually commit the attack and break in, and win the game right there. But assuming that doesnt happen, all we do now is simply wait for the dark swarm to run out. Once it does, we use only the fresh units that are in the terran main which will consist of vults, tanks, and mnm to clear out the nat and retake it. Continue from there. If there aren’t enough units to clear out the nat, then retreat the main army to help clear it out. Its not ideal but we are still in the game and just continue with the gameplan from there. If zerg does manage to do this attack, then there are 2 scenarios. Either we are behind or he is overcommiting and will be weak to a counter attack. If we are behind then fine, we are still in the game and can win, keep fighting. On the other hand, if he is weak to a counter attack then this actually favours terran. We can bust the zerg nat and wipe it out and then the zerg main too and after the dust settles we have basically secured victory.
3. You partly want to use wraith to take advantage of ai micro and I dont blame you at all. Its a good idea. A science vessel counters the muta flock with irradiate. It also detects at a point in the game when Terran has only 2 comsats. Lastly it has irradiate, one of the only things that can kill lurkers under swarm and is good at sniping defiler at range. These things outweigh what a wraith can achieve and what a wraith threatens. Zerg will be getting scourge to kill vessels and dropships. If Terran made a wraith instead, he will be happy that there is no dropship to worry about, and he will cast his darkswarm with impunity while feeling very safe. And then he will kill the wraith anyway with the scourge that he has already. Wraiths don’t really pose a threat on their own especially. However, there are a handful of example games where a small squad of wraith are used at this timing. Most often it is when terran opened with a wraith style in the first place. Very very uncommonly it is when terran is using the wraiths instead of tanks. I have only seen this style on semi island maps though. The idea is that you have enough wraith in a squad to micro against scourge somewhat safely, and they are able to snipe overlords in 1 or 2 hits, and are able to snipe sunken colonies (fulfilling part of the role of siege tanks). The threat earlier on is lurkers, and siege tanks are better vs lurkers than wraiths are. Then next problem is the sunken colonies at the zerg nat. Again, siege tanks are better than wraiths at solving this. The next thing is dark swarm. Siege tanks again are better at containing the zerg threats than wraiths are once dark swarm is out. This is because hydras under darkswarm are invulnerable to wraith and other things, like hydras don’t need to burrow to attack so they are fast and mobile. If zerg is facing wraiths, then he will start making hydras and Terran cant do anything to stop this. Soon, the hydras will have speed and overlords will have speed too. Now he can move out and clear the minefields and gain map control. We need tanks to stop this.
However, this is where its gets interesting. Everything I have been saying about wraiths is concerning human players. As we’ve said before, the perfect strategy might be completely different when using AI multitasking and reliability. Wraith are certainly a unit that would play into this. The option of getting 1 wraith before vessel is something I don’t really think is worth it because the dps is so low. You might kill an overlord yeah, but personally Id much rather have an irradiate sooner (or a dropship). The option of getting wraith instead of tanks is something that depends greatly on how perfectly they can dodge scourge. If wraith can dodge scourge in low numbers by being microed individually and using marine support to retreat to if needed then this could work (sort of). Personally Id stick with just getting siege tanks to work first though. Im sad to say I am an advocate of boring standard safe play to start with. But I think you are probably really interested in trying wraiths out. Obviously its up to you! I believe you already have the wraith micro coded and ready to go. I must warn though that I think lurkers would prove to be too strong ultimately. As I said before, we need vessel to hard counter muta, and then we need vessel to detect while on only 2 scanners, and then we need irradiate for vs lurker under dark swarm and for sniping defilers.
Another thing is the developing threat concept. When zerg sees a vessel, this means we present him with the possibility that we might start massing vessels. This is a huge threat for him to deal with, and irradiate is very good vs ultras and is a threat that only grows larger as the game goes on. However, wraith are the opposite to that. Once darkswarm is out, the threat drops off quite a lot, and when plague is out, the risk of using wraiths is very high for terran. As hydra numbers grow, the threat of wraiths dwindles further, and don’t forget spore colonies are strong vs them too. Its relatively simple for zerg to survive vs wraiths while the number of wraith builds up, and once there is a full squad of them its not really a problem for zerg to build a spore at each base and use hydra to zone the wraith out from the surrounding areas (in human vs human games anyway). So the remaining options that include wraiths would be some sort of vult and wraith opening that leads to a more standard lategame. Infact, I think this could ultimately prove to be the best AI strategy of all, as I think I mentioned in a previous message. Or a 1 base 2 port wraith which I think you have done already? So if you want we could switch to a different style altogether but I think its more solid to stick with something we know definitely works well at the highest human level and has a lot of evidence and example games to draw from. Don’t forget marines have a ridiculously high micro potential too. And you could do a dropship style lategame that would fully take advantage of AI multitasking.
Zerg might have some mutas left over (7 is actually quite common in a conservative zerg style and they get parked in the corner of a zerg main base so that they defend a potential terran dropship). And zerg has scourge so he does have anti air that can deal with a wraith. He also is making lurkers so he can choose to not morph a few of his hydra if absolutely needed. And if things went really badly for him and he had to make a spore at his nat, its not a problem, they are cheap, and they have good late game utility because they synergise very well with plague. In the meantime, by getting a wraith, terran has delayed his vessels (or dropships) which are much more potent.
