|
On March 03 2012 11:18 SniXSniPe wrote:Show nested quote +On March 03 2012 09:25 SeaSwift wrote:
Woah, in TvP Terran drops below 50% in Korea for the first time and remains on top everywhere else, and all the Terran whiners come crawling out of the woodwork. I don't know what you're looking at. International and Korean graphs both have at least 2 or more months where T has fallen below 50% in TvP. November, January, February for TvP have fallen below 50% for T to correct you (for the Korean graph). Show nested quote +On March 03 2012 11:01 xrapture wrote:On March 03 2012 10:57 Catatonic wrote:On March 02 2012 18:02 Molybdenum wrote:On March 02 2012 17:58 Megaman_X wrote:On March 02 2012 17:55 HaiFiSCH26 wrote:
I hope that people on ladder will stop whining now taht zerg is up,difference in ZvP is only 1% and as a whole it looks faitly balanced. or difference is ~18% in korea And just the month before it was about 14% different in favor of zerg. The metagame is shifting like crazy, and yet Blizzard is putting out balance patches when things aren't clearly in favor of one race or another. Cause Terran winning almost everything last year shows balance. These patchs are extreamly needed hence the shift in power where at one time Terran always won an now they're down to where every one else is terran winning everything? Last time I checked a foreign Terran hasn't won a major live event in well over a year... Have you ever heard of Polt?
have you ever heard of reading comprehension?
|
international looks really really good, korea on the other hand, not so much  (
|
On March 03 2012 11:18 SniXSniPe wrote:Show nested quote +On March 03 2012 09:25 SeaSwift wrote:
Woah, in TvP Terran drops below 50% in Korea for the first time and remains on top everywhere else, and all the Terran whiners come crawling out of the woodwork. I don't know what you're looking at. International and Korean graphs both have at least 2 or more months where T has fallen below 50% in TvP. November, January, February for TvP have fallen below 50% for T to correct you (for the Korean graph). Show nested quote +On March 03 2012 11:01 xrapture wrote:On March 03 2012 10:57 Catatonic wrote:On March 02 2012 18:02 Molybdenum wrote:On March 02 2012 17:58 Megaman_X wrote:On March 02 2012 17:55 HaiFiSCH26 wrote:
I hope that people on ladder will stop whining now taht zerg is up,difference in ZvP is only 1% and as a whole it looks faitly balanced. or difference is ~18% in korea And just the month before it was about 14% different in favor of zerg. The metagame is shifting like crazy, and yet Blizzard is putting out balance patches when things aren't clearly in favor of one race or another. Cause Terran winning almost everything last year shows balance. These patchs are extreamly needed hence the shift in power where at one time Terran always won an now they're down to where every one else is terran winning everything? Last time I checked a foreign Terran hasn't won a major live event in well over a year... Have you ever heard of Polt?
Have you ever heard of Polt? He said foreign terran, and polt is korean.....
|
On March 03 2012 10:30 HyperionDreamer wrote:Show nested quote +On March 03 2012 09:09 ePLocust wrote:On March 03 2012 09:00 GodZo wrote:
The game is pretty balanced, Zerg is slightly behind in Korea as I expeted.
The maps influence a lot. Too bad there aren't enough Korean games to make a good judgement off of them. The sample size is too small it would be nice to see a large sample of korean games and how it balances out then. To be honest a sample size of 1100 is large enough to make a lot of statistically valid conclusions, especially when the hypothesis is as extreme as an 8% variance in ZvP. Resorting to the "sample size of 1000 too small" argument is very weak at this point. Would a fair coin ever result in 8% deviance over 1000 flips? No. Of course a complex game like starcraft is completely different than the statistical flip of a coin, but saying a sample size of 1000 is too small is simply incorrect. I've heard of new drug trials conducted by the FDA on humans with a sample size of less than 200, which seems incredibly small considering how strict the regulations are on the production and sale of artificial drugs/supplements/medicines. I would be really interested to see more analysis with regards to maps, if that's at all possible. Note: I am not making any comments about units, races, or patches. I am saying that 8% over 1000 games is statistically significant this month, just as last month it was 6% the other way (zerg favored). I wonder if there's a new style/build that's becoming prevalent amongst Korean Protosses that's causing this 14% swing over one month.
The sample size isn't 1100. There were 1115 PvZ in korea in the time period of march 2011 to feb 2012. Feb 2012 pvz has a sample size closer to a tenth of that. Which, really, is too small.
|
On March 03 2012 06:10 ACrow wrote:Show nested quote +On March 03 2012 04:21 shizna wrote:
...
imo zealots, HT, colossus and archon are far too versatile compared to terran units.
