A Tortoise, a Serpent & Their Metronome

0. Information

Kit

Here are some numbers for Wuyang:

The staff is named for Xuanwu, the Black Tortoise-and-serpent of Chinese myth; the element is water. That is where the serpent in the title and the water orb in the hero both come from.

1. Rhythm

Always be moving

We always want to be moving.

Think of boxing. It's almost always better to be moving than standing still, but movement burns energy, and if you bounce around too much, you spend more than you can recover: a boxer learns to conserve.

In a game like Overwatch, the cost of movement is very low. Pressing a key is isomorphic to throwing a punch, but it costs a fraction of the energy of that punch, so there really shouldn't be much thought about conservation. So then, the trick becomes to make movement a reflex.

One way to do this is to latch onto a rhythm, some tempo, and then let your movement match that tempo. The bar pinned to the bottom of the page runs at a fixed BPM and speeds up as you scroll. It is a teaching accelerator, not Wuyang's literal fire rate. Every figure is locked to that bar, so you can feel what breaks when the tempo is pushed past what a stop-and-reset style can keep up with.

In real terms the staff fires about three orbs a second, and because a steered orb has no gravity and lives for up to 2.5 seconds, you can hold an unbroken line of fire, one orb handed off to the next. The only forced rest is the 1.5-second reload every 20 orbs. Whatever tempo you settle into sits inside those limits, and these drills are about making one effortless to hold.

First, reproduce a short burst of movement. Press Begin, watch the left hand flash a pattern, then repeat it on your own keyboard.

Now let's try it on the beat. Turn on Click in the bar at the bottom, hit Begin, and play along.

The guide hand runs a simple loop on the beat; match it, key for key. What you want ultimately is to press each key at the moment it lights up, right on the beat. The demo gives you a little leeway so you can follow along while you learn the feel, but the target is dead-on timing.

The readout scores your last ten presses, and if you change the BPM at the bottom the guide speeds up or slows with it, so you can hold the rhythm at any tempo.

Now, let's factor in the orb. When you fire, you can release at a rhythm as well, and see what works for you.

Hold to charge, release at full weight. Look at the statistics below to assess your progress.

When we think about firing at a tempo, the point isn't to fire slower or faster, the point is to fire with a structure and then modulate that structure to increase your proficiency. In other words, you learn to fire at a tempo and then you increase the tempo. When we are thinking too much about having to always send orbs, it becomes stressful, and it takes longer for the habit to become second nature. Hence, we practice to make it a steady state for ourselves. In isolation you might think that simply maximizing fire rate would produce the most damage, but it is more about keeping up a fire rate throughout an entire game than about any siloed situation within it, and a tempo, a rhythm, just makes that immensely easier.

The figure below shows exactly that: one fight, two players. One holds a steady tempo and never stops; the other keeps having to think about it and stalls. Watch the damage add up.

Movement on a beat, and the orb released on its own beat inside it. That's rhythm. Next we stop firing straight and start bending the path.

2. Curvature

Direct, or a curve first

A shot doesn't have to leave in a straight line. Because the orb steers for up to 2.5 seconds, you can keep aiming after you fire, and where you aim from changes.

Two ways to take the same shot. Direct: aim at the target and release, the orb goes straight. Curve first: release and keep steering, using the charge-up time to correct, and only commit the aim at the wind-up, the instant it hits max weight. The difference is subtle and it matters: in the second one the orb has already travelled forward, so you're effectively aiming from closer. A closer shot is an easier shot, and also harder to avoid.

Same target, reached two ways. The straight shot is a single commitment out of the staff. The guided orb is the same shot, but shaped: you spend the travel time steering it, so it can round corners a straight line can't, and it commits its aim later, from closer.

And the curve doesn't have to be lofty. Two orbs to the same point, both clearing the same cover: one arcs high, the other skims just over the top. Same speed, but the tight one travels less distance, so it gets there sooner.

Next we build on this: how to actually apply curvature in the way that helps you most.

3. In Practice

Pathing

Layers, built in order.