Example Games. A really good example game is Killer vs Ssak La Mancha. http://www.teamliquid.net/tlpd/korean/games/62779_Killer_vs_sSak/vod I recommend watching this one as I think you will like it and its more of a solid kind of style and it is also the type of build that I think an AI could excel at performing.
Im also going to link you to this game, Flash vs Jaedong Circuit Breaker. This is not a wraith game, its standard play using the same build Ive been recommending. This game shows Terran getting annihilated. It shows exactly what we don’t want to happen starting with the mutaling killing the initial 12 marines and 4 medics. Then mutaling kills another 12 marines and 3 medics. This gives lurkers free reign to walk around the map. Then Jaedong performs a perfect lurker surround attack on the 3 tank 1 vessel army. And all of this allows Jaedong to cast dark swarm into terrans nat. I don’t know but it might interest you to see exactly how it can all go belly up. But don’t get me wrong, Flash is still playing an excellent game. To make sure this doesn’t happen to your bot, don’t move out so far with the initial 12 marines. Keep them closer to the terran nat. This way they can retreat and link up with the 2nd squad of 12 marines. Once we have this 24 marine army and 7 medics we can roll out into the middle of the map and pressure the zerg 3rd after scanning and seeing 3 sunkens at the nat.
Back to wraith examples: An example game of the semi island map wraith style is Fantasy vs Hyun on Fortress SE. http://www.teamliquid.net/tlpd/korean/games/59893_FanTaSy_vs_HyuN/vod This island map also has a close by air 3rd base, which makes guardians good. When guardians are in play, Terran gets wraiths. Note that Fantasy doesn’t reveal his wraiths until he has quite a few. And note that Hyun doesn’t get lurkers but goes straight for guardians.
Another thing to talk about is wraith in the ‘really late game’. Imagine terran has successfully executed a mech switch and has managed to secure 5 base. Zerg meanwhile transitions to hydra queen. In this scenario, sometimes Terran gets a squad of cloaked wraiths (and has +1 air attack by now). The cloaked wraith are used to hunt queens and to discourage drops. Again though, plague makes them quite weak. (Its not really what you are talking about though, but I thought Id mention it anyway). An example game of the above is Hero vs Mong on Circuit Breaker SSL 11, Ro8 Quarter Finals, August 2nd 2015.
It’s a long game, you can skip to 24 minute mark in the vod to see the first queens. The first wraiths are seen at 25.50. Then at 29.00 they all get plagued. I know that AI controlled wraith aren’t stacked because they are being individually microed. But hydras are the backbone of the zerg army at this point and plague is still probably quite dangerous. Scourge are also still a threat and can zone out wraith.
I support Terran getting 5 valks in the 'really late game' with air upgrades though. They are mostly just kept in reserve incase zerg does a big muta switch but can also discourage drops and queens. But the main Terran army is tank/gol/vult with lets say 6 science vessels supporting and plenty of turrets.
Another example of the above (really late game wraiths) is Effort vs Mong on Neo Aztec Draemong starleague Semi finals day 2 Bonus Match June 4th 2015. The wraiths are seen at the 28 minute mark in the vod. Between 28.00 and 30.30 the wraiths get some work done and snipe some queens but they do get plagued and all die in the process, still was pretty good overall. At 31.50 Mong has a new group of wraith. They snipe 2 defilers and 2 queens then get plagued. A 3rd squad of wraith is bought and they do ok but end up getting plagued at 38.50 which is right before Mong resigns. He has lost several bases to ground armies.
Lastly I just mention that wraith styles are better on certain maps and spawns than others. But we have been focusing entirely on standard 4 players maps so far.
Another option for late game TvZ is to get a 2nd ebay and more barracks (7), and a 2nd starport. Marine medic with 3-3 upgrades and vessels, but also dropships. If you dont win with drops and busts, then you can start mixing in a healthy number of tanks and you can still use vults with mines too. Here is an nice modern example game of SK terran. You can see how he still transitions/mixes in plenty of mech later.
The problem with sk terran is darkswarm, plague and ultras. Spider mines are the critical unit to pin zerg. In the game I linked, Mong still gets spider mines from 1 fac. Ill copy paste the basics from my notes on that game: T opens 2 rax academy then 4 rax. He gets 3 tank 1 vessel then 5th rax. He goes into SK terran with 7 rax 2 stars. Gets mines and some vults. Gets vessels but also does quite a lot of dropship play. Effort defends very well though and then later he shuts down 4 dropships in quick succession. Mong transitions into a more management style from here as Effort is stable. T gets 4 more facs but keeps making marine medic all game. ------------------- I didnt write down much but it doesnt matter really. You can see he had 4 dropships at the same time at one point. You dont need more than that at any 1 time. You can get 8 or more vessels but the problem is they become more prone to plague as the number of vessels increases. This is another reason why mech switch is better. I dont have another vod right now sorry. You play it the same as before though: -Park mnm outside his natural. Poke to make him have to cast darkswarm. -Take the 3rd base at the normal time -Have another bunch of mnm outside his 4th, and do the same move of poking to make him cast darkswarm. -This is the phase where zerg is pinned on 4 base and you take extra bases. -Do drops with 1 dropship, then do a drop with 2 dropships, then build up to doing a 4 dropship drop. -If he is shutting down the dropships hard then you have transition to a management style and get more vessels instead. -Still plan to move into mech 'late-lategame' to deal with the threat of ultras.
