Okay, I’ve seen more people talking about batz responding to himself than supposed batz accounts talking to themselves. Please, either give evidence of this or stop the schizo posting. I’ll catch schizophrenia from you guys if you keep this up.
I have seen it a couple of times, it would be really hard to find considering the insane amount of posts he makes. Liking his own posts it is easier,you can look for threads he made,go to the likes section and many times you will see some of his accounts giving a morale boost. And that’s the things we know from his known accounts,who knows if he has other accounts,I wouldn’t be surprised.
I’ll buy the likes, that’s not absolutely insane. But talking to himself is really reaching into mentally ill territory.
Yep,when I saw it I couldn’t believe it all all, it was one of those accounts he made and didnt use for a long time,maybe banned or forgotten. Do not want to be an armchair psychologist,but it seems like he really craves attention and will do anything to keep the threads going.
The liked posts,just the first I searched has a couple of own likes and one from a new account that popped recently “flatulentdog” that is liking batz posts,who knows if its himself,but that’s not a proof,these days there is too many smurf accounts giving likes.
1 Like
I would want to say yes but it would not be a challenge. so they are allowed only if the action is randomly selected.
So you will have a software that will be generating random actions base on the previously generated actions and you can only perform this action. this will exclude your game knowledge but allow you to use APM.
Yeah, so there’s obviously proof he’s liking his own posts. Again, that’s far more sane than talking to yourself.
Wrong. Here is a good example. AIs have absolutely no clue how the game works, but they can punch in commands to win even at the pro level and they even do it better than professionals do. You are trying to equate inputting commands into the game as a form of “game knowledge” so are you prepared to claim AI’s understand the game and know how it works?
Or maybe you are willing to accept you can play this game by inputting arbitrary commands, and as long as you get these commands correct you win, even if you don’t understand what those commands are nor why they work. Are you willing to accept that this is an entirely valid way to play the game and not only that is the primary mode of playing the game for the vast majority of players?
SC2 is a game of multitasking and split second reactions. Decisions are made in a tenth of a second. To suggest that higher-order thought processes like decision making, strategy, analysis, play a role in these time-frames is pure lunacy. The only thing that matters is how fast you can do a list of memorized actions. That is 95% of the game. The problem is that Protoss players can do this substantially slower than T/Z and still win. So the primary facet of the game is easier for Protoss. That is the ONLY valid conclusion from the fact that Protoss have lower skill metrics per league.
AIs have a very good idea of how this game works. What they dont understand is how PEOPLE work. You stick an AI in a combat, and it can fight its way through better than any real person, because they understand things like perfectly timing blinks to dodge attacks, and have infinite APM to act on that understanding. And then it will send the entire army to attack a mined out base, because it sees a building and thinks its important to destroy it.
Starcraft 2 AI is not true AI but a bot that has a recorded list of commands to be performed. That way some units from LotV are not built by Starcraft AI.
True AI like deep mind AI had to be taught how to play the game by providing replays. At the same time when put in the ladder, it was adjusting the build if the new build was better than currently used.
So Deep mind AI had a rudimentary knowledge of the game.
Game knowledge is to know the effect of commands, and being able to select appropriate input commands to archive expected results (you need to know how blink works and when to used to save stalker from being killed).
If I would say that you aren’t allowed to use any starcraft 2 game knowledge, then you would not be able to generate APM, but it wouldn’t be a challenge.
So the challenge needed to be modified to allow you at least a basic understanding of the game, but not allow you to select commands as you please.
so yes, with this challenge you could win a game if you get lucky. but it’s not the primary mode of playing the game for the vast majority of players?
The vast majority of players try to select commands base on their game knowledge.
you contradict yourself. if you making a decision you doing “decision making” those you have higher thought processes.
But to be more precise:
1.) When you start the game you have a build order in mind - this is called planning.
2.) When playing the game you are collecting information regarding your opponent - this is called scouting
3.) You are processing collected information - analyzing
4.) you adjust the plan base on the information you collected and analyzed. - decision making.
for example:
the first scout using worker is used to adust your build order.
I would not deny that Protoss can’t archive the same avrage APM as T/Z but my question to you is why?
For me and most protoss players, the main reason for this is protoss builds mostly consist of units that have limited macro potential (Archons, Immortals, etc) those the value of performing the same action like split as P is less rewording than for T/Z.
for example:
For MMM it’s the much easier to gain value from performing a split vs bains than for Protoss due size and speed of the units.
OK. Go ask the AI what build is the best and ask it to explain why it is the best. Good luck!
I think it learned builds from others,so it wold sya some build is the best after considering the popularity of builds at ladder. At least I temember some comentary about devs saying to not cheese the AI to not create trash data,they wanted the AI learn and adapt from usual gameplay,not cheesers, abusers and random stuff.