... Absolutely. Marines totally need to be more versatile.
the difference is marines actually require skill to use late game while zealots/bl/archones/colo are just a-move units
|
On March 03 2012 11:37 dainbramage wrote:+ Show Spoiler +On March 03 2012 10:30 HyperionDreamer wrote:Show nested quote +On March 03 2012 09:09 ePLocust wrote:On March 03 2012 09:00 GodZo wrote:
The game is pretty balanced, Zerg is slightly behind in Korea as I expeted.
The maps influence a lot. Too bad there aren't enough Korean games to make a good judgement off of them. The sample size is too small it would be nice to see a large sample of korean games and how it balances out then. To be honest a sample size of 1100 is large enough to make a lot of statistically valid conclusions, especially when the hypothesis is as extreme as an 8% variance in ZvP. Resorting to the "sample size of 1000 too small" argument is very weak at this point. Would a fair coin ever result in 8% deviance over 1000 flips? No. Of course a complex game like starcraft is completely different than the statistical flip of a coin, but saying a sample size of 1000 is too small is simply incorrect. I've heard of new drug trials conducted by the FDA on humans with a sample size of less than 200, which seems incredibly small considering how strict the regulations are on the production and sale of artificial drugs/supplements/medicines. I would be really interested to see more analysis with regards to maps, if that's at all possible. Note: I am not making any comments about units, races, or patches. I am saying that 8% over 1000 games is statistically significant this month, just as last month it was 6% the other way (zerg favored). I wonder if there's a new style/build that's becoming prevalent amongst Korean Protosses that's causing this 14% swing over one month. The sample size isn't 1100. There were 1115 PvZ in korea in the time period of march 2011 to feb 2012. Feb 2012 pvz has a sample size closer to a tenth of that. Which, really, is too small.
Are you sure that's not just Feb? (I'm not sure, the graph for ZvP doesn't specify whether the number at the top is the total or the newest month).
Edit: The overall graph shows around 360 games for the month of Feb, so yeah, the sample size is about 130. However, you can still do some calculations using p = 0.05 reasonably well with 130 trials. For example, a perfectly balanced match up would be 50-50 for each race. If we want to consider the probability of seeing 8% deviance from the "fair 50/50" status, the math would be as follows:
Note: this probably requires first-second year university statistics knowledge.
8% deviance from 50/50 is 70.2 games won (130/2 = 65*1.08 = 70.2), but let's say 70 games. The probability of seeing at LEAST (this is required for using p-value calculations) 70 games won and 60 games lost could be checked using a two-tailed binomial distribution test. (We use two-tailed since we could just as easily see Zerg winning 8% more of the games as Protoss, the null hypothesis being to check the statistical significance of 8% win distribution assuming a statistically fair matchup, p level = 0.05, sample size 130).
We would "expect" (initially assuming null hypothesis is true) 65 games won for each race. However we see 8% in each direction, so a two-tailed 8% distribution would be a range from 60 to 70 games. Thus, we must use the binomial distribution to find the probabilities of finding the 8% result, or a more extreme result (this is required to use p-testing), and sum them up. So, we would be summing up all the probabilities from 0 games - 60 games, and 70 games to 130 games. This is represented by the following formula:
![[image loading]](http://i.imgur.com/LFs3K.gif)
In this expression, b(130;k;0.5) represents the binomial mass function with 130 trials, k successes (ranges from 0-60 and 70-130 as stated above) and a probability of 0.5, assuming the matchup is "fair". It represents the probability that in 130 games, there will be k wins by Protoss (or Zerg, it doesn't matter since the probability is 50%). Also, B(130;60,0.5) represents the cumulative distribution function, which represents the probability that in 130 games, there will be less than or equal to k wins by Protoss/Zerg. I use the 1-B(130;70,0.5) since I want the area above the 70-win line, and the cdf gives me the area below. Basic probability theory states that the total area under the binomial distribution is 1, so 1 - B(130;70,0.5) gives me what I want.
Plugging this into MATLAB, I get P = 0.3824, which shows that you are correct in saying this data is NOT statistically significant using a 5% confidence threshold (this is a pretty standard number). I wanted to just delete this post, but using some words of wisdom from xkcd, you can't not print something just cause you don't want it to be true.
Note: If you use the same method with 1100 games and 8% deviance (two tailed range of 506-594), then you get P = 0.0080, which would be HUGELY statistically significant, suggesting the original null hypothesis is incorrect. So it is correct to currently say that we do not have the data to suggest that the null hypothesis is false, and we can still consider it as a "statistically balanced" matchup. However, I might be able to do something with weighted averages of wins over the past 1 year, which would give me a sample size of 1100. I'd be able to weight the average based on how long ago the games occurred, so more recent games have more effect on what I'd call the time adjusted win total. After I finish studying for midterms, I might mess around with this. Hopefully this post was at least moderately understandable for anyone who has not take statistics.
|
On March 03 2012 13:59 HyperionDreamer wrote:Show nested quote +On March 03 2012 11:37 dainbramage wrote:+ Show Spoiler +On March 03 2012 10:30 HyperionDreamer wrote:Show nested quote +On March 03 2012 09:09 ePLocust wrote:On March 03 2012 09:00 GodZo wrote:
The game is pretty balanced, Zerg is slightly behind in Korea as I expeted.
