[Tom]

UPDATED WITH LATEST DATA FOR JULY/AUGUST - SEE PAGE 9




Introduction

The worm has turned. It began almost impercetibly. Protoss players such as myself started questioning whether the balance complaints of other races were really justified in light of the results we were seeing at high level play.

I have previously said it would be unwise to jump to conclusions based on the results of a couple of high profile games or one tournament. I am still of that view. What I would like to discuss is the emergence of an apparent trend in poor protoss results in high level play. I say "apparent" trend because I think it is still too early to conclude that what we are seeing is actually a "trend".

Conscious of not jumping to a premature conclusion, it seems to me that the emerging data raises a legitimate question: is protoss currently underpowered in high level play?

Disclaimer

As a protoss player I have an interest in this debate. I will try to keep this to a minimum, but I am conscious that the very fact that I am raising it as a topic of discussion demonstrates an inherent bias on my part. I hope you will forgive me!

Some definitions

First, my discussion is limited to high level play. By "high level play", I mean top tournaments such as GSL, and Grandmasters league, with particular emphasis on the leagues in Korea, NA and Europe, which I suggest are the most competitive.

Second, I have carefully and deliberately used the word underpowered. By underpowered, I do not mean "unplayable" or "uncompetitive". Strong players will continue to perform well because their inherent skill allows them to overcome possible shortcomings with their race. Underpowered also involves questions of degree - it may be very slight, such that the effect on lower level players is low or almost negligible, but enough to have a significant effect at higher levels.

Third, the expression underpowered in high level play is important. It may also be legitimate to ask "is protoss overpowered in low level play?", having regard to the race's arguably simpler mechanics (I'll leave this for others to decide).

Why it is important to look at data

In any non-mirror match up, it seems to me there are three key variables that affect the outcome of the game. The first, and probably most significant, is the player (or more correctly, players). The second variable is race. The third is map. In examining the relative "power" of a race, we are trying to eliminate the "noise" that is created as a result of differences in player skill and different maps.

For this reason, I suggest (although you may disagree) it is virtually useless to examine anecdotal experience - ie a player's personal experience of the race and their recent games. This is because the key variable here is the player, not the map or race. But, when we look at the results of hundreds (or thousands) of games, the significance of the "player" diminishes drastically, whereas the importance of "race" (and "map") dramatically increases.

I acknowledge that the following data does not attempt to isolate "map" as a factor, and that this is a significant limitation. I apologise in that the data is simply not available.

Recent tournament results [SPOILERS]

In early May, an author on TeamLiquid performed an analysis of global and Korean tournament results. The thread is here: http://www.teamliquid.net/forum/view...opic_id=218558

In summary, what those results showed was that in global tournaments, the win rate for all match ups had begun to approach 50% (although I would be interested to see the current results more than one month (and 1 patch) later). However, the results showed Protoss was been getting absolutely murdered in recent Korean tournaments, with win rates of 33% of ZvT and 30% of ZvP. I said at the time I would be interested to see whether this was the start of an emerging trend (Korea tends to lead the field in all things Starcraft).

The results of the current GSL "Super Tournament" have been, if anything even more dramatic. 16 protoss players qualified for the round of 64. This represents 25% of the field, despite the fact that Protoss players make up approximately 35% of active 1v1 players in Korean. Of these, 6 advanced to the round of 32 (two of whom advanced in mirror match ups). Three Protoss players have so far played in the round of 32 (Genius, HongUn and Trickster) and all have been knocked out. I pray that at least one of the remaining three advances to the top 16, so that there is at least someone I can watch to pick up some tips on how to play the race at the moment. But I think there is a real risk at the moment that we will have a quarter final (or even round of 16) with not a single protoss player.

I note in passing that the number of Korean protoss players complaining about balance has (as of yesterday), overtaken zerg players for the first time in many months (see http://www.playxp.com/sc2/jingjing/ - red = zerg, green = terran, blue = protoss, purple = nothing). Of course, all this demonstrates is "sentiment" (what people think about balance) rather than an actual indicator of balance.

Grandmasters League statistics

The number of Protoss players in the Korean Grandmasters League has remained the same as when I last examined the data. 32% of Korean Grandmasters play protoss, compared to around 35% of all players. They remain slightly underrepresented amongst Grandmasters. (It should also be remembered that random is dramatically underrepresented in GMs League, and as a result (statistically at least) Zerg and Terran are both significantly overrepresented. Zerg is the most overrepresented. There is a similar trend in SEA: see my earlier thread at http://www.sc2sea.com/archive/index.php/t-1242.html).

