After all the positive feedback I got for my post about the Phoenix micro (which I really apprecitate, by the way), I am back with another topic, Floating to the gold base (or Gangnam Terran, as some know it). The question about this strategy is, is it worth doing at all, and if yes, how much of an advantage does it give? In other words, isn't the delay in economy caused by the floating time greater then the boost the gold patches provide? Here is what I found.
Edit: the following calculation in the spoiler tag is flawed. Sorry for that! See the pictures below for correct info. + Show Spoiler +
First of all, the gold base provides an income only 5% greater than a regular base, which is not much at all. The key is that you need much less workers (less money invested, shorter time) to reach an income even higher (if not that much higher) than normally.
The way to examine the problem is to introduce the functions for total minerals mined as a function of time, for both the gold and the blue bases. These functions comprise little linear functions, each 17 gs long, because that is the time required for building an SCV. In such a 17-gs-long interval the mineral collection rate is constant, so the function is linear.
Now the gold function is delayed by the time it takes to fly to the gold base. If this delayed function intersects the blue function before the gold base is depleted, it means you have a window of opportunity, where you actually have an economic advantage, until the gold patches run dry.
For Habitation Station, the map where this strategy was frequently used, the graphs look like this: + Show Spoiler +
Surprisingly, the time it takes to fly to the gold is 45 gs on Habitation Station (much more than I expected), which results in the strategy only paying off after 6:17. The minerals are depleted around 13:58, when the economic advantage of the floating Terran is 227 minerals, so about 8 Marines, considering he needs 4 less SCVs, too. This may seem a good number of units, but at the 14 minute mark, it is not all that much (although it may decide the game in some cases). You can decide to attack earlier, though. At 6:17, you will already have a 200 mineral = 4 Marine advantage, which gradually grows.
On Foxtrot Labs, depending on the spawn location, it can last either about 58 game seconds or 1:30 game minutes to fly to the gold base. In the closer spawn, the gold base will first pay off at 10:50, and the base will deplete at 14:11, with an advantage of only 82 minerals. In the longer spawn, this strategy never pays off, with a deficit of 277 minerals when the base depletes. This build does not work on Foxtrot Labs, especially not on longer spawn. Do not try it!
The effect of MULEs may not be taken into account, becase they only start when the CC is upgraded; their extra income is partially compensated by the SCVs taken from minerals due to building; and they are not alway used (sometimes a scan is preferred). These three factors cancel out the effect of MULEs almost entirely, so the calculations should remain valid.
As in the previous one, I attach a (more) thorough pdf as well: here it is!
Edit: This is the correct calculation in the spoiler tag below: + Show Spoiler +
Is it really 47 seconds or a few less, as all your scvs will already have mined the gold once the CC landed. Also 4 less Workers translates to half a supply depot less needed. So I think the advantage is a tiny bit bigger than what you have shown.
Edit: aside from the eco, when it comes to allins there is also obviously the factor that you moved - at least on hab station - towards your opponent a bit so your barracks will also be closer which gives a few additional seconds as well.
Floating to the gold on Habitation Station does not take 47 seconds. Rather, I believe it is 42 or 43. In addition (I don't know if you accounted for this), the player's six SCVs will all have mined one set of minerals that they turn in the moment the CC lands.
On August 02 2014 22:44 EngrishTeacher wrote: Wait so TL;DR, not worth it?
Especially considering most all-ins hit much earlier than the 8 minute mark?
Korean pros should read this thread!
If you can pull it off well (Bbyong), it is worth it, but there are more factors than just the boost in mineral income.
You appear not to take into account that you need 4 less scvs to saturate the gold than the normal base. Well, I guess you did in the pdf as well as a few other caveats, but not in the post here. Perhaps it'll be better to post the contents of the pdf since it really isn't that long at all. At 7:53, the build is already "ahead" by 200 minerals simply by not spending that 200 minerals on scvs. Mineral income is not the only way to have more minerals for units; not spending on scvs is also part of it. Most strats using floating to the old base will be decide well before 12 minutes anyhow.