The maps influence a lot. Too bad there aren't enough Korean games to make a good judgement off of them. The sample size is too small it would be nice to see a large sample of korean games and how it balances out then. To be honest a sample size of 1100 is large enough to make a lot of statistically valid conclusions, especially when the hypothesis is as extreme as an 8% variance in ZvP. Resorting to the "sample size of 1000 too small" argument is very weak at this point. Would a fair coin ever result in 8% deviance over 1000 flips? No. Of course a complex game like starcraft is completely different than the statistical flip of a coin, but saying a sample size of 1000 is too small is simply incorrect. I've heard of new drug trials conducted by the FDA on humans with a sample size of less than 200, which seems incredibly small considering how strict the regulations are on the production and sale of artificial drugs/supplements/medicines. I would be really interested to see more analysis with regards to maps, if that's at all possible. Note: I am not making any comments about units, races, or patches. I am saying that 8% over 1000 games is statistically significant this month, just as last month it was 6% the other way (zerg favored). I wonder if there's a new style/build that's becoming prevalent amongst Korean Protosses that's causing this 14% swing over one month. The sample size isn't 1100. There were 1115 PvZ in korea in the time period of march 2011 to feb 2012. Feb 2012 pvz has a sample size closer to a tenth of that. Which, really, is too small. Are you sure that's not just Feb? (I'm not sure, the graph for ZvP doesn't specify whether the number at the top is the total or the newest month).
at the top it says 363 games overall in february (all 3 races, non mirror obv), so it can't be 1100 PvZ games.
It's not rocket science.
|
On March 03 2012 14:07 ChaosTerran wrote:
at the top it says 363 games overall in february (all 3 races, non mirror obv), so it can't be 1100 PvZ games.
It's not rocket science.
Edited the above to be much more comprehensive. No need to be condescending, I simply didn't see it the first time. 
|
On March 03 2012 11:01 ChaosTerran wrote:Show nested quote +On March 03 2012 09:25 freetgy wrote:On March 03 2012 09:00 ChaosTerran wrote:
The thing is that it's entirely luck based. You see 2-3 gates and a robo and no robo bay (maybe proxied or maybe just at the other side of the protoss base) so what you expect is either a warp prism all in or immortal all-in and in both cases the correct response is => make more marines. But then he comes with 2 collossi and the only reason you lost is because he proxied a building and was hiding his collossi. It's extremely annoying and can make terrans look like complete noobs simply because they reacted to the wrong build. But it's not even a mistake, you scan and see the buildings and you have to assume the all-in is coming according to the tech you scout, missing 1 building is actually deadly in this case. Protoss always acts like it's so easy to read their all-ins. I'll tell you what, I've had protoss players fake a blink stalker all-in and then died to DTs (yeah that's kind of dumb, but you scan his base see some stalkers and a chronoboost on the twilight -> what are you going to assume?). The thing that annoys me about tvp early game is that you have to prepare for absolutely everything as terran, you can scout something and it can be something completely different in the end, if you don't get an engineering bay and a turret you die to dts, but if hes not going dts and actually does what you scouted (blink stalkers) you are short 250 mins almost for absolutely nothing.
I seriously think that the game is broken in that regard and protoss should be abusing this alot more, it just keeps the terran guessing, spending money that he really doesn't want to spend and then lategame is a walk in the park for toss anyway. yeah and you can say 100% the same for if you switch that protoss with terran. and terran has allins are even easier to execute and with better winrates and alot wider arsenal of allins At least terran can hold an all-in easily knowing it is coming, it not like that the other way around. Also terran has way better scouting options earlygame. Oh really. Name the all-ins terran has, 1-1-1 and............. big void. there is nothing else. nothing viable that can potentially kill a protoss player, terran either has to play completely standard or do some 1-1-1 all-in and they always hit at around 10 mins, by then a protoss players has already scouted your base for minutes.
The closest analogue I see to a 1-or-2 base warpgate all-in is a 1-or-2 base Stimpack all-in. Both have similar tech costs and timings, and both are quite deadly on 1-2 bases if the defending player is teching significantly without cutting econ, and can be performed with varied timings. (Immortal + Warpgate all-ins involve more tech. A ghost academy with Concussive Shells, or an Engineering Bay with +1 Infantry Weapons, would cost you a very similar infrastructure investment.) Terran also has the option of a no-gas mass marine all-in off 1-2 bases, which can be quite difficult to tell apart from a one-rax expand.
