Improving Mechanics - TL oldie but goodie


Disclaimer: I do not know much about neurology, and this text gives an extremely simplified explanation of how the brain functions. The explanation exists for a reason though; it explains certain observed learning phenomenons very well. Please understand that there is plenty of room for improving these methods and this theory, as is the case with most complex topics.

Who am I?: I’v spent a lot of focus on trying to understand how I learn things in the last 4 years. When it comes to mechanical tasks in Starcraft I’v tried to build an understanding with experiences from both Starcraft and piano practice. After reading a book on optimizing memorization and doing memorization exercises I made a connection between all three fields. For the first time in my life, I was learning efficiently. I was always bad at memorizing things in school but recently learnt 500 Chinese characters in a month on my spare time. I previously failed utterly at learning “The Core hotkeys” after using them for more than a month, but a year I learnt it successfully for all races in a few weeks. I later reached 6k mmr on europe with random race macro play.

Introduction

During my coaching sessions I keep getting the question “How do I improve my mechanics?” and I always find myself struggling with giving a short answer since it’s a complex topic that personally took me years to figure out. I will in this text give my complete answer to the question so that I can refer my future Starcraft students to it, but also for you if you’re interested in my answer to the question.

This text will begin with a simplified background of how learning works. I will then write about how to learn mechanics which contains two big sections; one section explains a method that optimizes single mechanics efficiently and the second section describes how to make the most out of ladder games from a mechanical point of view.

Background

There is a catchy phrase coming from Hebbian theory; “Neurons that fire together wire together” which means that when we trigger an activity in the brain we can more easily trigger the same activity at a later time due to the new connections of the neurons. A Starcraft example of this is how when we want to build marines we need to actually perform several tasks:

  1. remember which production facility produces marines
  2. press the key that gives the barracks control group
  3. press “a” to build the marine

None of these steps are conscious to an experienced Starcraft player. The reason for this can be explained using “Neurons that fire together wire together”. I apply this analysis to the three steps above:

  1. Every time we see a barracks we also see the icon for the marine. We also have to recall this memory every single time we want to build a marine. In other words they fire together very frequently, and after they wire together the concepts “barracks” and “marine” are so strongly connected that we don’t need to think about it consciously anymore.
  2. This step is taken care of by most players by always binding the barracks to the same key. Every time we want to select the barracks we press that key. That’s the mutual fire. Eventually wanting to select the barracks and pressing the key becomes connected.
  3. We want a marine and then we press “a” many times every single game. This satisfies “fire together” so they quickly become connected.

These three steps are done over and over in the same order which creates a connection also between the steps, and together with the connections within each step we have an entire connected chain of actions that are triggered subconsciously by the thought “I want a marine”.

It’s possible to go even deeper in this analysis. Notice how “press a key” is not actually an action. What we do is move our hand (which can also be broken down into exact movements of many different muscles). If we have a “resting position” for our hand, a position to keep the hand in when nothing is being pressed, then this movement is identical every time. This strengthens the “fire together” compared to pressing the key from different initial hand positions. A resting position is implemented in the advanced hotkey layout “The Core”.

For efficient long term improving we want to find a way to create these connections with as little practice time as possible, as well as find a way to maintain the connections that we practiced for years to come. A method that has proved efficient for this is spaced repetition; it is better to learn for example glossary in short sessions every single day compared to doing a long session once a week. As a kid I incorrectly thought that the reason for this was purely that you should avoid long learning periods because you need breaks to rest (which are important as well). However, we can think in terms of the neurons again. A connection is made not only after a certain amount of fire. It takes time to build, and without a continuous fire over a longer period of time the brain will think that the connection is no longer important and the connection will be replaced by others. The longer a connection has been relevant to us the longer we can go without fire. In other words, we should keep our repetitions strategically spaced with increasing time intervals between them.

Depending on what we learn it can take different amounts of time to form a connection. For example, if you need to remember that your daughter has named a pet toad “Robert” you simply need create a connection, an association between two things you already have a good understanding of. This is a simple task that can be done instantly and therefore needs less repetition in the first few days. But even with this extremely simple task we still require spaced repetition over a longer period of time in order to remember it for a few years. If she stops using the toad’s name after a week you will have forgotten it after two months.

If however she calls the toad “Aregolethonith” you must first create an understanding of a word that you’ve never seen before which typically takes a 4-7 days with practice 3-5 times spread out daily before moving to the regular spaced repetition.

The Robert examples equivalent in starcraft can be to remember to produce an overseer to scout when the lair finishes, this is something you are familiar with and can do instantly, but it still requires spaced repetition to always remember it. An example of a task that cannot be done instantly in starcraft is often related to learning a new type of control, for example learning to control more spell casters or learning a new inject method.