1 Like
This is not how the meta works. The most common strategies are the ones you have the most data on. They are the ones you must exploit if you want the highest win-rates, which is what you need if you want to get to GM. So the AI by definition would have to learn builds that it doesn’t have data on in order to get to GM.
The AI does not understand what it is doing. It is punching a set of inputs into a black-box for the desired result of a high win-rate. That’s it. SC2 is a black box as far as it is concerned (as far as humans are concerned too).
In science, computing, and engineering, a black box is a system which can be viewed in terms of its inputs and outputs (or transfer characteristics), without any knowledge of its internal workings. Its implementation is “opaque” (black). The term can be used to refer to many inner workings, such as the ones of a transistor, an engine, an algorithm, the human brain, or an institution or government.
The AI knows that if it puts these inputs into the box, it spits out a higher probability of a “win” than other sets of inputs. That’s the AI’s entire “understanding” of SC2. The alphastar AI is complete, total, indisputable proof that you can play this game with absolutely no understanding of how it works. You can do simple build orders and have better micro and hit MMR levels that are humanly impossible by that merit alone. Why does this build work? What makes this micro strong? It has absolutely NO clue.
1 Like
AlphaStar uses blink yet if you asked AlphaStar why blink micro is good and how to do the best kind of blink micro, it couldn’t answer. It is entirely possible to play this game with no understanding of how it works, in fact the game is so complicated that even the most knowledgeable people only know 0.000001% of what there is to know. It is simply a fact that the primary mode of playing SC2, for the vast majority of players, is to punch in a series of commands that they’ve learned produces a win and that’s about it. SC2 is so complicated it is effectively a blackbox.
I did not contradict myself. I am not supposing that decision making is happening, I am PROVING that it CAN’T happen on the small time-scales that SC2 players are required to make decisions. SC2 gives you a tenth of a second. It is not possible to engage in analysis and decision making and strategizing on those time scales. SC2 is about short-circuiting your reflexes to react as fast as possible to an onslaught of inputs which it has to do so fast it is literally impossible to engage in higher order thought processes that are necessary to strategize.
Now if the game were focused around a 1-3 base economy, rather than a 5-7, and supplies ranging from 25-100, rather than 150-200, then you could make an argument that there is enough time between events that players can actually engage their critical thinking faculties. Right now, as the game is currently, it is designed to emphasize endurance & multitasking and that’s it.
And yet you didn’t provide any in-game proof to support your claim. Where I give you proof that you are wrong.
challenge for you:
14 0:18 Pylon
15 0:31 Pylon
17 0:50 Gateway
18 0:59 Assimilator
18 1:07 Assimilator
21 1:34 Assimilator
21 1:35 Zealot (Chrono Boost)
–Remove your knowledge about Sc2.
How do you know how to collect minerals?
How do you know how to build probes?
How do you know how to build Pylon?
How do you know how to build Gateway?
How do you know where to put Assimilater?
How do you know how to build Assimilater?
How do you know how to use Chrono
How do you know how to build Zealot?
How do you know how to select units/buildings?
How do you how cancel construction?
I can ask you multiple similar questions, but I hope you get the point.
I wouldn’t say we know 100% but for sure we know more than 0.000001% (assuming you not talking about how Sc2 was coded).
Regardless I give you the benefit of the doubt, so tell us what is in this 99.999999% of knowledge?
AlphaStar is can’t speak, so yes it would not answer your question but jokes aside.
If put AlphaStar in a blink micro scenario, Continues literation of this scenario would provide you information like % of killed stalkers, % of damaged stalkers, number of stalkers that are left.
Now for AI, those data indicate which literation of blink stalker micro is better and why.
AlphaStar also knows how the Blink is working because he needs to be able to understand output data.
Naturally under the assumption that Blizzard and Deep Mind didn’t cheat by allowing AlphaStar to scan the whole map.
1 Like
What on Earth are you smoking, kid? I’ve posted loads, and loads of ladder & replay data dumps. I stopped reading there. APM=Skill. This is a fact. Since Protoss have lower APM per-league, they are objectively, factually, lower skill despite achieving the same performance, period. If a race can achieve the same performance from a lower skill level, that race is overpowered, period.
I don’t smoke, but I would advice you to change your pils because I quoted your paragraph and yet you jumped to APM=Skill.
At the same time if APM=Skill you should be able to do this challenge:
Kid my points have already proven your points wrong. When you said:
I pointed to AlphaStar which has absolutely zero game knowledge (it just punches in a list of commands it knows produces a high win rate with zero knowledge of what those commands are or what they do) yet is able to beat even Serral.
I explained that this is the primary mode of playing SC2 for most players, since SC2 is so incredibly complicated that it is effectively a black box. I have effectively given you an introductory dissertation on AI/machine learning. DO NOT ACCUSE ME OF IGNORING YOUR POINTS WHEN I HAVEN’T ESPECIALLY AS YOU IGNORE MY POINTS IN THE PROCESS.