A basic all-in follows a normal opening until it suddenly stops spending on economy and tech, masses only units + unit-producing structures + supply for 2-3 minutes, and then attacks and streams reinforcements until someone is dead. When the opponent has pumped workers continuously and teched significantly while you built only units, your all-in will generally succeed. Recently, Terran players have been punished by bunker breaks after they spent a lot of resources to get a flourishing two-base economy with upgraded infantry and medivacs -- possibly with double engineering bays and/or a third in-base orbital -- but before their tech and economy can properly kick in, by two-base Protoss players who sat low on the tech tree and cut econ for a massive army at a particular attack timing between 8:00 and 10:00.
I'm fairly confident that a Protoss who makes similar deep investments into the future at the cost of the present (say, two-base +1/+1 with Charge and Obs before 10:00) will be similarly vulnerable to a two-base worker-cutting Terran all-in before the Protoss's tech completes.
|
On March 03 2012 16:37 Severedevil wrote:Show nested quote +On March 03 2012 11:01 ChaosTerran wrote:On March 03 2012 09:25 freetgy wrote:On March 03 2012 09:00 ChaosTerran wrote:
The thing is that it's entirely luck based. You see 2-3 gates and a robo and no robo bay (maybe proxied or maybe just at the other side of the protoss base) so what you expect is either a warp prism all in or immortal all-in and in both cases the correct response is => make more marines. But then he comes with 2 collossi and the only reason you lost is because he proxied a building and was hiding his collossi. It's extremely annoying and can make terrans look like complete noobs simply because they reacted to the wrong build. But it's not even a mistake, you scan and see the buildings and you have to assume the all-in is coming according to the tech you scout, missing 1 building is actually deadly in this case. Protoss always acts like it's so easy to read their all-ins. I'll tell you what, I've had protoss players fake a blink stalker all-in and then died to DTs (yeah that's kind of dumb, but you scan his base see some stalkers and a chronoboost on the twilight -> what are you going to assume?). The thing that annoys me about tvp early game is that you have to prepare for absolutely everything as terran, you can scout something and it can be something completely different in the end, if you don't get an engineering bay and a turret you die to dts, but if hes not going dts and actually does what you scouted (blink stalkers) you are short 250 mins almost for absolutely nothing.
I seriously think that the game is broken in that regard and protoss should be abusing this alot more, it just keeps the terran guessing, spending money that he really doesn't want to spend and then lategame is a walk in the park for toss anyway. yeah and you can say 100% the same for if you switch that protoss with terran. and terran has allins are even easier to execute and with better winrates and alot wider arsenal of allins At least terran can hold an all-in easily knowing it is coming, it not like that the other way around. Also terran has way better scouting options earlygame. Oh really. Name the all-ins terran has, 1-1-1 and............. big void. there is nothing else. nothing viable that can potentially kill a protoss player, terran either has to play completely standard or do some 1-1-1 all-in and they always hit at around 10 mins, by then a protoss players has already scouted your base for minutes. The closest analogue I see to a 1-or-2 base warpgate all-in is a 1-or-2 base Stimpack all-in. Both have similar tech costs and timings, and both are quite deadly on 1-2 bases if the defending player is teching significantly without cutting econ, and can be performed with varied timings. (Immortal + Warpgate all-ins involve more tech. A ghost academy with Concussive Shells, or an Engineering Bay with +1 Infantry Weapons, would cost you a very similar infrastructure investment.) Terran also has the option of a no-gas mass marine all-in off 1-2 bases, which can be quite difficult to tell apart from a one-rax expand. A basic all-in follows a normal opening until it suddenly stops spending on economy and tech, masses only units + unit-producing structures + supply for 2-3 minutes, and then attacks and streams reinforcements until someone is dead. When the opponent has pumped workers continuously and teched significantly while you built only units, your all-in will generally succeed. Recently, Terran players have been punished by bunker breaks after they spent a lot of resources to get a flourishing two-base economy with upgraded infantry and medivacs -- possibly with double engineering bays and/or a third in-base orbital -- but before their tech and economy can properly kick in, by two-base Protoss players who sat low on the tech tree and cut econ for a massive army at a particular attack timing between 8:00 and 10:00. I'm fairly confident that a Protoss who makes similar deep investments into the future at the cost of the present (say, two-base +1/+1 with Charge and Obs before 10:00) will be similarly vulnerable to a two-base worker-cutting Terran all-in before the Protoss's tech completes.
except force fields completely crush any early all in from terran. It's why 3 rax has been dead since a month after release.
|
Balance looks fine, even korea looks not that bad. Error bars in korea are that large due to the sample size. Even in a perfectly balanced game such situations would occur. The fact that zerg was a little low in korea over multiple month may indicate a slight problem, but it could as well be statistical noise (also single players include a lot into the statistics. Often a graph would look different if only one player wasn't there). If blizzard wan't to patch, then maybe patches that make certain matchups more interresting, otherwise I guess the game can be balanced by maps.
|
On March 03 2012 14:07 ChaosTerran wrote:
at the top it says 363 games overall in february (all 3 races, non mirror obv), so it can't be 1100 PvZ games.