3a. Discrete

In this first part we consider discrete actions. We've already met a couple: moving to a tempo, and sending the fired shots. Those individual movements need to be smooth, accurate, and automatic. Practice them as individual pieces, and find the continuity inside each fundamental. When you press the movement keys, for instance, they should be played legato, one flowing into the next. Once each movement follows that rhythm, you can begin putting them together.

As a function, the discrete style is path = pick(pₙ): each orb is its own aimed choice, picked from a list of targets. Every orb is independent, it shares nothing with the one before it, so for each one you pay the full cost again: find a target, aim, judge the release, commit. Nothing carries over. That is why N orbs cost N decisions, one complete choice per orb, and why attention scales directly with volume: the more you fire, the more you have to think. That is the cost the next two layers exist to drive down, from N decisions, to 1, to roughly 0.

The other fundamentals, moving with the team, positioning, taking space, and so on, should all be mastered individually and then re-learned as you interpret them together. Being on high ground might make sense in the average case, but conditioned on where your other support is, who your tank is, and so on, it may not be optimal. You have to understand that in real time, and understanding in real time is exercising your fundamentals all together as one system.

Get each of these good on its own, then good in combination:

3b. Continuity

With the individual parts down, continuity is about moving them all together, like an orchestra rising into a crescendo. Every fundamental you drilled in isolation now plays at once, in time, and the whole thing flows. The mental shift for the demo below is this: each orb you throw is not a separate event, it is one continuous action that happens to go solid for a moment. So be relentless, keep sending them, and let the stream read as a single line rather than a series of shots.

Curvature is also what makes the kit non-linear. A straight shot can only reach points on a ray out from the staff. Curving the orb lifts that limit: instead of a few straight lines, you can reach a whole area. Every bit of curvature you control adds more of the map you can hit from the same spot.

This is where the discrete list collapses into a single function. Because the orb steers for up to 2.5 seconds with no gravity, its whole trajectory is produced, one sample per frame, from your camera angle:

The input is your aim over time, θ(t); the output is the curve. You are choosing a function, not a point, and it keeps resolving for the full window after release. So the whole stream is one decision, the function, evaluated every frame, and attention no longer scales with the number of orbs.

3c. Modulation

Now you decide when to be fluid and when to be direct, and you're the master of that switch, which also makes you unpredictable, because it can't be read.

As a function this splits into two pieces:

The carrier C(t) is the base cycle, fixed by the beat and free to run, it costs no attention. Everything that decides whether you are readable lives in the modulation on top: the amplitude A(t) and the phase φ(t). So the decisions per orb fall to roughly zero, and you spend attention only on those two. That is the sense in which it is both mathematical and improvised: the carrier is the fixed mathematics, the modulation is played live against a real opponent, who is running loops of their own.

Thinking of the stream as a function is what lets you shape it. The same orbs can read completely differently depending on how you bend them: tight or wide, fast or slow, on the beat or deliberately off it. The figure below is one carrier shown two ways, bend the three knobs and watch the same object take on a different character.

The three knobs above are the three parameters you actually control. Amplitude is sweep width: narrow to stack orbs on one point, wide to contest a whole area so it reads as occupied. Frequency is cycles per beat, it does not change your rate (the beat fixes that), it changes how many distinct directions the orbs arrive from per unit time, which is what pressure actually is. Phase is the offset between your cycle and the opponent's peek, fire, hide, cooldown loop: slide it so your orbs land in their gaps, mid-peek, mid-cooldown, mid-reload. Amplitude and frequency you set; phase is read off your opponent. Press Carrier only to flatten all three, that bare, regular waveform is fully readable within a few cycles, which is exactly why modulation is the point.

This is also where sensitivity matters. The staff steers off your camera, so a higher sensitivity moves the orb more in the same amount of time: more control over fine paths, more of the space actually reachable. Lower sensitivity shrinks what you can do, less range with the orb per second, so any path that needs a tight turn is off the table. Steer the ball through the course below at each setting and watch how much harder it gets when the sensitivity is low.

Which is the case for keeping it high. A larger value reaches more of the field inside the orb’s short life and keeps the tight turns available, and a sharp flick swings the orb further, which will matter more than it looks like it should here. Bias it up. (Hint: there is a whole other reason to run it high, waiting in the Extra after the guide ends.)