I highly recommend trying out spaced repetition if you have not already if there is anything in life you want to remember (google it, try flashcards like anki for example). This will teach you about your brain and which length of time intervals are appropriate for you, something that will be useful in many parts of life. But that’s enough background for now, let’s start the discussion on mechanics!

Mechanics

Good mechanics in Starcraft is about being quick at going from wanting to perform an action to making the movement(s) necessary to perform the action. This can to be split up in two parts. The first part is having the brain translate the wanted action to what movement(s) needs to be made to perform the wanted action. The second part is performing the movement. We obviously have to figure out which movement to make before we can make the movement. This means that we can always separate the mechanical analysis into these two separate steps.

The first step takes place purely in the brain and the stronger the connections in the neurons the faster the translation. The more you repeat it the faster you will get. Period. An additional way of speeding it up is to simplify the task. Suppose you want to select your infestors, vipers, queens, ravagers, brood lords in quick succession in a battle. Having to think about which control group the units are in will slow you down significantly. A way to simplify this task is to always keep each unit in the same group so that before the game even starts you already have a strong connection between wanting to select a unit and pressing a certain key. A different commonly used system is to add the units in the order that they are built to the next available army key, which will optimize the time required to press keys since you more frequently use the control groups that are the easiest to press, in other words this will reduce the time of the second step. This will come at the cost of having to think about where unit is located. If you want to learn to control an army with many control groups I encourage you to position all control groups in a reachable position and prioritizing simplifying things for the brain by having a clear control group system. The time that you lose having to press keys slightly further away is much lower than the time you gain not having to think about it. This will also give you a clear mind which allows you to observe the engagement and make the correct micro decisions. After training a very simple system I have no problem controlling a 7 control group army with zerg, while i struggle with 3-4 with terran where I have not yet made a clear system. There are however different mechanical tasks where performing the movement is the dominant delay, for example when you produce a round of terran bio units.

The second step of performing the movement is less forgiving to practice, which I learnt from playing the same song over and over on the piano for over a year while knowing it by heart, but still missing the keys over and over. I noticed that I made some mistakes almost every single time I played the song; even in parts of the song that should have been easy to play. Other misses simply originated from a lacking fundamentals and were more random. The simple phrase “Neurons that fire together wire together” explains the recurring errors; because I had made the error many times I had made connections that made me likely to make those again. Spending all my focus on avoiding errors (and still failing) I was also unable to speed up my play without making more errors. Because of this effect it is important to start out very slow and slowly force yourself to play faster and faster while making sure that you don’t repeat your errors. Many people start out too quickly. This applies for learning new hotkeys, learning to control more control groups, camera locations and much, much more.

Learning mechanical tasks individually

Now let’s use this analysis on an example: getting faster at moving raxes onto reactors that have been built by the factory. The purpose of this example is to explain the method, not to get faster at this task specifically.

The first thing to do is figure out which steps you want to take. It can be hard to figure out the most efficient method of doing something. You can spend time analyzing it yourself, but you can also ask people and even watch replays and first person streams of pros to look at how they do things in detail. Make sure to briefly analyse the methods you find though; a lot false information is spread in this topic. People will say silly things like “amazing player does X and therefore X is the best thing to do”.

Ideally you choose the most efficient method here, but you can also choose a method you are already using or a method in between. A suggestion is given below:

The thought-steps:

  1. select factory
  2. build a new reactor with factory in a good spot (or simply land it)
  3. select the correct barracks
  4. lift the barracks
  5. tell barracks to land at reactor
  6. (Delayed step!) Set the rally points after the barracks has landed and build 2 marines.

In terms key-presses this is broken down into:

  1. select factory control group
  2. lift (factory)
  3. build reactor (factory)
  4. lift (barracks)
  5. land (barracks)
  6. Since step 6 is delayed we don’t need to optimize its speed, instead I simply note that while waiting for barracks to fly over it can be a good habit to build scvs, look at supply, check upgrade progress, watch minimap, or simply take a breath and think about what your game plan is.

The fastest way to press keys depends on your hotkeys, and it is a good rule to avoid using the same finger twice in a row. If you for example use the same finger for selecting barracks, produce marine and produce marauder you might want to make a small hotkey change. But given your hotkey layout you can practice pressing the keys in the order given by the steps above.

For mouse movement we have to move the cursor to a good reactor location during steps 1-2. We then have to move the cursor to the barracks between steps 3 and 4 as well as to the reactor during step 4. In other words the mouse movement will slow down step 3 to 4 the most. You should learn to be very fast at pressing key 2 after key 1 since there is no mouse movement in between. Pushing down key 4 should be triggered after the mouse movement has been successful rather than directly after key 3 is pressed, and the keyboard hand movement to key 4 should be made at the same time as mouse is being moved.