Really though, the best way to compare gangnam terran is to try to compare it with when gangnam terran is not used at all, and the differences in the unit/tech count and safety. Isn't there like a replay program that can count the amount mined? Mathematical modelling in this case isn't useful since you can physically go out and do it and then compare results.
On August 02 2014 22:49 Zealously wrote: In addition (I don't know if you accounted for this), the player's six SCVs will all have mined one set of minerals that they turn in the moment the CC lands.
No, he didn't account for it.
If he did, the mathematical model immediately breaks down. As soon as the CC lands on the gold, it receives 300 minerals and is ahead of the amount of the normal base mined. So the gold base is ahead at 47 game seconds (gs). A failure in mathematical modelling by oversimplification.
Edit: Uh not 300, 30 lol. Ignore what I wrote. Using the modelling, the amount mined is equalled at 7:00 mins.
On August 02 2014 22:57 Dangermousecatdog wrote: You appear not to take into account that you need 4 less scvs to saturate the gold than the normal base. Well, I guess you did in the pdf as well as a few other caveats, but not in the post here. Perhaps it'll be better to post the contents of the pdf since it really isn't that long at all. At 7:53, the build is already "ahead" by 200 minerals simply by not spending that 200 minerals on scvs. Mineral income is not the only way to have more minerals for units; not spending on scvs is also part of it.
I did take it into consideration in the pdf. The problem is, I can't upload it here on TL as a pdf, and it is full of tables and equations which I couldn't include in the post, so it has to be a pdf.
On August 02 2014 22:57 Dangermousecatdog wrote: Edit: Uh not 300, 30 lol. Ignore what I wrote. Using the modelling, the amount mined is equalled at 7:00 mins.
I believed that 300 for a moment! Actually, though, that 30 counts little in the grand scheme of things, but I included it in the pdf and I corrected the post as well. There seem to have been mistakes anyway, now it should be all right.
On August 02 2014 22:49 Zealously wrote: Floating to the gold on Habitation Station does not take 47 seconds. Rather, I believe it is 42 or 43.
You are right, it isn't 47, nor 43, but 45 seconds. I tested it again.
On August 02 2014 22:44 EngrishTeacher wrote: Wait so TL;DR, not worth it?
Especially considering most all-ins hit much earlier than the 8 minute mark?
No, TL;DR: it is worth on Habitation Station, so long as you attack after around 6:20. It is not worth on Foxtrot Labs in the long spawn. In the short spawn, if you attack at 8:00, you are about 100 minerals behind, but since you don't have to build 4 SCVs, you are barely ahead. If you attack much earlier, your situation is worse, of course.I think, though, that the map is worse for this strategy than Habitation Station in other aspects as well, and you restrain yourself to one build for minimal advantage, so I think even on the shorter spawn, it should not be used.
Yeah, sorry about the 300 minerals thing. I didn't expect to read mathematical language in the pdf and my brain got scrambled. 30 minerals is significant in that it basically cuts off around 50 seconds in the calculation for equal amount of minerals mined.
Anyhow, I don't think mathematical modelling in this case can prove useful as this is something that is best compared by physically doing it and comparing it. There's probably some replay programs that can measure income mined at certain points in the game, though you'll probably need bbyong's replays for accuracy.
I think one thing you might want to consider is that in the gold base you probably want to "oversaturate" your base: 3 extra scvs at 33.6 m/gmin is not that bad comparing to those 39 m/gmin at regular patches. I am not sure, but you can look at some replays see if pro players do it.
I can't do precise math here, but consider you spend extra 150 minerals around 3-4 gmin mark and gain ~100 m/gmin back per minute, you earn your extra investment back already before 6 minute mark, and every game minute after before you deplete the base you gain 100 mineral advantage. Seems worth it to me.
On August 03 2014 05:56 timchen1017 wrote: I think one thing you might want to consider is that in the gold base you probably want to "oversaturate" your base: 3 extra scvs at 33.6 m/gmin is not that bad comparing to those 39 m/gmin at regular patches. I am not sure, but you can look at some replays see if pro players do it.