It's not rocket science.
117 PvZ/ZvP, see my post earlier in this thread:
On March 02 2012 23:44 ulan-bat wrote:Some stats about the match-up specific graphs for Korea: ZvT or TvZ 106 games 57 different players -  Noblesse 8.5% (9 games) -  Curious 8.5% (9 games) ZvP or PvZ 117 games 69 players -  Extreme 9.4% (11 games) PvT or TvP 140 games 79 players - Tear 8.6% (12 games)
|
On March 03 2012 16:45 xrapture wrote:Show nested quote +On March 03 2012 16:37 Severedevil wrote:On March 03 2012 11:01 ChaosTerran wrote:On March 03 2012 09:25 freetgy wrote:On March 03 2012 09:00 ChaosTerran wrote:
The thing is that it's entirely luck based. You see 2-3 gates and a robo and no robo bay (maybe proxied or maybe just at the other side of the protoss base) so what you expect is either a warp prism all in or immortal all-in and in both cases the correct response is => make more marines. But then he comes with 2 collossi and the only reason you lost is because he proxied a building and was hiding his collossi. It's extremely annoying and can make terrans look like complete noobs simply because they reacted to the wrong build. But it's not even a mistake, you scan and see the buildings and you have to assume the all-in is coming according to the tech you scout, missing 1 building is actually deadly in this case. Protoss always acts like it's so easy to read their all-ins. I'll tell you what, I've had protoss players fake a blink stalker all-in and then died to DTs (yeah that's kind of dumb, but you scan his base see some stalkers and a chronoboost on the twilight -> what are you going to assume?). The thing that annoys me about tvp early game is that you have to prepare for absolutely everything as terran, you can scout something and it can be something completely different in the end, if you don't get an engineering bay and a turret you die to dts, but if hes not going dts and actually does what you scouted (blink stalkers) you are short 250 mins almost for absolutely nothing.
I seriously think that the game is broken in that regard and protoss should be abusing this alot more, it just keeps the terran guessing, spending money that he really doesn't want to spend and then lategame is a walk in the park for toss anyway. yeah and you can say 100% the same for if you switch that protoss with terran. and terran has allins are even easier to execute and with better winrates and alot wider arsenal of allins At least terran can hold an all-in easily knowing it is coming, it not like that the other way around. Also terran has way better scouting options earlygame. Oh really. Name the all-ins terran has, 1-1-1 and............. big void. there is nothing else. nothing viable that can potentially kill a protoss player, terran either has to play completely standard or do some 1-1-1 all-in and they always hit at around 10 mins, by then a protoss players has already scouted your base for minutes. The closest analogue I see to a 1-or-2 base warpgate all-in is a 1-or-2 base Stimpack all-in. Both have similar tech costs and timings, and both are quite deadly on 1-2 bases if the defending player is teching significantly without cutting econ, and can be performed with varied timings. (Immortal + Warpgate all-ins involve more tech. A ghost academy with Concussive Shells, or an Engineering Bay with +1 Infantry Weapons, would cost you a very similar infrastructure investment.) Terran also has the option of a no-gas mass marine all-in off 1-2 bases, which can be quite difficult to tell apart from a one-rax expand. A basic all-in follows a normal opening until it suddenly stops spending on economy and tech, masses only units + unit-producing structures + supply for 2-3 minutes, and then attacks and streams reinforcements until someone is dead. When the opponent has pumped workers continuously and teched significantly while you built only units, your all-in will generally succeed. Recently, Terran players have been punished by bunker breaks after they spent a lot of resources to get a flourishing two-base economy with upgraded infantry and medivacs -- possibly with double engineering bays and/or a third in-base orbital -- but before their tech and economy can properly kick in, by two-base Protoss players who sat low on the tech tree and cut econ for a massive army at a particular attack timing between 8:00 and 10:00. I'm fairly confident that a Protoss who makes similar deep investments into the future at the cost of the present (say, two-base +1/+1 with Charge and Obs before 10:00) will be similarly vulnerable to a two-base worker-cutting Terran all-in before the Protoss's tech completes. except force fields completely crush any early all in from terran. It's why 3 rax has been dead since a month after release.
