No. You never once could come up with one. You came up with pro players crying about imba. I want to hear a pro say “Terran is easier than Zerg.” Or are you not smart enough to understand that?
BNet never had a vote. You made an obvious troll post, so nobody upvoted it.
I wonder if my Terran bias would ever let me post something this low IQ. Probably not.
You keep using averages on the pro scene as if they mean something when they don’t.
Terran also has the lowest representation on the Pro scene by a pretty significant amount, which makes that a questionable argument at best, considering you limited the sample to players with 100+ games. You’d expect the TvZ win rate to be ~55.8% if Terran actually had a 26% higher performance than Zerg in a player pool where the average skill of players is equal, but they don’t. In fact, the TvZ win rate has been below 50% for most of the year, which would indicate that Terran having higher average ELO vs Zerg on that graph is likely a result of the players that fit the metric you’re using to qualify for that graph actually being better players on average, which would make sense given that there are significantly less of them.
Given recent tournament performances, and all the other data available, there’s little reason to think that a Terran is favored vs a Zerg of equal skill. If you look at the top 10 players on aligulac for example, the Zerg players have a slightly higher average rating in the match up than the Terran players, and there are more of them.
If you selected only those players on your graph and matched them against each other at random, you might get something like that ~55.8% win rate over enough games; too bad the data from the graph doesn’t mean what you seem to think it does.
There are more stats that would indicate Terran as the most difficult race to pick up and learn, and Zerg the easiest, than the other way around. I won’t make any definitive claims, as there’s too much going on for something like that to be applicable to anyone, but… using a 22 person vote as “proof” of anything is pretty silly.
I changed the requirement to having at least one game played and the trend stayed the same. Requiring at least 100 games played is very important, because it filters out noise. A players ELO vs Zerg can be predicted with ~80% accuracy from their vs Protoss and vs Terran ELOs, with the exception of people with few games played:
This is PERCENTAGE of error. The trend that applies to the rest of the data is so wrong on these people it is estimating their ELO as being 5000% larger or smaller than it is. There is a very good reason why I exclude people who have played very few games.
No, you wouldn’t. For example, let’s imagine Zerg is favored by 10%. Now lets imagine Innovation is 15% better than Zerg and he goes into a tournament packed with Zergs and he is the only terran. TvZ win-rates are going to favor Terran, even though Zerg is favored by 10%. You MUST control for the skill variable and win-rates don’t do that.
For the same reason you can’t look at a win-rate and tell if the result is caused by balance or skill, you can’t take a result caused by imbalance and project it into a win-rate. Imbalance is not the only factor that affects win-rates. If you want to project this into a win-rate, you need to account for all the other factors. If you did, you’d see an identical result to the data on Aligulac because this data is compiled from the same data that win-rates have. That means you can transform this data to that data and visa versa. If you don’t, it means you messed up the process somehow.
I did this at one point in time, just not between average ELO and general win-rates, but rather odds of winning tournaments. Taking average TvZ ELOS and a mock round of 32 tournament, I showed Zergs would win tournaments 5.5x less often than terrans. Guess what the actual data showed? Terrans win tournaments 5.5x more often than zergs. I can link you to the math if you’d like.
The long and short if it is that your predictions are wrong because your math and logic are bad, not because the data is unsound.
You are going off of Aligulac? Here’s the thing about that. It splits win-rates in periods. Some periods have more games than other periods. So you can have, for example, period 0 have 1000 TvZ games with a 55% win-rate for Terran, and then the next period be 200 games with ZvT being 70%. Now, if you average that you get a 58% win-rate for Zerg even though that’s not what the data says. The data says a 49% win-rate for Zerg. You have to add up games won/lost for each period individually, then calculate win-rates. You cannot average the win-rates.
Right, so, there is a group of people in Aligulac who perform 26% worse than another group, and it’s pure coincidence that the people who perform worse are the people who play zerg? That is a possibility, but it is a one in a million possibility. That is HARDLY a “little” reason - to the contrary, it is a very compelling reason. Either Zerg is underpowered, or a one in a million scenario has occurred. By nature of common sense we must reject the hypothesis that a one in a million scenario has occurred and thus zerg is underpowered.