There continues to be very few Protoss players in the top 10 of any of the regions. Based on my search this morning, of the top 10 players in each of the major regions, NA has 1 Protoss, EU has 0 Protoss, Korea has 2 Protoss and SEA has 1 Protoss. By contrast:

- 5 of the NA top 10 are Terran and 4 are Zerg;
- 6 of the EU top 10 are Terran and 4 are Zerg;
- 7 of the Korean top 10 are Terran and 1 is Zerg; and
- 3 of the SEA top 10 are Terran and 6 are Zerg.

Discussion

These results hardly provide a definitive answer to the question I have asked. But, I think, the data legitimately entitles me to ask the question: is Protoss currently underpowered in high level play?

I am very curious to see what Blizzard's overall data shows, and really wish they would release this (as they have done in the past). The most problematic scenario, I think, is one in which the data shows Protoss is overperforming in lower level play (whether that be overperformance in bronze, silver, gold, platinum, or overperformance all the way up to high Masters). At the moment I don't have any data and could only speculate on the position below Grandmasters. If this were the case, would raise the question, as many have already suggested, of who it is that Blizzard should be balancing the game for - professionals or the average player? Personally, I tend to think it should be balanced for high level play and everyone else should try and catch up by looking at what the professionals do. However, at the moment we do not have data to suggest any underperformance by Protoss below the parameters I have identified.

On a sad note, it will be difficult for me to pick up much to "imitate" from the GSL super tournament. Unfortunately, most of the Protoss games have been terribly one-sided and not really worth watching (for my part, it is starting to look brutal and somewhat bleak out there for the pros). However, I remain optimistic that one of the remaining Protoss players will show us something new and spectactular to stop the rot. My hopes are now pinned on SlayerS_Alicia (who, for those of you who don't know, more or less gave us the aggressive 3gate expand which revolutionised PvT a few months ago).

Tom please add at least one picture to your articles so it has a thumbnail!

Image Credit: http://sanggene.blogspot.com/

Quick Comments

	nirvAnA:
	Apth:
	EveRorschach:
	TCPLemminks:
	EveJeonSa:
	TAScarecrow:
	duanloves:
	Digby:

[AlphaWhale]

Originally Posted by dennistoo

Well, i guess cheesing is the only way to win now.

1 base DT rush ftw.

Yeah I hear you can win a GSL with that. Oh wait...

Personally I'd like to see the new protoss meta trend towards templar/mass warpgate play. (When you see protoss warp in 21 units at once it makes hatcheries look weak). Terrans are pretty much making blind vikings in PvT and zergs are learning how to use infestors.

But robo gives you a lot of mid game security so how that's going to come around I do not know.

[Benji]

My only issue with your post is that you have such a small sample size with your data. In the TL thread the Korean graph even has a note saying that it is a small data size. For reference, the PvZ graph on the Korean graph has a sample of 269 games over 6 months. That means theres around 45 games a month, meaning that the 30%/70% difference, which makes P seem amazingly weak is really misleading. If Protoss won 10 more games that month, it would be 50%/50%.

The other graph showing global trends has 10x the data, and the win% is within 1% which is acceptable in my opinion for game balance.

What about protoss do you think needs to be changed?

[Tom]

Both your points about the data are valid Benji, but I think you have not allowed for some of the other data I have presented.

Korean GSL data

The fact that the Korean April/May GSL sample size for was small and limited to the results of one tounrament is the reason why I wanted to wait and see what happened in the next GSL. If the April/May GSL results were an aberation, I would have expected to see some reversion to the mean (that is, I would have expected Protoss to "bounce back" to some degree in the GSL Super Tournament). However, it is already clear the opposite has occurred - Protoss has performed even worse. That is why I am concerned the results may be part of an overall trend rather than a "once off".

International tournament data and other international data

I have also been careful to look at a range of data sources, not just the results of GSL in Korea. These include various Grandmaster Leagues and International Tournaments.

As you say, the analysis of TLPD data on Team Liquid for international tournaments showed results approaching 50%. However, this analysis is almost 2 months old, and it is necessary to look at more recent results. It has been my suspicion that Korean results are generally a "leading indicator" in this area.

Sure enough, if you examine the results of NASL and MLG Columbus (see my earlier post/analysis in this thread), the results are now consistent with the possible trend we are seeing in GSL and in various Grandmaster Leagues. While there are a lot of Protoss players starting out, a disproportionately low number are qualifiying for higher rounds. The proportion qualifying for MLG nearly halved (in comparison, the proportion of Terran and Zerg qualifiers both increased by around 25%).