The goal now is to trigger this entire chain of actions from the wish to swap a barracks over to a reactor-factory and to have the entire process be subconscious and fast so that you can think about strategy while doing it. In order to achieve this I propose the following method:

Go into a solo game or unit tester and follow the steps above without deviating from the points and without making mistakes. Think about what is slowing you down. Is it pressing a key? Not remembering the next step instantly? Bad coordination that’s preventing you from moving the mouse and pressing keys at the same time? Other? If you take too long to press a key you can simply ignore the mouse and thinking and just write down the key order and practice pressing keys in that order. If it’s the thinking that slows you down then try to simplify tasks, for example make up a system for how to position structures so that you don’t need to think about where to build the reactor (for example in a vertical line). If you lack coordination between hands then force yourself to move the mouse and press keys at the same time painfully slowly. Speed will come quickly in the first 4-7 days of practice.

Don’t forget to use spaced repetition here, it is impossible to learn a long chain of movements to a subconscious level in a single session, it will take 4-7 days with 3-5 shorter sessions each day where you practice until it feels comfortable at the end of the practice. You will be faster at the task you choose to improve after only a week. After a week you can start learning more mechanics, but don’t forget to get back to training the first mechanic again after 2 weeks and then after 4 weeks so that the practice is remembered by the brain. Keeping a schedule for this is highly recommended. When you go back into real games it is important that you keep performing the routine that you practiced in solo games when reactors need to be built. It is easy to get back into bad habits under pressure but the only way to remove the old bad connections is to stop doing it completely. Continuing to do it correctly every time is much more important than winning the practice game, remind yourself of that and make yourself feel good about having practiced effectively regardless of the outcome of the game.

It might feel like a waste of time to spend a few hours over a week just to get faster at building reactors, but saving some time and being able to do tasks more subconsciously in every single game does make a difference. Especially if you do this with more and more tasks in upcoming weeks. It might also sound boring to train very simple actions, but then you are probably doing it wrong. You need to stay focused just like in a real game and you also need to be observant of your thoughts and actions. This can be emotionally rewarding, almost meditative. Additionally you will be rewarded after speeding up the mechanic at which stage you can think about strategy while building reactors instead of having to spend repetitive thought effort on it every time. Even if your current thought process of building reactors is subconscious it will still block your conscious thoughts if it is inefficient. After decreasing the size of this block you can play Starcraft more consciously, which feels great.

You might also think that performing 3-5 sessions a day is impossible, but keep in mind that a lot of these repetitions are only 5-15 minutes long and can be done before and after a ladder session. In other words you only need to play Starcraft twice a day during this most intense period. After the initial 4-7 days a single repetition per day is sufficient to maintain the level you reached, and after 2-3 weeks regular ladder practice will be all you need to maintain that level for years. As a final note to convince you to try this method I want to add that feeling improvement is incredibly motivating and rewarding. It is hard to feel improvement on a competitive ladder when your mmr might be the same as it was a year ago. Splitting practice into smaller pieces like this will showcase your improvement which feels great.

Ladder

Other than the special training to improve specific mechanical tasks one at a time we also want to make the most out of our training games on the ladder. An important mindset is to treat ladder as long term practice and not focus too hard on winning. People often say “The best way to improve mechanics is to just play ladder.” but playing mindlessly will not necessarily give you better mechanics after playing for more than a year. A useful tool when laddering is to focus on a single thing, for example creep spread to get in the habit of doing that efficiently; this should also be done with spaced repetition in mind. This directed focus can be used simply to get into the habit of doing something that you otherwise forget, but your focus can also be to spread creep mechanically faster than you currently do without giving up accuracy.

The final key component of ladder practice that I will bring up is to play outside of your comfort zone which makes for a much greater variety of mechanical practice. It is also important to switch up your builds and strategies occasionally. If I play song A on the piano during the entire year and play song B only during autumn then song B will be player better because I had better fundamentals when I started learning it. This can be counter-intuitive and is of course not true if the time periods were very short. It is however harder to get rid of errors that come from bad fundamentals in song A compared to learning to play song B from scratch when the fundamentals are stronger. The same applies to Starcraft. You don’t become good at a build by doing it for 3 years straight, but instead by working on your fundamentals and mechanics and then applying them to learn the build. If your fundamentals and mechanics are up to par, then learning a build is a very simple task that will only take a few days, not years. If you struggle with performing a build or strategy after doing it for more than a week then that is because of weak fundamentals, not because you haven’t done it long enough. So with term practice in mind, keep switching up the way you play to improve fundamentals faster. This will help get rid of bias and mistakes. As a bonus it will also improve your game understanding significantly faster as well.

Summary

To summarize, using our understanding of the phrase “Neurons that fire together wire together” we can learn any task we set our mind to. The method that I proposed helps us break down mechanical tasks and speed them up individually. Using spaced repetition we can learn it with a low total amount of work. And we can make better use of the ladder by directing our mental focus and playing strategies that are new and challenging.

1 Like