I think you missed my point entirely. I know why it’s necessary for your graph to be able to measure anything.
My point is that there being more Zerg players will inevitably make their performance look worse than it actually is when using averages to measure the graph (for obvious reasons if you understand the concept of normal distribution and where a pro scene is with regards to it).
Yes, which is why I pointed out the average for the entire year, and not just the last 2 weeks (as each period is 2 weeks).
Aligulac also rates Zerg as the race with the best overall performance, so your “performance” measurement is probably missing something critical.
Representation plays a huge factor in averages. All your data really does is support the idea that there are more Zerg players of lower skill on the pro scene when taken into context with all the other data…
The average being relevant to a discussion of balance is dependent on the actual skill of the players (not their ELO), having the same average, independent of race. We do not know that that is the case, and there is data that suggests otherwise.
Just flag troll posts.
Just flag troll posts.
Goba would have so many he could call himself the UN.
This chart disproved that thoroughly. That’s in fact the exact reason I created this chart. Your theory is that when having a larger sample of zergs increases the number of low performers in the sample because low performers are more common than high performers. This theory is wrong and has been thoroughly debunked in many ways:
You don’t understand how their performance metric works. All it is is a net elo gain per-race for a period of time. All it means is zerg’s performance is going UP which we would expect with the recent buffs.
You have absolutely no clue what you’re talking about. This data says literally the exact opposite. It disproves your theory very thoroughly.
lets take forum oppinions over pro players and top gm random players, obviously will be more accurate.
Watch avilo’s latest vod (starts around 53:19) and let us know how “easy” terran has it.
Also watch avilo vs really good player like one of the many games against beastyqt or even Pig.
Avilos style eventually just stops working because its not a good style.
You could prob get to gm with any race playing low apm campy styles however i agree mech terran tends to favor that the most (less now as theres no longer any raven god unit).
Also watching avilo play vs watching anyone play bio is like night and day so i guess if you wanna get technical you have to look at three terran styles:
Bio, Mech (actual mech with harassment and timings) and then theres avilo style.
This is functionally no different from sampling the entire pro scene, so of course it does. The pro scene is itself, not a random pool of players. In order to play on the pro scene, a player needs to have a performance above a certain level.
A higher rate of occurrence of one race is an indication itself that that race performs better relative to skill on average.
Only if it is the case that the skill of players is actually equal, and all the other match ups are equal. If Zerg has an increased overall performance, resulting in a larger number of players, then the difference in skill between actual Zerg players at the bottom and players at the top will be larger, which will result in the appearance of worse performance by Zergs when measured by something like elo.
Ex: PvZ is bad for Protoss, ZvT is close to balanced, and TvP is bad for Terran.
TvP brings down the elo for terran, while bringing up the elo for Protoss, which, in turn, brings up the elo for Zerg due to their strong match up into protoss, making TvZ look Terran favored when measured by elo, despite being balanced.
Even if the opponent’s average elo is lower, beating them will raise the elo for the winner of the match. By playing against players of lower elo at a higher frequency, a players elo will naturally end up appearing higher in the match up if they have a ~ 50% win rate vs a Zerg of ~ equal skill. Because more Zerg are in the pool, and, per normal distribution, more of them will be lower in skill (and thus have worse performance), the elo of players that are able to beat those Zergs will end up higher if the game balance is reasonably good.
Measuring balance based on performance and claiming any single measurement as proof of something is a fallacy - balance is a factor in player performance. It can, at best, give a very loose idea of the state of things when put into the proper context. A single measure cannot “prove” anything on its own, because such a “proof” is reliant on circular reasoning.
You’re missing my entire point.
The pro scene is not an entire normal distribution curve. It is the top end of a normal distribution curve. I don’t see why that’s so difficult for you to understand.
The performance of the population is weighted towards the bottom, which will pull down the average for the race with the highest representation more than the other races over time (as player skill increases, the differences in skill between players also increase, compensating more for the differences in balance) unless it is proven that all 3 races have the same average skill within the sample (Which is effectively impossible with the data available), then it does not actually prove anything.