Looking at a variety of data sources

I agree that we cannot safely reach a conclusion the basis of one data set for one point in time. However, I think we can conclude that:

1. A variety of data sources now show the same apparent trend in poor Protoss results; and

2. The most recent data shows the trend is apparently accelerating.

My point is, if a whole lot of different data sources are starting to say the same thing, it is time to ask the question.

What do I think needs to be changed about Protoss?

I am flattered anyone would actually ask for my opinion on this question. Unfortunately, I feel I am unworthy to speak with much (if any) authority on the subject. I also feel that if I were to start talking about changes I think should be made, it would detract from the objectivity with which I am attempting to approach the data.

There are really two steps in balance:

1. See if there is a problem (look at the data)
2. If a problem is identified, try to find ways to correct it.

I am only really dealing with step 1. Step 2 is really hard, given the potential for an apparently small tweak to have a large and/or unexpected effect on balance.

[nGenLight]

I don't really think any units need changing as opposed to us Protoss just figuring out how to play. However, if there was something I could change, I would rework Ventral Sacs - Maybe actually showing if there is units in there. Floating an overlord just a little away from each Protoss mineral line just completely cripples our economy.

[crAzerk]

If the Ventral Sacs change was done, then wouldn't Medivacs have to be tweaked as well for fairness?

I understand how it's a complete mindf--- for Protoss to have something like that being a burden on your mind, but to express what I feel about this would warrant another thread.

[nGenLight]

There are some pretty significant differences between Medivacs and Ventral Sac'd overlords that makes me feel okay with the former but not the latter.

But as you said, let's not get into it because it was not my intention to discuss the balance of ventral Sac, nor get into a debate with yourself.

[Benji]

Light, Put a cannon down, it'll deter fake overlords as they will lose them, meaning they really don't gain much out of you losing 5 seconds of mining time (<100 minerals I would say), and also puts a timer on the baneling filled ones, as they generally can't keep following your probe line for long before they are killed.

Tom, Blizzard said in an interview during the HOTS reveal that they feel race balance is fairly balanced, and future changes will be regarding specific builds they feel are too strong (Can't remember which video it was). Which is why I ask where the issue is, if theres something Protoss have a glaring weakness for with regards to builds, it would be what needs to be addressed. I feel blizzard won't make any blacket changes for any of the races anymore, and looking at win %'s doesn't take any of this into account.

[Tom]

Benji could you please provide a source as I would be interested to read exactly what Blizzard said.

To say "race balance is fairly balanced" sort of begs a whole lot of further questions - do they mean across all skill levels etc (of course I am saying this in the absence of details of precisely what Blizzard said). Also, just because Blizzard say something doesn't necessarily make it true - I would assume they would relying on some sort of data of their own, which I hope they will make available.

[nGenLight]

It's not just mineral line baneling bombs. Ventral sac is so good against Protoss in so many ways, I really can't be bothered to list them all.

[Monk]

@My last post giving MLG Colombus Mirrors, I have to correct myself. I only counted half of the mirrors instead of the full amount of mirrors and made an error in one of the countings. I can't get the numbers to match so let's ditch that last working.

@Benji I honestly can only think of 1 build that should be nerfed. And it's the MC 4-gate. It's hard to scout, and if you do scout it, it's generally way too late.

All-ins and any other form of dominant builds(2 rax, 4 gate, 7 roach rush) are easily scouted, and easily defended if you know what you're doing.

@Tom
Korean Protoss in GSTL(http://wiki.teamliquid.net/starcraft...r_Tournament):
P__T__Z

Ro64
16_29_19
Ro32
7__17_8
Ro16
1__7__6

GSL - May
http://wiki.teamliquid.net/starcraft..._II_League_May

Look at this Tom. You can't deny that protoss aren't doing well. Using the above pattern.
Code A - Ro32
13_13_7
Code A - Ro16
3__9__4
Code A - Ro8
0__6__2

Terran dominating GSL Code A Zerg ain't even doing well. Let's see Code S. Using the same pattern (PTZ)

Code S - Ro32
10_14_8
Code S - Ro16
5__7__4
Code S - Ro8
3__3__2
Code S - Semi
1__2__1
Code S - Finals
1__0__1

How can protoss be underpowered if they are the highest retainers of Code S? 10 protoss at ro64, 50% retained to ro16 then 60% retained at ro8. Korean Protoss look at this thread and laugh, with that being said Code A still needs some work.