One more, and it's the real test. In a real fight the orb is usually off your screen, your eyes are on the enemy or the objective, not on it. But you know the map, so you can steer it through cover you aren't even looking at, running the path from memory while you watch something else. That is what Lights off is for: generate a course, study it until you have it memorized, then turn the lights off and steer the orb through it from the map in your head, smooth or direct as the remembered path demands.

3d. Automation & Thought

When all of it runs itself, the charge, the firing, the path, the switching, something opens up. You stop spending attention on the orb and start spending it on the game: the macro, the timings, what they're thinking, their intentions whether they actively know them or not. The mechanics went automatic, so now you have the capacity to think. That's the ceiling, and you only get there once everything beneath it runs without you.

The demo below trains exactly that split. You still steer the ball through the course, but now a snake and a tortoise are each tracing a curved path across the grid. Once you reach the end you draw the path each one took, and it's matched against the real trajectory. If steering still costs you attention, you won't have watched closely enough to reproduce them. Only when the guiding runs itself do you have the spare capacity to read the board. That is the whole point: the mechanics take care of themselves so your mind is free for the decisions, the intentions, the fight. Then turn the lights off and do it without seeing the cover at all.

Here's why it's worth it. What actually matters isn't how accurate any single orb is, it's how much you put out across the whole game. Aiming every orb costs attention, and attention runs out, so a pure-precision style starts sharp and falls off as the game goes long. A steady stream starts a little less accurate but holds there, since there's no per-orb decision draining you. Over a full match the stream just does more, and the accuracy you traded away comes back as pressure, the kind that makes people misplay but never shows up on the scoreboard.

Roughly, this is how I feel progression looks across the ranks:

Extra

Aiming a thinking opponent

Everything so far assumed the target keeps its mind. Release at the earliest point of maximum charge, on the beat. But a good opponent is aiming back, and the instant they can dodge, a second layer opens: not when to release, but how to aim a thing that curves at someone who won't hold still.

Say I see a target. They are looking right at me, and the moment I commit, they jump to their left. I have already sent an orb. There were two ways to have taken that shot.

Predict. Throw straight, but lead: aim not at where they are, but at where they will be when the orb arrives. If they keep going, we converge. It is a bet on their path.
Pursue. Do not guess the endpoint. Fire and keep steering, bending the orb toward wherever they actually are, correcting for the whole charge window. It is not a bet; it is a conversation.

That is the whole distinction. Prediction is open‑loop: you commit once, then go blind. Pursuit is closed‑loop: you keep a hand on it the entire way.

The straight lead has a clean geometry. If they strafe sideways at speed u and the orb flies at speed v, you lead by an angle whose sine is u/v; point that far ahead and, if they hold the line, you meet. The trouble is the “if.” A juke is them changing their mind, and your lead was a bet on the old one. Worse, the error grows with range: the orb spends time d/v in the air, so a full reversal in that window throws your point off by about 2u·d/v. At six meters that is already a couple of body‑widths of empty air, a clean miss on a spot they already left.

Pursuit does not care about the endpoint. Because you correct live, you only have to out‑turn the juke, not out‑guess it. Inside the steering window, roughly the first 7.5 meters, the orb bends after them faster than they can re‑decide, so the miss never gets a chance to accumulate. Prediction fails more the farther out you are; pursuit fails only when they leave the window, or juke harder than the orb can curve.

Here is the part worth the whole section. Pursuing costs something: to steer the orb after a target juking to my right, I have to turn. My view, and Wuyang's body with it, swings right to keep the orb on them. Naively that is a loss: I have spent my orientation, and I cannot now snap a fresh straight shot back the other way.

But the turn is not spent; it is banked. My body is now pointed right, exactly the amount I steered. So I release a second orb immediately, into that rightward facing, and curve it hard back left, through the same angle I just turned, so it arrives at the spot the target left, where the rest of their team is still standing. Orb one chases the one who moved; orb two, a breath later, curves to the ones who did not.