I can't do precise math here, but consider you spend extra 150 minerals around 3-4 gmin mark and gain ~100 m/gmin back per minute, you earn your extra investment back already before 6 minute mark, and every game minute after before you deplete the base you gain 100 mineral advantage. Seems worth it to me.
Yeah, it seems to be worth it. I'll check some pro matches and test it.
Really nice post and very well detailed, it seems it can be made work into a sizable advantage as long as you hit at the right window. The secret is probably attacking at a timing when you both have an econ/army lead and also attacks right before the opponent can get crucial tech out.
I don't like Gangnam Terran style, the main reasons is that its doing an all in that is harder to macro out of than other all-ins due to building placements(sacking depots and the big opening you leave if you lift off all your buildings). Also the all in is strongly telegraphed, its like saying in your face that I will kill you within 14 minutes or lose. It would many times be better to do an all in without the opponent knowing and have an easier way of expanding.
I'm guessing it might be better at pro level though were the players really can squeeze out the extra value out of those 4-8 marines and where the opponent would scout whats coming anyway. In a normal diamond level game however I would say, not worth.
I don't like Gangnam Terran style, the main reasons is that its doing an all in that is harder to macro out of than other all-ins due to building placements(sacking depots and the big opening you leave if you lift off all your buildings). Also the all in is strongly telegraphed, its like saying in your face that I will kill you within 14 minutes or lose. It would many times be better to do an all in without the opponent knowing and have an easier way of expanding.
I'm guessing it might be better at pro level though were the players really can squeeze out the extra value out of those 4-8 marines and where the opponent would scout whats coming anyway. In a normal diamond level game however I would say, not worth.
To be fair, the protoss that Bbong plays against are among the best and they know it's coming, but they can't stop it anyways. So it really doesn't matter if the opponent scouts it.
There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
Alright, so what about proxy/early pool?
I don't have the actual facts to back this up, but given that Bbyong beat sOs, Parting and Stork (at least one of them knew he'd play Gangnam Terran) with this I would imagine that proxying, too, is ineffective.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
Alright, so what about proxy/early pool?
havent tested it because if they go reaper and dont float to the gold the game is over, but I assume the one depot barracks wall would not die in time and the one marine would eventually finish even if scvs have to run around forawhile.
I think psychologically floating to the gold terrifies me. If I scout that my mind instantly thought 'quick hit hard and fast, end it before somehow he overpowers me with mass production' and for some reason it has worked for me by making me harass like a boss and apply constant pressure. But like with the Phoenix micro, great job with the theory behind it,
Can we get some numbers on old country, the new 2v2 map? I've been toying around with the float to gold on that map, but I'm not sure if it's beneficial or not.
I absolutely love the discussion for hab stat but it makes me extremely sad to see all this theorycrafting yet the map isn't in the season mappool anymore. As much as it might seem selfish that I want Hab Stat to still be in the map pool because it's my creation, honestly I want it to still be in the map pool because it has been the only map in the entire HotS expansion that has actually produced different, meta changing & overall very entertaining games compared to every other map.
You look at all the memorial games on it, bomber vs scarlett, ty vs stork, pigbaby vs bomber & major vs violet to name a few and you see all these different playstyles which just wouldn't work in any other map. Overall, I think the main reason it got removed was because terran had a fairly poor win ratio on it, yet now we got all these terran buffs which overall would most likely improve the win rate.
Oh Psione, please PM me why you guys at blizzard decided to remove Hab Stat...it's not like it was getting boring or too "stale" to watch...
On August 03 2014 08:50 dsousa wrote: 1.4.3 ... got it
MULEs don't need accounting because they mine the same on gold/blue patches.
They actually do. If both the gold and the blue base mines faster by a certain amount, it changes things. Imagine if MULEs provided an additional income of, say, 10,000 per gmin. Then the workers' mining would be nothing compared to MULEs, and the delay caused by the floating would be so substantial that this strategy would never pay off. Of course, this is not the case; it is merely an illustration that MULEs can make an impact on the strategy even if they mine at the same speed on both patch types.
On August 03 2014 09:18 Jer99 wrote: Can we get some numbers on old country, the new 2v2 map? I've been toying around with the float to gold on that map, but I'm not sure if it's beneficial or not.