I just don't see why you're saying Korean Protoss don't know how to win with Protoss, where clearly they do.

[Benji]

Tom, I can't remember which interview it was, I will spend some time trying to find it, but they implied at the top level of play. They won't release numbers, they don't need to prove anything to us.

Light, Apart from baneling drops, what else do you need to see? Roach drops or Hydra drops and such you can scout, or react to. You need to be more specific on what you think is broken with regards to drops.

[nGenLight]

Because units that have no food cost, very low cost creates a threat that requires about half your army to deal with. While the Zerg player subsequently throws cheap Roaches at your third and natural at the same time whilst they have baited your army to your main, goodluck getting that deathball up. With super high levels of insight, unit management and control, it is possible to defend well enough to get your deathball up before Zerg has a deathball busting unit composition, but this is ridiculously hard, especially on a map like XNC.

Medivacs can be dealt by mineral dumping on some zealots or making 4 stalkers to prevent a drop. Zealots against Zerg units? 4 Stalkers to target fire 20 overlords heading your way (with no idea if theres any units in there)?

I'm not saying Ventral Sac is OP or anything, but it is probably the one single feature of Zerg that drives me insane because it is ridiculously hard to really gameplan for it. I'm sure some Zerg players may have the same opinion about Blink or FF.

[Monk]

Originally Posted by nGenLight

Because units that have no food cost, very low cost creates a threat that requires about half your army to deal with. When the Zerg player subsequently throws cheap Roaches at your third and natural at same time whilst baited your army to your main. Goodluck getting that deathball up. With super high levels of insight, unit management and control, it is possible to defend well enough to get your deathball up before Zerg has a deathball busting unit composition.

Medivacs can be dealt by mineral dumping on some zealots or making 4 stalkers to prevent a drop. Zealots against Zerg units? 4 Stalkers to target fire 20 overlords heading your way (with no idea if theres any units in there)?

I would be looking at your macro if you're working towards a deathball and only have 4 stalkers when your opponent has 160 supply worth of overlords, unless you mean what you meant with the medivac.

[nGenLight]

^ Dude WTF? I was comparing about dealing with medivacs to dealing with overlords. That was my exact point, you have 20 overlords flying to your base, you need like 40 stalkers to defend that, and what if it was just a diversion? I just placed 40 Stalkers completely out of position to watch my expos get destroyed.

Maybe you could teach me some silver zerg macro.

[Tom]

@Monk: Um, I'm not sure what point you're trying to make with your example. In one of your examples there were 1 Protoss, 7 Terran and 6 Zerg by the Ro16 and in another example there were 0 Protoss, 6 Terran and 2 Zerg in the Ro8. The third example shows the races "breaking even", with each race halving in number each round, until the Protoss/Zerg final.

In any event, arguing about the data at this level of specificity is a waste of time. The data you have referred to is included in the Korean Tournament Results data for May (which is accounted for in the Team Liquid graph). What you appear to have have done is revisited the data used for this graph, selected a segment that apparently suits your argument (or at least doesn't look as bad as the rest of the data), and then used this to assert there is no trend.

With the greatest of respect to you Monk, I am not going to debate the figures with you anymore. It is really not worth my time pointing out gross errors of arithmetic (eg overcounting mirror match ups) or trying to explain basic statistical principles such as "don't pick a small part of a data set that suits your argument and then ignore the rest of it". I also suggest it may be wise not to lecture nGenLight (possibly the best Protoss player in Australia) on his macro ;p

[Monk]

Originally Posted by nGenLight

^ Dude WTF? I was comparing about dealing with medivacs to dealing with overlords. That was my exact point, you have 20 overlords flying to your base, you need like 40 stalkers to defend that, and what if it was just a diversion? I just placed 40 Stalkers completely out of position to watch my expos get destroyed.

Maybe you could teach me some silver zerg macro.

Everyone has to start from somewhere. And you should have game sense and scouting. There is a reason why Blizzard gave you observers and gave some maps xel'naga watchtowers. You see many units under mass overlords? Base defense, set your observer to follow which base has the most overlords heading. You see a ton of units and overlords heading to you? Could be a diversion or a small drop. I pull overlords to take the first few hits from stalkers giving my roaches and hydras time to get into position.

With the same thing being said about overlords and medivacs, a diversion medivac or a real drop, how do you know? You don't. That's where your scouting comes in. I'm pretty sure there is this thing called static defense. As Benji said, cannons work well. I find it's near impossible to micro a just dropped baneling from splashing on a cannon if it's placed in the mineral line. A simple probe pull and let the cannon take the hits.