Letabot asks: What to do against a 1 hatch lurker opening ( 9 pool into hydralisk den ) when you opened 1 rax FE? And what about 1 hatch muta?
We wont be opening with a command centre when we scout zerg is on 1 base. The first scv scout goes to 1 base and if zerg isnt there, a 2nd scv scout is also sent so that the remaining 2 spawns are scouted at the same time. This gives us the information of whether zerg is going 1 base or not.
If 1 base zerg, we also stay on 1 base and get a 2nd barracks, gas, academy. Get range first, and scanner. And get ebay. Make a bunker at the top of the main ramp, and a turret near it but slightly to the side so it is out of range of lurkers but gives vision of them. This makes us safe from lurkers. Assume that the terran scv scout will not be able to scout if it is lurker or muta. The scanner can be used to scout the tech (muta or lurker). If its muta then get a couple of turrets by the main min line.
If its lurker, get factory next, then star and 3rd barracks, then science facility (all on 1 base). Get a dropship and drop marines in the zerg main. Get 3 tanks and siege mode, and science vessel. Only get irradiate if he switches to muta. So basically once you defended the lurker, prioritise tanks and dropship.
If its muta, get 3rd barracks next, then fac, star, science facility. Get marine medic and vessel with irradiate. So basically once you defended the muta, prioritise more marines and vessel.
In both cases you are still working towards a 3 tank 2 vessel marine medic kind of army all on 1 base. This army should win the game but build natural cc as you move out with this army. And from there flow back into our previous standard builds, this is a good time to flow into sk terran like in the Mong videos with 2 ebay, 2 star, 7 rax and continue from there if needed.
Note how I mentioned get range first, not stim. This is something I recommend but if you get stim first it is workable. Check out this vod, Zero vs Last Destination Feb 2015 41TV Live: "https://www.youtube.com/watch?v=KpI1L8YOTY8"
You can see how Last gets stim and is actually really aggressive, I dont really recommend that style, I prefer much more defensive and then counter attack with the dropship. He also makes a bunker at his nat rather than in his main on the ramp which I dont prefer.
Its not 1 hat lurker but its a 4 pool. The reason it might be interesting is that Fantasy goes 1 base 3 rax and tech afterwards, so you can see another perspective on what Terran is able to put out from 1 base. Fantasy should have been able to kill the natural hatch during the first mnm attack.
Lastly, the reaction to 2 hat (2 base) lurker is very different. If it is 2 hat (but still 1 base) lurker then do the above and go 1 base terran.
I'd love to see something this from the zerg perspective, but I'm surprised this didn't get any good responses. Obviously a lot of game watching and analysis going into these thoughts, love it!
Nice writeup, read it all. I missed mentioning Fantasy's name from the mech switch, he was really pioneering that.
However, I think this is really theorycrafting for high level. On C rank, i can't do half of this shit. Funnily enough, i guess this is just as unusable for 'not-there-yet-developing' bots as for untalented humans, because somehow bots just can't seem to do the other half. It is too difficult to code these strategic elements into actual action. Or maybe I'm not up to date with bots, but checking some recent Letabot uploads, i think we are not quite there yet.
Somehow i would imagine that against a human you need to 1. teach the bot how to hold off all the cheese 2. code it to play safe with little but constant harass because once it gets to 2-3 base full macro mode, the perfect macro and the individual microing of units will be unholdable for any human player. Even if they expanded faster. I might be wrong. But in order to achieve these, we need to scout the buildings and units and based on that information somehow try to find out a proper reaction. We just can't skip this. To counter each individual strategy and unit type. They are not really infinite. If you need help, we can try to gather everything with the proper reaction.
Thinking about it, that's what CardinalAllin wrote down in detail, a basic BO and gameplan for T. And that is all good, but as he also wrote, the AI strategy might differ from the human strategy, and i think you can differ a bit more, and you should really build on the strength of the AI, which is multitasking. So constant but light harass would be really good i think, because its just small resources for the bot, but it distracts the human player, causing mistakes. I think squeezing in 1-2 wraith here and there would really be good. If you want to go a little bit further you might also change that basic BO for a factory port start to have vults and wraiths, from the start. But really not overdoing it, just getting a few while expanding and switching to M&M while maintaining the harass. I'm sure CardinalAllin has vods for this build. Another important thing would be constant scv or other unit scouting, and all map control/base denial. Another: once we have 2-3 base, expand even more (with that amount of money, squeezing in 400 money for a new base is not a big deal, and AI multitask can handle new bases).
As for how to implement this into the bot, i don't know. Pezsmapatkany (sas.Pez) has some interest in this and he is a programmer with excellent game knowledge.
EPT
