Once the 4th rax has finished, pull half your SCVs and poke his ramp with your marines.
Pulling means you take them away from what they are currently doing.
So, you pull half of your SCVs away from your resources, but to do what?
This is pointed out in the next sentences:
You can first place a bunker outside his ramp to give you micro space. Once you're up the ramp, build another bunker and start targeting workers. Make sure to keep producing from your barracks and making depots.
Basically, you send some SCVs along behind your marines so that they can build bunkers and repair those bunkers, the other SCVs you have pulled are useful for supply depots and taking damage.
Every time that a SCV dies instead of a Marine when you are pushing him, it's in your advance.
Why does this work? As you are solely producing marines that cost 50 minerals and placing some buildings that cost 100 minerals, you are able to work with half the mineral income and still have constant production running.
When Terran pulls SCVs, it is strongest against zerg; however Mega Rax is also the hardest to hide from zerg.
In contrast with other races, Zerg is the only race that needs to decide whether he needs to spent his larvae on workers or on army units, as he is under pressure he will need to produce more army units at the cost of his workers. So, if you pull SCVs which gets your economy down, your opponent his economy will also be down because of the pressure that you apply on his production...
Short answer: Scout, detect the roach build before the Roach Warren, and change your build as necessary. Alternatively, start with marine/marauder and change to your hellion build if you detect a non-roach build.
Long answer: Let's assume you detected the Roach Warren 10 seconds after it went up.
If you detect the roach warren early enough, you can immediately throw a tech lab up (25 seconds) on your Barracks and build one marauder (30 seconds) in the same amount of time it takes the Zerg player to build their Roach Warren (55 seconds).
- 0: Roach Warren goes up.
- 10: Roach Warren detected.
- 15: Barracks starts building a tech lab; 2nd Barracks starts to be built.
- 40: Tech lab complete; Marauder 1 started.
- 55: Roach Warren complete, Roach 1-3 started.
- 70: Marauder 1, new, larva, Roach 4 started.
- 82: Roach 1-3.
- 95: Barracks 2, tech lab started.
- 100: Marauder 2. Marauder 3 started.
- 102: Roach 4.
- 120: Tech lab complete; Marauder #4 started.
- 125: Marauder 3. Marauder 5 started.
If at this point you see 4 roaches on your doorstep, you'll have 3 marauders complete (assuming a rush takes 23 seconds from when their 4th roach came out).
- 4 roaches = 32 DPS, 580 HP
- 3 marauders = 40 DPS, 375 HP
Statistically-speaking, the roaches will win without stim. If you researched stim, it'll be very close. If you had a bunker, you'd be fine.
Now, if they wait until they have 6 roaches:
- 85: Roach 5 started.
- 100: Roach 6 started.
- 112: Roach 5 completed.
- 127: Roach 6 completed.
- 150: Marauder 4 completed.
- 155: Marauder 5 completed.
Given the rush of 23 seconds, 6 roaches would appear in between your 4th and 5th marauders completing.
- 6 roaches = 48 DPS, 870 HP
- 5 marauders = 66.65 DPS, 626 HP
You're slightly better off. Without stim, it's close. With stim, you win. With a bunker or two, you are fine.
Note that this is all theory. Going for a hellion rush and having to bail out on it is not good news. Wasted time, resources, and a sense of panic can cause you to falter in this position. Even if you're able to stick to the example above, you won't be in an excellent position. You'll be forced into trying to scout to detect when the Zerg player changes their build to counter it and you'll be in a position where you're constantly reacting rather than being proactive and causing the Zerg player to react.
The best solution is to detect the roach build before the Roach Warren drops and get a head start.
If you watch a standard roach opener, you'll notice that rather than expanding after their 13/14 pool, the Zerg player is forced to save the minerals, wait until 18 to throw out the roach warren, and build up minerals for roaches rather than an expansion. When you don't see that expansion after the pool, it's a good sign that Zerg is going with a roach build. This is especially true if you do not see an upgrade happening at the spawning pool right after it is done. The upgrade happening at the spawning pool, in combination with an extractor being built early, usually hints at a baneling bust build instead.
If you swap over early, you'll be in a much better position than you would be if you waited for the Roach Warren to start building. You'll be proactive, the Zerg player will be reacting, and you'll have the upper-hand.
Best Answer
Dealing with cloaked banshees as any race on any map
Lets make it more generic, I'll leave the references to TvT in it, but I'm going to expand this answer to 'dealing with cloaked banshees as any race on any map'
Your Build
First off, a 1-1-1 as Terran is an excellent way to defend the banshee rush. The best builds with all races are the ones that can defend you against the most variations of attacks. Chances are you wont be able to scout the banshees before you have to commit to a build (like a 1-1-1 or hatch on 15). Stick with what you are most comfortable with in terms of build, then determine how best to defend with that build (all Terran builds can defend banshees with the exception of maurader heavy).
The Defense
Someone who commits to 2-3 banshees with cloak and doesn't do sufficient damage is punished enough. In terms of a pro-game, it is almost a GG to make that type of commitment and not win or cripple the opponent. This is called an all-in. The fact of the matter is that any of your solutions (as long as you live) will punish him! He is making a significant investment and you surviving with minimum loses puts him behind.
Relying solely on static defense like 2 turrets, cannons, or spore crawlers: You defend against cloak but will have a hard time killing the banshees. You need an army to deal with the mobility of the banshees. Killing the banshees is key, chances are he has maybe ~6 marines + 2-3 banshees as the entire extent of his army. If you throw down two turrets you will be able to detect him and prevent damage to your worker line, but chances are his banshees will escape alive and be repaired. You don't want them coming back...
A proper defense of banshees with cloak is minimum detection (maybe 1-2 turrets max for any race, or a cloak detection unit like the overseerer or observer) and the ability to kill the banshees. You can do this with most ranged units if you know the banshees are coming:
It is important to note that a scan or a static defense to kill his banshees isn't the end of the world as they are such a commitment on his part. Just make sure you kill them.
The Follow Up
How you follow up is up to you.
Macro
If you are confident in your macro play you have already put him behind by surviving his attack. You can put him even more behind by expanding. Good players recognize that going for the instant kill is often a high risk situation, the way to exploit an advantage in starcraft is to get 'more ahead'. Advantages build exponentially in Starcraft 2. You can safely expand knowing he has a smaller army then you due to his commitment to high tech. This lets you grow a larger economy earlier leading to an even larger army later. This is especially true with Zerg as their economic advantage can painfully punish mistakes like this from Terran later in the game. I will pick this solution every time. A definite win in ten minutes is better then a 90% chance to win now.
Micro
If you think your opponent is better in a long term macro game then you, you can attempt a counter attack immediately after you kill his banshees. This has a much higher risk to losing your advantage and could leave you on equal footings with him (or worse!). There are variations to the extent you counter attack though, you could probe (which is always good) and if he looks weak enough throw into an all-in with scv's and hope for a win or you can simply probe, back off and choose the macro alternative.