Machine learning maps inputs to outputs:
Supervised learning (SL) is the machine learning task of learning a function that maps an input to an output based on example input-output pairs.[1] It infers a function from labeled training data consisting of a set of training examples.[2] In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal). A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples. An optimal scenario will allow for the algorithm to correctly determine the class labels for unseen instances. This requires the learning algorithm to generalize from the training data to unseen situations in a “reasonable” way (see inductive bias). This statistical quality of an algorithm is measured through the so-called generalization error.[3]
Humans do the same thing, which is called “concept learning”:
The parallel task in human and animal psychology is often referred to as concept learning.
In a concept learning task, a human or machine learner is trained to classify objects by being shown a set of example objects along with their class labels. The learner simplifies what has been observed by condensing it in the form of an example. This simplified version of what has been learned is then applied to future examples. Concept learning may be simple or complex because learning takes place over many areas. When a concept is difficult, it is less likely that the learner will be able to simplify, and therefore will be less likely to learn. Colloquially, the task is known as learning from examples. Most theories of concept learning are based on the storage of exemplars and avoid summarization or overt abstraction of any kind
The brain:
- Identifies a situation as one of a number of pre-classified groups.
- Maps that group to a series of inputs that are fed back into the game using a keyboard/mouse.
Nowhere in here does this require:
- Understanding what the groups are.
- Understanding what the actions are.
- Understanding how the actions relate to the groups.
- Understanding how the actions impact future groupings.
It requires NONE of these things. It only requires:
- Classifying game outputs into groups.
- Mapping those groups to sets of actions fed into the game as inputs.
- Nothing else.
So if you are going to argue that being able to click a mouse is a form of “game knowledge” then you’ve conceded my point because you’ve excluded all modes of “knowledge” except the most simple kinds.
It is entirely possible to treat SC2 like a blackbox, playing a build you know works without any real or meaningful comprehension of how or why it works; in fact this is the primary mode of playing the game for the vast majority of players since the game is so complicated their attempts to analyze it are futile. They can list their reasons why the builds work the way they do, but the bottom line is they have scratched the 0.00000000001% surface of the strategy of the game and their predictions are therefore meaningless.
The strategical element of the game is too vast and complicated for any meaningful analysis, especially on the time-scales that decisions must be made in sc2 (<0.1 seconds), which is why it is so fruitful to focus on the simpler aspects of the game like staying on top of your macro, avoiding supply blocks, etc, all of which are a function of APM skill. Since it is much easier to differentiate yourself using APM, APM becomes the primary driving force behind the mmr distribution of the leagues.
1 Like
Arguing with you is like talking to a politician. So let me you speak BS by answering my own challenge:
How do you know how to collect minerals? - you don’t know how to collect minerals, because you do not know with unit collect them
How do you know how to build probes? - you don’t know how to build probes, because you do not know with building build them
ETC.
So no you need a rudimentary understanding of units, buildings, and SC2 mechanics to even start generating APM (Actions per Minute). So APM is not the only factor in SC2 those APM /= Skill
Starcraft 2 is BlackBox. AlphaStar and Players learn to play SC2 via mapping Input to output (Supervised learning). At the same time, we do not need to understand the interworking of SC2
So your point is meaningless that is why I didn’t respond to it.
To simplify: SC2 knowledge is understanding the output information and reacting to it with the appropriate input information.
example:
–if you receive an invalid build location (output) at a particular location you need to move construction (input) to a valid location.
–if you have the start of the game your output information workers and main building (output) then you need to react to it by providing appropriate input.
And I give you a challenge showing that without a rudimentary understanding of units/building alpha star would not be able to play the game.
So Does AlphaStar know the difference between worker and combat unit?
Yes, because we do not see AlphaStar selecting a combat unit and trying to build a building.
you are the one who claims APM=Skill not me.
I am a proponent of APM /= Skill because it is closest to the truth than APM=Skill.
a.) you need to understand the difference between workers, combat units, buildings,
b.) you need to understand where you can and can not build/move
c.) you need to understand how to react to your opponent’s actions. (Yes, this can be simplified to learning how to switch from build order A to build order B)
ETC.
So we go from this:
to this.
So I consider it my win 
2 Likes
Already covered this one. Please read:
I didn’t read the rest of your post. It’s clear you aren’t reading mine.
1 Like
I Quote the exact paragraph, I respond and I write answers to everything that has sens in this paragraph.
For example, I do not respond to paragraphs like the paragraph you quoted because they have 0 sens
How do I exclude any modes of knowledge?
I never say there don’t exist other modes of knowledge in SC2. I only say, that base on the level you are, the importance of those modes is different.
For example, up to the master, APM is much less important than other modes of knowledge. OP even confirms it with his Chart showing that Avrage Player with 100 APM can get to masters.
In Master/GM the APM starts to be more important because at this level other forms of knowledge are equalized for playerbase.