Introduction

Reading is weird. The community has come up with pretty good theories for playing osu! which can be backed up either by common knowledge, or by science. Aim is just muscle memory. Snap/flow is about forces, acceleration, and momentum. Tapping stamina is... pretty much in line with just other kinds of exercise. However, we've never been able to come up with a rigid explanation for reading; it just seems to escape us. Luckily, some cool studies in neuroscience might be able to help us know what reading actually is, so we could pinpoint our weak spots and overall become much better players. This guide will deal with osu! reading, scientific explanations for reading, and how we can connect the two ideas and get a better picture of how to improve.

Reading, as we know it

Reading is about patterns. We see a pattern. We know which order to click the circles. Then we play it.

We're very used to seeing the same old patterns over and over, and mapping meta has done a pretty good job of making sure we see the same old patterns over and over. This seems to be no problem as long as the farmers stick to the same old AR10. But, when they encounter AR9 (or god forbid, AR8!), everything falls apart. There seems to be this strange divide between learning how to play high AR and how to play low AR. If you go higher than you're used to, it feels like circles are blazing by and you can't hit them in time. Go lower, and then the playfield becomes a tangled mess, and you just can't work out which circle to click in time. In both of these cases, it seems to be a mental delay which causes us to go out of sync with the rhythm.

Since it takes time for our mind to process patterns, it follows that we actually read ahead of what we play. It's not easy to notice, but it makes sense: For us to play naturally, our brain wants us to be reading patterns ahead of time, so that we have the time to process them and play properly.

From this, we can say that there are two major obstacles to reading: "reading time" (high AR) and number of notes (low AR). We still don't know the exact steps; the way our brains actually do it, even if we have some idea of how it goes. We also don't know the exact skills involved; which part of reading is just made-up stuff, and which part is a real thing that we can train and get better at.

Eye-Hand Span

Eye-hand span is one of the lesser known concepts of neuroscience. It's been developed as one of the explanations on how our brain handles taking in chunks of info, alongside "peripheral vision" and "saccades & fixations". On closer inspection, it seems to fit snugly with our description of reading, the same way muscle memory fits into aiming.

A study (Furneaux & Land) from the Sussex Centre for Neuroscience, about how skill affects the way pianists read sheet music, introduces us to the concept. Eye-hand span is the mental delay caused by the entire process of acting on live information in four distinct steps: Detecting info, processing it, storing processed data in an internal buffer, then translating it into signals at the last second, telling the body to move. In short, it is an analogue of what we osu! players call "reading", except in a much clearer way. Most players simplify (and think of) reading as just the first part - essentially "reacting" - and minimize the middle steps as much as possible. However, this may actually hurt their improvement.

There are two ways to measure eye-hand span. "Note index" is how many objects the performer reads ahead. "Time index", on the other hand, is how far ahead in time the performer reads. The study shows that, the note index gets bigger as the pianist becomes more skilled. Novice players can only read at least two notes consistently, but professional players can go up to at least four notes. This gives us a clue that professionals could process or store info a lot more efficiently and thus play better than the novices. Time index is a lot more interesting, however.

In time index, novices and professionals all played with the same average delay. Instead, the delay changed when the tempo (BPM) of the performance changed; higher BPM meant a lower time index, and vice versa. This makes sense, since you can end up filling your note index with less time if the song was faster. This confirms the idea that BPM and AR are both factors in reading difficulty, as shown by some charts made in the osu! community.

Though average delay is not related to performers' skill, what matters is actually the fluctuation in their actual time index. Novices played at a common average with professionals, but their reading was very erratic; they often read too far ahead or too far behind. This means that another clue to skill is if performers can keep a consistent pace with how far along they read.

Another study (Truitt et al.) from the University of Massachusetts provides more data using "moving-window techniques". The pianists' view of the sheet music is limited and only shows a few notes ahead during play (similar to AR in osu!). The general trend is that smaller window sizes (higher AR) resulted in poorer performance. This is reasonable since pianists do not play with a restricted view unlike osu! players. Overall, professionals delivered more consistent timing, better eye-hand spans, and shorter fixations (meaning they spent less time focusing on a single object) than novices in both unrestricted and restricted cases. In this case, skill seems to correlate both to no window (low AR) and small window (high AR) performance, even if the latter was never practiced.