Yeah, of course! I totally forgot there were 2v2 maps as well.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
Alright, so what about proxy/early pool?
havent tested it because if they go reaper and dont float to the gold the game is over, but I assume the one depot barracks wall would not die in time and the one marine would eventually finish even if scvs have to run around forawhile.
The way to deal with it was around for a bit and was said by Catz, at the moment you spot it you charge in with all your drones, someone did it in proleague and won.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
Alright, so what about proxy/early pool?
havent tested it because if they go reaper and dont float to the gold the game is over, but I assume the one depot barracks wall would not die in time and the one marine would eventually finish even if scvs have to run around forawhile.
The way to deal with it was around for a bit and was said by Catz, at the moment you spot it you charge in with all your drones, someone did it in proleague and won.
That's why Z has the easiest auto-win. Your overlord spots it at a set time, and when you see it you pull all drones (unless you VERY early pool) because you have MORE workers and T don't have the $ to wall-off. You should have 1-2 drones popping to "macro". I've seen lots of "responses" both in pro-games than on player streams, and this one never failed. Sadly T - P don't have that reliable automatic scout, but worker pull is still and option if seen early.
Anyway I don't think the OP understand the real strength of this technique.
-Shorter rush distance without the risks of proxies (once wall is up it's very easy to defend vs low tech units), because what hits earlier hits harder. The fact your econ is catching up right when you're about to hit means that in theory you also have more reinforcements.
-Psy advantages, at least in the first week(s) peoples went on ape-shit craziness in response, that don't work well.
-If cheese doesn't kill you opponent, then you don't have this exposed base to defend after 15min where drops/ warp-prism, mutas and army-position near watch tower can really screw the gold.
-All-in-all a pressure build that is safe (except very vs early rush/worker pull) both early and in the late game, that can win if your opponent is greedy and give you an edge if he's too careful.
May I also doubt a little the numbers:
Less SCV = Less $ to spend, but don't forget SCVs don't cost "50minerals". You need supply to produce them, and supply depots cost not only 100 minerals but also lost mining time (20 normally, so about 28 on gold), amounting for a real SCV cost of 65 and 71 respectively (par 11 supply at least). Considering we start with 6 worker and 11-6 "free" supply, we get a total (2per patch) saturation cost of 5*50+(5*65)= 575 minerals for normal and 5*50+(1*71)= 321 for gold, 254 minerals, or 5 (not 4) marines, 4 of those without hidden cost since they take place of the SCVs not built in a normal game.
It could also be argue that since each SCV is worth about 40% more mining, then each "hidden" construction cost should be 40% higher, bringing the price of a rax from 193 to 210 and the price of a supply from 15 to 21. The supply shouldn't matter much since most follow-up from flying to gold are very aggressive.
MULES can't be dismissed, they're an integral part of Terran economy, there's a good reason you always build orbital after 1st rax complete, unless you're cheesing or defending one. I don't play T, so I don't know if gold into rax wall-off vs normal 10 supply 11rax brings a faster rax or not. It's worth noting they worth a little less in gold base senarios because comparativly an SCV is worth more. Also "supply calldown" is less punishing, while not being as efficient than mules+building a depot.
I'd be also happy to consider the effect of good worker micro or even mineral boosting with a much more scalable income. When I see koreans doing it, the SCVs are already waiting the CC to land with gold mineral, then try directly saturate closer patches. If you plan those early "cheeses" at that level, every mineral counts.
All in all good work, very interesting. Sadly you seems to have dismissed everything other than scv saturation and that makes the experiment totally void of practical knowledge in my opinion. While your math is sound, it fails to answer your question : " is it worth doing at all".
I regret to tell everyone that my previous calculations seem to have been wrong. I updated the pdf and I will update the post itself as well soon. According to the new results, strictly economically, the Gangnam Terran strategy is very well worth on Habitation Station, fairly good on Foxtrot short spawn, barely OK on Old Country, and not worth on Foxtrot long spawn nor on Overgrowth (of course). Also, always 15 SCVs should be built, rather than 12.