Any sampling that includes players on the pro scene, randomly determined or otherwise, will run into this as an issue. It’s been a problem for every single one of your “proof” threads. A player’s position on ladder or the pro scene, their Elo/mmr/rating, or them being on the pro scene at all, are all measurements of performance, not skill.
Skill level isn’t a metric you’ve used to measure anything. Performance level (which can be measured by Elo/rating/mmr/whatever) and skill level (which would rely on a measurement of performance in a state of perfect balance) are two different things.
Skill is not a variable that we have enough data to actually measure in a meaningful way, so we do not know the actual skill level of players; we only know their relative performance, and can only make somewhat definitive conclusions of skill relative to players of the same race, as it is assumed balance is not perfect, and balance is a factor in performance that we don’t know the value or consistency of.
No, you don’t. Anyone can sign up for open enrollment tournaments. Anyone can submit “pro” games to Aligulac. Aligulac has 18k players in its database. Are you really going to argue that there are 18,000 pro players?
Correct! And you’d see that skew on the chart. There is none on the chart. Terrans get a flat buff vs zerg and this isn’t dependent on skill level in any way.
*normally distributed. Not equal - normally distributed.
Yes, but it is all proportional. If you have two people, one being 500 elo in skill and one being 300 elo in skill, and they play lots of games vs each other, their elo rankings will converge on 500 and 300. The same is true for a person of 500 elo and -500 elo. If they play enough games, their elo will converge on 500 and -500. That is a property of the algorithm. You cannot change the average skill level of an opponent and net an increase in your own elo. You will be more likely to beat these people, and will have a higher winrate, but your elo ranking will stay the same.
Nope. The point of a statistical test is to test whether or not the results correspond with a variable and not others. When you prove mathematically that there is a 99.999999% chance that these results are caused by balance, and not the other factors, you’ve proved imbalance.
A) yes it is unless you want to argue that there are 18k pro players.
B) the linearity of the chart and the random sampling produce the same results thus proving it is normally distributed.
C) Even though it IS normally distributed, I entertain your ideas and show that even then they cannot explain these results.
These results correspond with race with a 99.999999% confidence, but you say we haven’t accounted for skill? You’re nuts. The chart shows zero relationship between skill and being buffed vs Zerg. The only way you can make these is claims is to deny evidence. The only variable that controls being buffed vs Zerg is the race variable and that correspondence is so strong it is undeniable.
If your argument is based on fact denial, which it clearly is, then it is trash. That puts you in the same realm as flat-earthers. You put on an elaborate show of words but when you strip away all the frivolous material your entire argument is simply fact denial.
kelthard is a plantinam protons 4 long time and anastasia is in mustard in wing of liberny
I just wanna know what damaged this guys ego so bad that he has been constantly coming to the forums for YEARS now to complain about Terran. Like even if he’s right, with the amount of time he’s spent on this…what does he get out of it? I really don’t get it. I guess people’s obsessions are hard to understand. It’s a strange world out there.
I am a software engineer who is developing new statistical analysis software and tests it out on all sorts of data sets. Definitely have a gargantuan ego, though, you’ve got that much right!
Another rabiDrone thread. Flagging as trolling and moving on.
Ah, yes, organizing brigades of posters to report posts to remove content that doesn’t violate the rules - that sure will tell blizzard that the guy you hate is the one abusing the forum system! Oh… wait…
(Insert BIG BRAIN TIME meme here)
OMG, this post of BatZ is priceless. So much water has passed under that bridge. Today BatZ sings a very different song.
It reminds us that a long time ago BatZ was sane and rational. Bad company caused BatZ to loose the way…
I always love how he suddenly became insane and irrational the instant he decided to use almost the exact same metrics to instead define Protoss as the easy race.
Spare us the “indignation”. I would take you more seriously the day you would reject BatZ Anti-Protoss diatribes (that yourself call them the same as antiterran ones) with the same force that you did when your race was attacked.
I never hided my position as protoss-apologist, much different that your hypocritical mask.
Even a broken cloak shows the correct time twice a day, BatZ is correct in this one.