Okay, but what does that actually mean?

The idea of eye-hand span lets us give a more solid meaning to "reading". We can now say that reading is a four-step process: "detect, process, store, translate", and that better reading simply means being better at the process/store parts, which shows in being able to read further and more consistently. We have definite proof that BPM should be considered when talking about reading difficulty.

We can also draw parallels with the research and approach rate in our circle-clicking game. When we play with lower AR, there are more objects and the playfield becomes cluttered. This forces the player to actually process and store info instead of reacting to it. Professional pianists, despite not being exposed to window size (higher AR), was able to adapt better using their experience and efficiency in breaking down data. While there is no supporting data for higher AR improving reading proficiency, it can still be argued that being able to detect and translate basic patterns effortlessly is as important a skill, and may in fact be a stepping stone as it allows the brain to focus completely on deciphering more complicated patterns. It seems that practice for the two works well with each other.

Along with anecdotes from regular and top players alike that low AR practice has improved their overall play, there may be benefits to having a wider comfort zone for both lower and higher ARs, rather than sticking to a single comfortable AR that many players have.



(note: this guide is kinda rushed due to a tight schedule, it may be polished as time goes.)

I agree with everything, this post and the linked studies are explaining well and confirming what i'm feeling when i play osu!

a nice read, refreshing to see something like this on forums than the old classic "how to git gud"

Nice, you took on my advice of posting links!

I like some parts of this but there are a couple of things I'd like to point out.

Antiforte wrote:

Reading is about patterns. We see a pattern. We know which order to click the circles. Then we play it.

First, reading is not completely about patterns to everyone. It really depends on who you ask. I personally do not read patterns most of the time and break everything down to individual circles. Obviously, I can still see patterns but when playing extremely high density maps, I actually find it counterproductive to try chunk everything into patterns because there's just too much going on. There's a time and place for it.

Antiforte wrote:

The idea of eye-hand span lets us give a more solid meaning to "reading". We can now say that reading is a four-step process: "detect, process, store, translate", and that better reading simply means being better at the process/store parts, which shows in being able to read further and more consistently.

Second and going along with the first point, reading in osu! is not about processing and storing everything. It would be cool if I could get some other low AR players (EZ mod or better) to confirm but when I play high object densities, I don't actually know what's going on. Improving reading object densities at those levels is actually all about being faster at distinguishing the useful information from the junk and then ignoring the junk until it's relevant. To help with this, you use a bunch of heuristics such as following follow points or counting numbers. Additionally, I often will not know which circle I have to hit next and have to actively find it as I need to hit it!

This is of course talking about sight reading, once you add in any sort of memorization, all steps of reading become less and less relevant.

Antiforte wrote:

In time index, novices and professionals all played with the same average delay. Instead, the delay changed when the tempo (BPM) of the performance changed; higher BPM meant a lower time index, and vice versa. This makes sense, since you can end up filling your note index with less time if the song was faster. This confirms the idea that BPM and AR are both factors in reading difficulty, as shown by some charts made in the osu! community.

Third, object density already accounts for reading difficulty increases due to BPM and AR. Object density generally increases as BPM increases and AR decreases.

Finally, there's a disconnect between comparing piano sight-reading to osu! and you should add an explicit disclaimer to inform readers of this since you're essentially comparing apples to oranges. I do agree there are items of merit here (like the 4 step process and time index fluctuations) however the way we receive information is completely different as well as the way we process and act on it is also likely different.

Antiforte wrote:

Along with anecdotes from regular and top players alike that low AR practice has improved their overall play, there may be benefits to having a wider comfort zone for both lower and higher ARs, rather than sticking to a single comfortable AR that many players have.

Just my 2 cents here, the merit in practicing low ARs comes from the ability to practice higher object densities. If you were not to play low ARs, the only way to practice high object densities would be to play harder and harder maps. At the same time, those maps are also extremely challenging to your ability to aim and accuracy. This makes improving in reading higher object densities a challenge. I also wouldn't recommend trying to get better at a larger range of ARs. It's better to be a master of a single AR than be a jack-of-all ARs. Each AR has very specific timing aspects to them which you only learn by playing them.