On August 03 2014 07:56 darkness wrote: There is always a risk with this I think, what if you do this so frequently that your opponent sends their first worker to block? Even worse, proxy gateway or any early pool. I think it's a build that is good for one game only.
at one point every terran did this and I sent a drone at 0 seconds, and by the time I got to the watch tower they were already mining.
Alright, so what about proxy/early pool?
havent tested it because if they go reaper and dont float to the gold the game is over, but I assume the one depot barracks wall would not die in time and the one marine would eventually finish even if scvs have to run around forawhile.
The way to deal with it was around for a bit and was said by Catz, at the moment you spot it you charge in with all your drones, someone did it in proleague and won.
It must have been another match, but I clearly remember a Zerg pulling drones immediately after scouting it and then loosing terribly. That was in Proleague as well, if I remember correctly.
On August 04 2014 01:22 varsovie wrote: -Shorter rush distance without the risks of proxies (once wall is up it's very easy to defend vs low tech units), because what hits earlier hits harder. The fact your econ is catching up right when you're about to hit means that in theory you also have more reinforcements.
This is true, the calculations are strictly economic.
On August 04 2014 01:22 varsovie wrote: -Psy advantages, at least in the first week(s) peoples went on ape-shit craziness in response, that don't work well.
I don't think this should really matter at pro level.
On August 04 2014 01:22 varsovie wrote: Less SCV = Less $ to spend, but don't forget SCVs don't cost "50minerals". You need supply to produce them, and supply depots cost not only 100 minerals but also lost mining time (20 normally, so about 28 on gold), amounting for a real SCV cost of 65 and 71 respectively (par 11 supply at least). Considering we start with 6 worker and 11-6 "free" supply, we get a total (2per patch) saturation cost of 5*50+(5*65)= 575 minerals for normal and 5*50+(1*71)= 321 for gold, 254 minerals, or 5 (not 4) marines, 4 of those without hidden cost since they take place of the SCVs not built in a normal game.
This is mostly true. However, those if you work with 15 SCVs, you really don't have much less supply. The part on hidden costs is true, though.
I decided to disregard MULEs beacause they work in a way too complicated. Claiming that the 3 mentioned factors nullify the effect of MULEs is not exactly true, but it is approximately, I believe.
As for worker micro, it can provide about an extra 50 minerals compared to unmicroed mining on a regular base. The mineral functions I introduced already supposed perfect worker pairing, though.
I think you used the far spawn for old country; I think there are 4 possible spawn locations and they will each have different float times, the ones I am interested in are of course the spawns right next to the gold base, I'm almost 100% certain that the float time should be similar to habitation station just by looking at the map
On August 04 2014 22:46 Jer99 wrote: I think you used the far spawn for old country; I think there are 4 possible spawn locations and they will each have different float times, the ones I am interested in are of course the spawns right next to the gold base, I'm almost 100% certain that the float time should be similar to habitation station just by looking at the map
Uhm, yeah. In my defense, though, I have never played this map. I think the shorter spawn may even be closer than on Hab Station, but I'll have to check.
On August 04 2014 22:46 Jer99 wrote: I think you used the far spawn for old country; I think there are 4 possible spawn locations and they will each have different float times, the ones I am interested in are of course the spawns right next to the gold base, I'm almost 100% certain that the float time should be similar to habitation station just by looking at the map
The float time on Old Country close spawn is 41 gs, which is even better than Habitation Station (that being 45), so this style is even slightly better on Old Country. I don't know how it would play out in 2v2, especially that there are rocks in the middle of the map, iirc. Edit: pdf updated.
EPT![[image loading]](http://i.imgur.com/WlWVAfw.png)
![[image loading]](http://i.imgur.com/JKLN7Dp.png)
![[image loading]](http://i.imgur.com/7b2aAlW.png)
![[image loading]](http://i.imgur.com/OM66QUF.png)
![[image loading]](http://i.imgur.com/kzcHc51.png)
![[image loading]](http://i.imgur.com/wvceJkm.png)
![[image loading]](http://i.imgur.com/Tk7UGUe.png)