The angle I paid to track the juker is the exact angle orb two's curve uses to get home. Nothing is wasted. The consequence of the first shot is the setup for the second, which, if you have been reading, is the entire thesis of this guide, now pointed at two enemies at once.

It looks like it should cost fire rate: surely holding and steering one orb means the next comes late. But the second orb is a curve, not a re‑aim. The alternative, straight orb one and then a flick of the mouse all the way back to the cluster for straight orb two, costs the same fire‑rate floor plus the flick: the reaction to re‑aim, the precision to land it, and the chance that someone jukes during it. The spin‑throw pays the floor and skips the flick. The curve is the aim, and if you know the map and can see the arc in your field of view before you throw it, a memorized curve is steadier than a panicked flick. You keep your rate of fire and buy accuracy with geometry you were going to move through anyway.

So which wins? It is not taste; it splits cleanly by range, and by whether they are thinking:

Point‑blank, under ~2 m. Just throw straight. The orb arrives before a juke can matter (2u·d/v is tiny), and straight is faster. Steering here is pure overhead.

The steering band, ~2–7.5 m, against a juker with their team near. The spin follow‑up is objectively better. Prediction is fragile exactly here (enough air‑time to punish a lead), pursuit is available (inside the window), and the turn you spend tracking one enemy is repaid as a free curve into the others. This is its home.

Long, past ~7.5 m. You cannot steer that far, so you are forced back to the straight lead, accepting that a hard juke may beat it. Prediction is not better here; it is just the only option left.

And two quieter cases keep the straight shot even inside the band: a target moving in a committed straight line (no juke to punish, so leading is clean and cheaper), and any moment you cannot afford to rotate your view (a flank to watch, a duel to hold), because pursuit turns your body, and that turn is only an asset if there is a second shot worth setting up.

Does it hold up? It does. Strip the names off and it is a textbook trade, interception versus pursuit, open‑loop versus closed‑loop, with one genuinely clever addition: the aim‑rotation, usually counted as the cost of steering, is re‑spent as the aim of the follow‑up. Two orbs, one turn, and the second lands because of, not in spite of, the first.

Push that to its limit. The juke was gentle a moment ago; make it enormous: the target breaks almost behind me, a near‑180. To keep orb one on them I do not bank a soft turn, I whip the camera around, and Wuyang’s body snaps with it. Orb one lands on the runner with my back already half‑turned to where they started.

So I let the turn keep going. I release orb two while I am still facing the wrong way, pointed away from the cluster entirely, and curve it so hard it rides the rest of my spin the whole way round. Orb one swept the first half of the circle; orb two sweeps the second. Together they close a full 360, and orb two arrives back at the spot it all began from, where the team never left.

None of that spin is free, and here the earlier note on sensitivity comes due. A near‑180 whip only lands inside a charge window if the camera actually gets there in time, and how far the view swings per centimeter of mouse is exactly what sensitivity sets. Overwatch even breaks the orb out on its own: Relative Sensitivity While Controlling the Water Orb scales your turn rate while an orb is out, on top of your base value. I run it at 160%, so the orb steers hotter than my aim does and the whip that closes the 360 is a flick, not a heave. Drop that number and the second half of the circle never arrives: the body cannot come around fast enough, and orb two dies facing the wrong way.

The plainer example from earlier points the same direction. When the honest shot is too hard (the angle is ugly, the lead unreadable), you do not force it; you curve to a different answer the moment hands you: a softer target, a cleaner lane. That, too, is bought with turn rate, because the harder you can bend, the more of those escape hatches exist. And a curved orb carries a bonus the straight one never will: it arrives from a direction nobody watched it leave. A shot that starts pointed away and lands in your side is not just a hit, it is a surprise, and surprise is a hitscan’s whole certainty turned into the water orb’s edge.

Both roads lead to the same dial. The reinvested turn, the full 360, the curve to the target you can hit, the shot that lands off‑guard: every one of them is really an argument for keeping sensitivity high. There: the reason the hint promised.

This is probably the definition of over-engineering, but you can take away the ideas that stick and revisit this guide to see if you can take away some more if your judgement demands it! Thanks for reading; 🙂 🐍🐢