Almost wrote:

Nice, you took on my advice of posting links!

I like some parts of this but there are a couple of things I'd like to point out.

Antiforte wrote:

Reading is about patterns. We see a pattern. We know which order to click the circles. Then we play it.

First, reading is not completely about patterns to everyone. It really depends on who you ask. I personally do not read patterns most of the time and break everything down to individual circles. Obviously, I can still see patterns but when playing extremely high density maps, I actually find it counterproductive to try chunk everything into patterns because there's just too much going on. There's a time and place for it.

Antiforte wrote:

The idea of eye-hand span lets us give a more solid meaning to "reading". We can now say that reading is a four-step process: "detect, process, store, translate", and that better reading simply means being better at the process/store parts, which shows in being able to read further and more consistently.

Second and going along with the first point, reading in osu! is not about processing and storing everything. It would be cool if I could get some other low AR players (EZ mod or better) to confirm but when I play high object densities, I don't actually know what's going on. Improving reading object densities at those levels is actually all about being faster at distinguishing the useful information from the junk and then ignoring the junk until it's relevant. To help with this, you use a bunch of heuristics such as following follow points or counting numbers. Additionally, I often will not know which circle I have to hit next and have to actively find it as I need to hit it!

This is of course talking about sight reading, once you add in any sort of memorization, all steps of reading become less and less relevant.

Antiforte wrote:

In time index, novices and professionals all played with the same average delay. Instead, the delay changed when the tempo (BPM) of the performance changed; higher BPM meant a lower time index, and vice versa. This makes sense, since you can end up filling your note index with less time if the song was faster. This confirms the idea that BPM and AR are both factors in reading difficulty, as shown by some charts made in the osu! community.

Third, object density already accounts for reading difficulty increases due to BPM and AR. Object density generally increases as BPM increases and AR decreases.

Finally, there's a disconnect between comparing piano sight-reading to osu! and you should add an explicit disclaimer to inform readers of this since you're essentially comparing apples to oranges. I do agree there are items of merit here (like the 4 step process and time index fluctuations) however the way we receive information is completely different as well as the way we process and act on it is also likely different.

Antiforte wrote:

Along with anecdotes from regular and top players alike that low AR practice has improved their overall play, there may be benefits to having a wider comfort zone for both lower and higher ARs, rather than sticking to a single comfortable AR that many players have.

Just my 2 cents here, the merit in practicing low ARs comes from the ability to practice higher object densities. If you were not to play low ARs, the only way to practice high object densities would be to play harder and harder maps. At the same time, those maps are also extremely challenging to your ability to aim and accuracy. This makes improving in reading higher object densities a challenge. I also wouldn't recommend trying to get better at a larger range of ARs. It's better to be a master of a single AR than be a jack-of-all ARs. Each AR has very specific timing aspects to them which you only learn by playing them.

counter 100

First things first. Sorry that I didn't post any links in your thread. That was intentional on my part as I was saving them to make this thread with "reading" itself as the main focus.

Overall, I think this response raises some good counter-arguments. These are my takes on them:

Almost wrote:

First, reading is not completely about patterns to everyone. It really depends on who you ask. I personally do not read patterns most of the time and break everything down to individual circles. Obviously, I can still see patterns but when playing extremely high density maps, I actually find it counterproductive to try chunk everything into patterns because there's just too much going on. There's a time and place for it.

The "Reading, as we know it" section doesn't reflect my personal view. Rather, the plan was to take the community view of reading and see how far I could run with them. The end result is it realizes the limits of our definition with reading, and that becomes the problem to solve + the main focus/purpose of this thread. I'm in the camp that all reading is best treated as individual circles, object density and clutter than in terms of patterns, but I failed to mention this directly in the post.