I can tell you from personal experience that 3-rax (i'm talking about 2 tech 1 reactor) is definetly NOT dead. It pwns nexus first, 1 gate fe, and beats basically every all-in (except 1 base collosi and dt's). I use it as my standard build in high masters, and i have a pretty good winrate just using the 3-rax (2 tech 1 reactor variation) every game.
|
On March 02 2012 17:47 Ctuchik wrote:Hey Guys, My charts with TLPD results for February are done: http://imgur.com/a/1aAfuNote that the y-axis is now consistent between regions (30-70). Versions for R/G color blind are in the gallery. Edit: Thanks Mods!
Can you provide the datatables from which you are extracting these charts ?
And do you keep track of the length of the games in them ? (thus making it easy to have a separate chart
for short and late games).
|
On March 02 2012 22:04 DarQraven wrote:Show nested quote +On March 02 2012 21:54 Mentalizor wrote:On March 02 2012 21:33 StarscreamG1 wrote:Blizzard, if you're reading this..  TvP fixes to early/late game: Banshee from light to armored; Buff Viking ground attack. Lol, you're kidding right? Making banshees armored would actually make stalkers better against them... Making vikings better on ground? Are you kidding? It's not their real purpose, so it's obvious they should not have good ground attacks. TvP doesn't seem super imbalanced to me. What league and server are you? General concensus seems to be that PvT is not necessarily imbalanced percentage-wise, but timing wise. Short games are almost always won by the T, while, the longer the game goes, the more the P start winning. This shows a dynamic where Terran are very, very threatening to Protoss during the early and midgame, but it becomes hard to beat Protoss once they get their macro going. There can be all sort of explanations for this, though. Terrans love to jump on the "Protoss a-move OP, deathball untouchable" bandwagon, but as many players have demonstrated already, it is perfectly possible to beat P lategame. Harder, yes, but the same can be said about PvT early game and it is not impossible (see: Genius proxy Stargate vs Alive). This mostly seems to be players projecting their own ladder experiences onto these graphs. Meanwhile, statistics like game length do not take into account how the game actually got to that point - if the only reason PvT's ever get long is because a T failed their early/mid aggression and are behind economically, it should come as no surprise that they are losing out in the late game, for instance. Possible explanations like these would never surface from stats alone, so I really think this is up to Blizzard to analyze and fix. It really is a damn shame we have so few good Random players around, since I'd say those are the only ones who could really see the matchups honestly from both sides. From the Protoss perspective, Terran is just frustratingly strong early on. Your units are going to be running after them, not really hitting anything. You'll get dropped in multiple locations and even if you split your army up perfectly, you're still going to take more losses than the T because of the "T>P in low numbers" dynamic. This viewpoint completely discounts how the Terran is managing two drops/fights as well, though. Meanwhile, from the Terran perspective, deathballs can seem untouchable - as long as you're on the receiving end. From the P side of things, even with a 200/200 3/2 upgraded deathball, I still shit my pants right before I engage a Terran because even slight mistakes like getting EMP'd can completely destroy you. As for my own opinion: Tone down Terran early game threat/flexibility a bit (so that Protoss actually has reliable non-cheese ways of threatening a Terran early on, then take a look at if T can still beat P and make changes to P lategame accordingly. I don't suggest this approach because I want Terrans to stop allinning me or anything, but it just doesn't make sense to address lategame issues without a solid early game - the latter shapes the former, so you can't see them as separate issues. It seems like very bad game design to me to have one race play dominant 'attacker' for most of the game, while the other race only gets to play their strong suit once the other player has failed/been thwarted. That would create a very volatile and unforgiving matchup.
Early game TvP tends to favor Protoss as long as the Protoss in question is good with force fields. It's the midgame where Protoss is trying to transition into either robo or templar tech that Terran gains a momentary but heavy advantage. The problem stems from Terran's forced reliance on T1 units (and the relative quality of said units) thanks to the uselessness of mech against Protoss. There's no reason for Terran to tech up to anything expensive, and so mass production of marine/marauder continues unabated and creates a nice window where Terran can often do a ton of damage simply by having a bunch of units out while Protoss is trying to get to T3. And then of course after that Protoss gains the upper hand with the deathball that crushes my soul with how anti-StarCraft it is.
The matchup is pretty much a mess on all levels, and most of it stems from Protoss being poorly designed. Warp gate necessitates stalker/zealot being bad against barracks units early otherwise 4gate would be unstoppable. This requires existence of a spell like force field to keep Protoss from dying if Terran decides to attack. The issue is compounded by the ludicrously overpowered colossus having to make up for the poor DPS of gateway units compared to MMM but unlike tanks and brood lords, it's pretty mobile and much less risky to field in addition to its superior firepower.
|
On March 03 2012 17:21 forsooth wrote:
The matchup is pretty much a mess on all levels, and most of it stems from Protoss being poorly designed. Warp gate necessitates stalker/zealot being bad against barracks units early otherwise 4gate would be unstoppable. This requires existence of a spell like force field to keep Protoss from dying if Terran decides to attack. The issue is compounded by the ludicrously overpowered colossus having to make up for the poor DPS of gateway units compared to MMM but unlike tanks and brood lords, it's pretty mobile and much less risky to field in addition to its superior firepower.
