What's inside
- What counts as an anomaly
- The four channels at a glance
- Stall and counter anomalies: the full cue list
- Environment, audio and state anomalies
- The check routine, step by step
- Which response fits your situation
- Worked example: what an unnecessary report actually costs
- Mistakes that produce false reports
- Frequently Asked Questions
- What counts as an anomaly
- The four channels at a glance
- Stall and counter anomalies: the full cue list
- Environment, audio and state anomalies
- The check routine, step by step
- Which response fits your situation
- Worked example: what an unnecessary report actually costs
- Mistakes that produce false reports
- Frequently Asked Questions
As an Amazon Associate I earn from qualifying purchases. This post may contain affiliate links at no extra cost to you.
Short answer: Bakso Malang anomalies fall into four channels — sight, count, sound and state — and one rule settles all of them: report what differs from the reference frame, ignore what only differs from the previous pass. A sight scan takes 4–6 seconds, a count check 2–3 seconds, and an audio or state check about 2 seconds each, so a complete pass should land between 8 and 12 seconds. Anything slower is usually a sign you are comparing the scene to your memory instead of the baseline.
Sections
- What counts as an anomaly
- The four channels at a glance
- Stall and counter anomalies: the full cue list
- Environment, audio and state anomalies
- The check routine, step by step
- Which response fits your situation
- Worked example: what an unnecessary report actually costs
- Mistakes that produce false reports
- Frequently Asked Questions
What counts as an anomaly
An anomaly is any detail in the stall scene that differs from the reference frame — the single clean version of the scene you are shown at the start of a shift — while every other detail stays consistent. That definition matters more than any individual cue, because the two failure modes pull in opposite directions: report too little and you walk past a real anomaly, report too much and you burn the shift on resets. Both come from the same root cause, which is comparing the scene against the wrong baseline.
Anomaly pools rotate between builds, so treat the cue list below as a taxonomy rather than a fixed checklist. The four channels stay stable across updates; only the specific cues inside them change.
The four channels at a glance
Sorting cues by the sense you use to catch them is faster than memorising a flat list of thirty items, because each channel has its own scan order and its own error profile.
| Channel | What it covers | Scan method | Time per pass | Share of false reports (practical estimate) |
|---|---|---|---|---|
| Sight | Colour, shape, text, placement | Left to right, counter level first | 4–6 s | ~25% |
| Count | Quantities of repeating objects: meatballs, bowls, chairs, queue length | Count in pairs, not one by one | 2–3 s | ~50% |
| Sound | Ambience, radio, boiling, dialogue | Stand still, listen two full seconds | ~2 s | ~10% |
| State | Doors, lights, steam, shadows, your own HUD | Glance at the four corners | ~2 s | ~15% |
Counting is the channel that generates most mistaken reports, and it is also the one people rush. Slow down there and speed up everywhere else.
Stall and counter anomalies: the full cue list
These are the cues tied to the stall itself. They are the ones a first-time player misses most often, because the stall is busy and the eye tends to land on movement rather than on static detail.
| Anomaly | Visual cue | Correct response | Check time |
|---|---|---|---|
| Broth colour shift | Clear or grey broth where the reference is amber-brown | Report | 1 s |
| Meatball count off | Any number other than the reference count in the pot | Report | 2 s |
| Noodle swap | Rice vermicelli where yellow noodles should be, or the reverse | Report | 1 s |
| Bowl stack height | Stack taller or shorter than the reference | Report | 1 s |
| Price board digits | Any digit changed, added or dropped | Report | 2 s |
| Signage spelling | Missing, doubled or mirrored letters in the stall name | Report | 2 s |
| Vendor missing or doubled | Empty serving position, or two identical figures behind the cart | Report | 1 s |
| Condiment swap | Chilli sauce and soy sauce exchanged left to right | Report | 1 s |
| Cart position | Wheels off the kerb line, cart rotated | Report | 1 s |
| Extra or missing garnish tray | Fried shallots, crackers or tofu absent or duplicated | Report | 1 s |
| Queue length | More or fewer customers than the reference queue | Report | 2 s |
| Duplicated customer | The same figure appearing twice in the queue | Report | 2 s |
Environment, audio and state anomalies
These are quieter and easier to argue yourself out of. They are also the cheapest to verify, because sound and state cues do not require you to hold an image in your head.
| Anomaly | Cue | Correct response | Check time |
|---|---|---|---|
| Ambience dropout | No clatter, no boiling, no street noise | Report | 2 s |
| Radio or music change | Different track, wrong tempo, looped line | Report | 2 s |
| Dialogue repeat | A line delivered twice by the same speaker | Report | 2 s |
| Steam frozen | Steam column static instead of drifting | Report | 1 s |
| Light temperature | Warm bulb replaced by a cold green or blue cast | Report | 1 s |
| Door or shutter state | Open where it should be closed, or the reverse | Report | 1 s |
| Shadow direction | Shadows pointing against the light source | Report | 1 s |
| HUD change | Timer font, counter digits or objective text altered | Report | 1 s |
| Nothing changed | Scene matches the reference frame exactly | Do not report | — |
The check routine, step by step
- Step 1 — Stand still and listen for two seconds. Footsteps and menu navigation mask ambience, so audio anomalies disappear if you keep moving. Two seconds of stillness is enough to catch a dropout.