Almost wrote:

Second and going along with the first point, reading in osu! is not about processing and storing everything. It would be cool if I could get some other low AR players (EZ mod or better) to confirm but when I play high object densities, I don't actually know what's going on. Improving reading object densities at those levels is actually all about being faster at distinguishing the useful information from the junk and then ignoring the junk until it's relevant. To help with this, you use a bunch of heuristics such as following follow points or counting numbers.

Processing and storing isn't everything. I just emphasized them because they are more relevant to the average player who is used to basic AR10 patterns. Based on Furneaux & Land's study, filtering out irrelevant data is a necessary component of eye-hand span so the limited internal buffer is not overloaded. In the formal definition of reading, this filtering of irrelevant junk falls under the "process" step.

Almost wrote:

Third, object density already accounts for reading difficulty increases due to BPM and AR. Object density generally increases as BPM increases and AR decreases.

I wasn't trying to take credit for object density as a measure. I wanted to confirm it, framing it into the concept of eye-hand span as a decreased average time index meaning less time to process each note. That was actually what those "charts" I mentioned were; they were object density charts, but I failed to come up with the term during writing.

Almost wrote:

Finally, there's a disconnect between comparing piano sight-reading to osu! and you should add an explicit disclaimer to inform readers of this since you're essentially comparing apples to oranges. I do agree there are items of merit here (like the 4 step process and time index fluctuations) however the way we receive information is completely different as well as the way we process and act on it is also likely different.

Yes, piano sight-reading has differences from osu! reading, there is no disputing that. But, rather than comparing apples to oranges, I think this thread is more about taking apples and oranges to study eating. We can agree that apples are different from oranges, but we can also agree that both are foods and share the property of being eaten. Same with pianos and osu!. We can agree that they're different things, but they both involve live reading tasks and so eye-hand span is relevant to both of them.

I feel that the analogies set up were fairly good, and discrepancies between the studies and typical osu! play were addressed. To my attention, there isn't any major difference between the two that weakens the analogies. It's also important to note that eye-hand span also applies to typists where the "type of information" and "the method of processing info" are, arguably, pretty different from pianists. Eye-hand span is not a one-trick pony.

Almost wrote:

I also wouldn't recommend trying to get better at a larger range of ARs. It's better to be a master of a single AR than be a jack-of-all ARs. Each AR has very specific timing aspects to them which you only learn by playing them.

I believe that all aspects of reading are interlinked, and that this is just not taken advantage of because players aren't using newer training methods discovered by neuroscience. Again I wanted to save this for another neuroscience-related thread and so I do not have the sources compiled yet, but the gist of it is this:

Many studies have proven the existence of benefits from practicing two related skills back-to-back. This process of "mixed practice" results in the phenomenon of "interleaving". Interleaving makes mental improvements solid, promotes skill transfer (meaning improving one skill also improves the other), and even boosts improvement in related tasks that were never done before. This phenomenon has been observed for mental skills like mathematics, as well as physical skills with mental aspects like different ball tricks which involve tracking the ball and hitting/catching it in time.

It is still good to stick to one approach rate in the middle range you call your best, since this is the condition where you will improve the fastest ("specificity"). In some respect, you could say those players who stuck to AR10 are already ahead of the game. But, it is just as important to extend this skill to different conditions or approach rates, which may even improve your performance in that middle AR thanks to interleaving. Arguably the ultimate goal for any competitive sport is to reach high "variability", or great performance across a wide variety of situations. What we call "well-rounded" players.

Antiforte wrote:

Almost wrote:

Third, object density already accounts for reading difficulty increases due to BPM and AR. Object density generally increases as BPM increases and AR decreases.

I wasn't trying to take credit for object density as a measure. I wanted to confirm it, framing it into the concept of eye-hand span as a decreased average time index meaning less time to process each note. That was actually what those "charts" I mentioned were; they were object density charts, but I failed to come up with the term during writing.

I think basic math is enough to really prove it honestly. There's no point adding complexity where it isn't needed.

Antiforte wrote:

Yes, piano sight-reading has differences from osu! reading, there is no disputing that. But, rather than comparing apples to oranges, I think this thread is more about taking apples and oranges to study eating. We can agree that apples are different from oranges, but we can also agree that both are foods and share the property of being eaten. Same with pianos and osu!. We can agree that they're different things, but they both involve live reading tasks and so eye-hand span is relevant to both of them.