Pretty much summarized all the problems with Tvp...
|
On March 03 2012 16:37 Severedevil wrote:Show nested quote +On March 03 2012 11:01 ChaosTerran wrote:On March 03 2012 09:25 freetgy wrote:On March 03 2012 09:00 ChaosTerran wrote:
The thing is that it's entirely luck based. You see 2-3 gates and a robo and no robo bay (maybe proxied or maybe just at the other side of the protoss base) so what you expect is either a warp prism all in or immortal all-in and in both cases the correct response is => make more marines. But then he comes with 2 collossi and the only reason you lost is because he proxied a building and was hiding his collossi. It's extremely annoying and can make terrans look like complete noobs simply because they reacted to the wrong build. But it's not even a mistake, you scan and see the buildings and you have to assume the all-in is coming according to the tech you scout, missing 1 building is actually deadly in this case. Protoss always acts like it's so easy to read their all-ins. I'll tell you what, I've had protoss players fake a blink stalker all-in and then died to DTs (yeah that's kind of dumb, but you scan his base see some stalkers and a chronoboost on the twilight -> what are you going to assume?). The thing that annoys me about tvp early game is that you have to prepare for absolutely everything as terran, you can scout something and it can be something completely different in the end, if you don't get an engineering bay and a turret you die to dts, but if hes not going dts and actually does what you scouted (blink stalkers) you are short 250 mins almost for absolutely nothing.
I seriously think that the game is broken in that regard and protoss should be abusing this alot more, it just keeps the terran guessing, spending money that he really doesn't want to spend and then lategame is a walk in the park for toss anyway. yeah and you can say 100% the same for if you switch that protoss with terran. and terran has allins are even easier to execute and with better winrates and alot wider arsenal of allins At least terran can hold an all-in easily knowing it is coming, it not like that the other way around. Also terran has way better scouting options earlygame. Oh really. Name the all-ins terran has, 1-1-1 and............. big void. there is nothing else. nothing viable that can potentially kill a protoss player, terran either has to play completely standard or do some 1-1-1 all-in and they always hit at around 10 mins, by then a protoss players has already scouted your base for minutes. The closest analogue I see to a 1-or-2 base warpgate all-in is a 1-or-2 base Stimpack all-in. Both have similar tech costs and timings, and both are quite deadly on 1-2 bases if the defending player is teching significantly without cutting econ, and can be performed with varied timings. (Immortal + Warpgate all-ins involve more tech. A ghost academy with Concussive Shells, or an Engineering Bay with +1 Infantry Weapons, would cost you a very similar infrastructure investment.) Terran also has the option of a no-gas mass marine all-in off 1-2 bases, which can be quite difficult to tell apart from a one-rax expand. A basic all-in follows a normal opening until it suddenly stops spending on economy and tech, masses only units + unit-producing structures + supply for 2-3 minutes, and then attacks and streams reinforcements until someone is dead. When the opponent has pumped workers continuously and teched significantly while you built only units, your all-in will generally succeed. Recently, Terran players have been punished by bunker breaks after they spent a lot of resources to get a flourishing two-base economy with upgraded infantry and medivacs -- possibly with double engineering bays and/or a third in-base orbital -- but before their tech and economy can properly kick in, by two-base Protoss players who sat low on the tech tree and cut econ for a massive army at a particular attack timing between 8:00 and 10:00. I'm fairly confident that a Protoss who makes similar deep investments into the future at the cost of the present (say, two-base +1/+1 with Charge and Obs before 10:00) will be similarly vulnerable to a two-base worker-cutting Terran all-in before the Protoss's tech completes.
On maps with a ramp to the natural....Those 1 and or 2 base all ins don't work. You only need like 4 forcefields to stall enough to get more units, or you just cut the army in half. Whatever, it just doesn't work either way.