You just sound like a complete whiner. Benji gave you a tip on how to stop it. You then come around saying ventral sacs is good overall? You have to spend 200/200 to get ventral sacs, and it takes forever for that to upgrade, and you have to upgrade overlord speed too. If it's against your 'death'ball then you have your 40 stalkers to target fire them down. Blink is pretty much a get upgrade in PvZ anyways, so just box blink if your micro isn't good. You're just giving excuses and looking for the easy way out.

Let's nerf Warp prisms so they can't warp in your base. Let's nerf Medivacs so they tell you what is in it and make it so it can't heal at all. Let's show what is inside an overlord too, and make it so they have no speed upgrade. Ohh wait, if we're doing supply giving units/structures let's do this. Pylons can only power things touching them. Supply depots cannot be lowered and have lower HP, Overlords cannot drop units and move slow. Man I should be in charge of SC2 balancing. I'd make you so happy.

I may just be some silver zerg(btw I'm not really silver any more, but let's roll with it) But I watch Tournaments, Replay and streams of high-level players to know some type of high-level game and ways to stop certain things, and ways to do certain things.

In the end, you've just gotta have eyes on your opponent. You should see a drop coming. Stop it. You see multiple drops coming? You should have cannons.

Instead of mineral dumping zealots why don't you mineral dump with cannons, giving you more supply and you don't essentially waste warpgate cooldowns.

TL;DR:
Get better.

Originally Posted by Tom

@Monk: Um, I'm not sure what point you're trying to make with your example. In one of your examples there were 1 Protoss, 7 Terran and 6 Zerg by the Ro16 and in another example there were 0 Protoss, 6 Terran and 2 Zerg in the Ro8. The third example shows the races "breaking even", with each race halving in number each round, until the Protoss/Zerg final.

In any event, arguing about the data at this level of specificity is a waste of time. The data you have referred to is included in the Korean Tournament Results data for May (which is accounted for in the Team Liquid graph). What you appear to have have done is revisited the data used for this graph, selected a segment that apparently suits your argument (or at least doesn't look as bad as the rest of the data), and then used this to assert there is no trend.

With the greatest of respect to you Monk, I am not going to debate the figures with you anymore. It is really not worth my time pointing out gross errors of arithmetic (eg overcounting mirror match ups) or trying to explain basic statistical principles such as "don't pick a small part of a data set that suits your argument and then ignore the rest of it". I also suggest it may be wise not to lecture nGenLight (possibly the best Protoss player in Australia) on his macro ;p

I don't care who's who. If they're whining about a thing they are obviously asking for help in dealing with it.

And in regards to the data, you said that everything I did before was around april, and you were meaning now so I pulled the most recent GSL results out. And I was meant to say you can't deny protoss aren't doing well* in regards to Code A. But in Code S the protoss were doing pretty well. I was having fun throwing data at you and receiving data back (:

So on topic: Is Protoss considered Underpowered in high-level play still?

Quick Comments

FaDeBadger:

You are lecturing Light on how to play the game when unless you are one of the best zergs on the server, you shouldn't be. Not to mention you are wrong on almost every point

[Monk]

double post on accident D:

[nGenLight]

I was offering my opinion because it was asked of me to give one. I was just telling my own personal experience of Ventral Sacs giving me fits, in no where was I calling anything OP. But all I get from you is insults and you talking out of your ass about a level of play you wouldn't have a clue about.

1. You have no clue what a multi-prong attack is do you? I can stop 20 overlords if it was the only thing coming at me.

2. Do you honestly think 20 medivacs flying towards your base is the same as 20 overlords flying towards your base? Maybe you want to start thinking about resources like food/building time/availability/opportunity cost.

3. Having the ability to scout = map control. Once you take that third base you give away that map control. Regardless, I am still able to scout through hallucinated phoenix, well placed units and usually can see things coming at me before they do.

4. You need to read, I didn't say nerf Ventral Sacs, I was asked a question, my answer was that If I HAD to nerf something that would be it because Ventral Sacs is what gives me the most trouble.

5. Yes I need to get better, it's just a little weird being told so by somewhat who is so ridiculously ignorant.

[nirvAnA]

an interesting observation of the reputation in this "protoss underpowered" topic



well except for apth

[Tom]

The fact that a non-Protoss player approves of this thread leads me to doubt that there is trend in Protoss players approving of this thread!