I feel that the analogies set up were fairly good, and discrepancies between the studies and typical osu! play were addressed. To my attention, there isn't any major difference between the two that weakens the analogies. It's also important to note that eye-hand span also applies to typists where the "type of information" and "the method of processing info" are, arguably, pretty different from pianists. Eye-hand span is not a one-trick pony.

My point was more just about informing people who aren't necessarily science-y of possible discrepancies in the comparison. There are many things that seem right in theory but don't match up in reality. I just wanted to make a point of it on ethical grounds as many people out there misportray science (especially in the media) which actually results in more ill-informed people.

The analogy is fine but just remember, we are dealing in complex systems and any type of discrepancy can cause large changes down the line (chaos theory). For example, caffeine from coffee effecting the body differently to caffeine from a pill. An example of a discrepancy I was alluding to earlier would be the way the information is given. In piano sight reading (and even for typists), information is processed in a linear fashion (reading left to right) whereas in osu! the information is presented all over such that noise filtering becomes an even greater challenge. I'm not attacking the eye-hand span idea at all, it's just some food for thought.

Antiforte wrote:

I believe that all aspects of reading are interlinked, and that this is just not taken advantage of because players aren't using newer training methods discovered by neuroscience. Again I wanted to save this for another neuroscience-related thread and so I do not have the sources compiled yet, but the gist of it is this:

Many studies have proven the existence of benefits from practicing two related skills back-to-back. This process of "mixed practice" results in the phenomenon of "interleaving". Interleaving makes mental improvements solid, promotes skill transfer (meaning improving one skill also improves the other), and even boosts improvement in related tasks that were never done before. This phenomenon has been observed for mental skills like mathematics, as well as physical skills with mental aspects like different ball tricks which involve tracking the ball and hitting/catching it in time.

It is still good to stick to one approach rate in the middle range you call your best, since this is the condition where you will improve the fastest ("specificity"). In some respect, you could say those players who stuck to AR10 are already ahead of the game. But, it is just as important to extend this skill to different conditions or approach rates, which may even improve your performance in that middle AR thanks to interleaving. Arguably the ultimate goal for any competitive sport is to reach high "variability", or great performance across a wide variety of situations. What we call "well-rounded" players.

I agree that reading aspects are linked. For example, being good at reading high object densities on low ARs definitely translates well to reading high object densities on higher ARs. However, practicing one aspect actually reduces practice in another aspect. The main question is "is it worth the time?". I personally haven't gone over any studies for this so take what I say as a potential criticism. In terms of striving to become a "well-rounded" player, I would have to say that comes with it's own set of advantages and disadvantages. The advantage being you can do a lot with the drawback being you won't be better than someone who specializes. It's up to the individual to decide what they want based on their own goals and what they find enjoyable.

I knew I would find Almost in this thread. The resident reading analyst lol. I really enjoyed this writeup. I'd like to see the implications for this as far as how someone might apply this to playing in order to improve. I understand you touched on that a bit with the idea of mixed practice and ARs. From my understanding, the idea would be to practice low AR and then high AR to improve your medium AR which is your preferred?

You mentioned that not understanding this process can hurt our improvement, but how can we use this knowledge practically besides mixed practice? Not that anyone has the answer, but I think it's a question worth looking into.

I'm interested in what we can learn from the detect, process, store, and translating steps as far as to how we can approach our practice. This is just a random hypothesis, but for example I wonder if improving the translation of a pattern properly then makes it easier to process and store because your brain already knows what it needs to translate and thus you can take things in by way of chunks of movements instead of one by one, and then, if you can properly execute those chunks, it makes it easier to smoothly read, which is one of the things you mentioned professionals being good at. That's just an example and not even what I really think may be happening, but it's interesting to theorize about these things.

That theory comes from Cookiezi saying he treats things like patterns where if he can do the pattern then he can play it. I know not everyone plays this way but there's reason to believe (cookiezi's skill and the prevalence of patterns in mapping) that it may be an efficient way to improve - practicing the proper execution of patterns that is.