|
On March 03 2012 16:59 kofman wrote:Show nested quote +On March 03 2012 16:45 xrapture wrote:On March 03 2012 16:37 Severedevil wrote:On March 03 2012 11:01 ChaosTerran wrote:On March 03 2012 09:25 freetgy wrote:On March 03 2012 09:00 ChaosTerran wrote:
The thing is that it's entirely luck based. You see 2-3 gates and a robo and no robo bay (maybe proxied or maybe just at the other side of the protoss base) so what you expect is either a warp prism all in or immortal all-in and in both cases the correct response is => make more marines. But then he comes with 2 collossi and the only reason you lost is because he proxied a building and was hiding his collossi. It's extremely annoying and can make terrans look like complete noobs simply because they reacted to the wrong build. But it's not even a mistake, you scan and see the buildings and you have to assume the all-in is coming according to the tech you scout, missing 1 building is actually deadly in this case. Protoss always acts like it's so easy to read their all-ins. I'll tell you what, I've had protoss players fake a blink stalker all-in and then died to DTs (yeah that's kind of dumb, but you scan his base see some stalkers and a chronoboost on the twilight -> what are you going to assume?). The thing that annoys me about tvp early game is that you have to prepare for absolutely everything as terran, you can scout something and it can be something completely different in the end, if you don't get an engineering bay and a turret you die to dts, but if hes not going dts and actually does what you scouted (blink stalkers) you are short 250 mins almost for absolutely nothing.
I seriously think that the game is broken in that regard and protoss should be abusing this alot more, it just keeps the terran guessing, spending money that he really doesn't want to spend and then lategame is a walk in the park for toss anyway. yeah and you can say 100% the same for if you switch that protoss with terran. and terran has allins are even easier to execute and with better winrates and alot wider arsenal of allins At least terran can hold an all-in easily knowing it is coming, it not like that the other way around. Also terran has way better scouting options earlygame. Oh really. Name the all-ins terran has, 1-1-1 and............. big void. there is nothing else. nothing viable that can potentially kill a protoss player, terran either has to play completely standard or do some 1-1-1 all-in and they always hit at around 10 mins, by then a protoss players has already scouted your base for minutes. The closest analogue I see to a 1-or-2 base warpgate all-in is a 1-or-2 base Stimpack all-in. Both have similar tech costs and timings, and both are quite deadly on 1-2 bases if the defending player is teching significantly without cutting econ, and can be performed with varied timings. (Immortal + Warpgate all-ins involve more tech. A ghost academy with Concussive Shells, or an Engineering Bay with +1 Infantry Weapons, would cost you a very similar infrastructure investment.) Terran also has the option of a no-gas mass marine all-in off 1-2 bases, which can be quite difficult to tell apart from a one-rax expand. A basic all-in follows a normal opening until it suddenly stops spending on economy and tech, masses only units + unit-producing structures + supply for 2-3 minutes, and then attacks and streams reinforcements until someone is dead. When the opponent has pumped workers continuously and teched significantly while you built only units, your all-in will generally succeed. Recently, Terran players have been punished by bunker breaks after they spent a lot of resources to get a flourishing two-base economy with upgraded infantry and medivacs -- possibly with double engineering bays and/or a third in-base orbital -- but before their tech and economy can properly kick in, by two-base Protoss players who sat low on the tech tree and cut econ for a massive army at a particular attack timing between 8:00 and 10:00. I'm fairly confident that a Protoss who makes similar deep investments into the future at the cost of the present (say, two-base +1/+1 with Charge and Obs before 10:00) will be similarly vulnerable to a two-base worker-cutting Terran all-in before the Protoss's tech completes. except force fields completely crush any early all in from terran. It's why 3 rax has been dead since a month after release. I can tell you from personal experience that 3-rax (i'm talking about 2 tech 1 reactor) is definetly NOT dead. It pwns nexus first, 1 gate fe, and beats basically every all-in (except 1 base collosi and dt's). I use it as my standard build in high masters, and i have a pretty good winrate just using the 3-rax (2 tech 1 reactor variation) every game.
I demand replays! I seriously haven't seen any 3 Rax in ages. I did see a few win with 2 Rax Conc against 15 Nexus though.
Even then, it would mean that Terran has two all-ins : 1-1-1 and 3Rax stim.
|
I wouldn't even call the 3rax an all-in. Its purpose is to deny the opponents expansion while you get your own, but with stim timings being nerfed since forever it's probably fairly easy to hold off nowadays even if you fast expand, although I've had good success with it the few times I've used it.
|
On March 03 2012 06:10 ACrow wrote:Show nested quote +On March 03 2012 04:21 shizna wrote:
...
imo zealots, HT, colossus and archon are far too versatile compared to terran units.
... Absolutely. Marines totally need to be more versatile.
i don't want a protoss nerf, but sadly that seems to be blizzard's preferred way to deal with balance is to continually nerf stuff. the terran units are badly designed, therefore protoss feels imba compared to terran, therefore blizzard deny any responsibility for dumb terran units and nerf protoss instead.
i'd like them to preferably give terran some kind of semi-viable build besides blind marine marauder medivac... TvP is mind numbingly boring and every single game that goes past 15 minutes is like carbon copy of the last - because there's simply no other viable way to play it out :/
|
|
|
|
|
|
EPT