- Step 2 — Scan the counter left to right at eye level. A fixed direction stops you from double-checking the same object and gives you a repeatable order to compare against the baseline.
- Step 3 — Count repeating objects in pairs. Meatballs, then bowls, then chairs, then people in the queue. Pair counting roughly halves the error rate of single-item counting, because you hold one number instead of two.
- Step 4 — Glance at the four corners. Door, exit sign, light source, shadow direction. These are the state cues, and they sit outside your central vision, which is exactly why they are missed.
- Step 5 — Read the text once, left to right. Signage, price board, menu. Text anomalies are easy to miss and easy to confirm, so they are the cheapest points you can score.
- Step 6 — Decide and hold. If nothing differs from the reference frame, submit nothing and begin the next pass. If something does, report once and wait for the scene to reset before checking again.
Which response fits your situation
Most wrong reports come from uncertainty, not from missing a cue. Use the situation, not your gut feeling, to pick the response.
| Your situation | Correct response | Why |
|---|---|---|
| You are about 60% sure a cue changed | Do not report; re-check on the next pass | A missed pass costs nothing, a false report costs a reset |
| Two cues differ at the same time | Report once, then watch whether both clear | Paired cues are usually counted as one submission |
| The change appears only in peripheral vision | Stop, look directly, then decide | Peripheral detection is the largest source of imagined changes |
| The cue survives a reset | Stop reporting it; it belongs to the reference frame | Repeating a correct-looking report is still a false report |
| The cue is audio-only | Listen for one full loop before reporting | Ambience loops are long; a short listen catches false dropouts |
| The cue is in your own HUD | Report it like any scene cue | Interface changes count as anomalies in current builds |
| You already reported the same cue this shift | Check whether the scene reset, then treat it as new | The same anomaly can legitimately reappear after a reset |
Worked example: what an unnecessary report actually costs
The arithmetic is what makes the accuracy-over-speed argument concrete. Assume a pass every 20 seconds, which gives 30 passes in a 10-minute shift, and assume a wrong report costs about 45 seconds of progress through the verification screen and reset.
- At a 5% false-report rate: 30 × 0.05 = 1.5 wrong reports, costing 1.5 × 45 = 67.5 seconds — roughly 11% of a 10-minute shift.
- At a 20% false-report rate: 30 × 0.20 = 6 wrong reports, costing 6 × 45 = 270 seconds — roughly 45% of the same shift.
- Correct reports are nearly free by comparison: three real anomalies at about 5 seconds each to submit is 15 seconds, or 2.5% of the shift.
The cost curve is steep and non-linear, which is why “when in doubt, leave it out” is not laziness — it is the mathematically cheaper policy until your confidence passes roughly 80%.
Mistakes that produce false reports
- Comparing to the previous pass instead of the reference frame. Random variation reverts between passes; an anomaly does not. If the scene “changed back”, it was never an anomaly.
- Counting one by one under time pressure. Single-item counting is where roughly half of all wrong reports originate.
- Scanning in a different order each time. A drifting scan order makes normal objects look new.
- Reporting motion. Steam, queue shuffle and idle animations are supposed to move. Only a frozen or reversed state is anomalous.
- Ignoring the HUD. Players watch the scene and forget the interface, which is where a small but consistent share of anomalies hides.
- Re-reporting after a reset. Once the scene resets, the baseline is restored; anything still different is a fresh anomaly, anything that reverted is not.
Frequently Asked Questions
How many anomaly types are there in Bakso Malang?
Build-dependent, but the working taxonomy is four channels covering roughly 25 to 30 distinct cues. Updates rotate the pool rather than expanding the categories, so a player who has the four channels and a fixed scan order can handle a new build without relearning a list.
What happens if I report an anomaly that isn’t there?
It is treated as a false report and typically triggers a verification screen followed by a reset of the current scene. In practice that costs around 45 seconds of progress, which is why an uncertain report is almost always the worse choice.
Can two anomalies appear in the same pass?
Yes. Paired cues are usually counted as a single submission, so the correct move is to report once and then watch whether both cues clear together. Reporting twice for one paired event is a common source of unnecessary resets.
How long should a single check take?
Between 8 and 12 seconds for a full pass: 4–6 seconds for the sight scan, 2–3 seconds for counting, and about 2 seconds each for sound and state. Consistently taking longer than 15 seconds usually means you are second-guessing rather than checking.
How do I tell an anomaly from ordinary random variation?
Persistence is the test. Random variation changes between passes and reverts to the baseline; an anomaly stays until it is reported and the scene resets. If a cue disappears on its own, it was variation.
Do I need to memorise the whole reference frame?
No, and trying to is counterproductive. Memorise structure instead: the counts of repeating objects, the text on the boards, the colour of the broth, and the state of the four corners. Those four anchors catch the large majority of cues without requiring a photographic memory of the scene.