Selection changes during the Answer phase of Multiple Object Tracking Test never count as errors. Only the act of completing a wrong unordered final pair and pressing Submit records a miss, and even that miss is only logged on the first distinct wrong submission. Every other interaction with the six numbered position buttons — clicking one to add, clicking the same one again to remove, replacing one with a different one, or pressing keys 1 through 6 to toggle — is a free action that does not touch the score or the miss counter. The Submit control stays disabled until exactly two positions are selected, so there is no risk of an accidental submit while a player is still reconsidering. The same rule holds for mouse, touch, and keyboard players: every position toggle is a non-event from the game's perspective. As long as the same wrong pair is on the board and the player has not pressed Submit, or the player has made a different selection and not submitted, the round is still alive. The deliberate design is meant to remove an obvious penalty for hesitating before committing, so the only thing a player should focus on is getting the unordered pair right at the moment of submission.
What counts as an error in this game
The miss counter in Multiple Object Tracking Test tracks exactly two things and nothing else: a submitted wrong unordered pair, on its first distinct occurrence, and any subsequent distinct wrong unordered pair submitted in the same round. Every other interaction with the game — pressing keys, clicking buttons, changing selections, skipping frames, watching the disc paths, restarting, or using the boss key — is logged in the ledger but never advances the miss counter. This is part of the deterministic entertainment game design: the disc paths, frame intervals, target pairs, and target rings are all fixed product fixtures, so the only variable a player controls is the final unordered pair they choose to submit.
The framing matters here. The page where this game lives uses the familiar task description "Multiple Object Tracking Test," but the actual product is an entertainment game rather than a clinical, scientific, medical, vision, attention, or diagnostic assessment. No score is compared with population norms, no claim is made about a cognitive trait, and no result is interpreted as a measure of attention. Treating the error counter as a fair but unweighted tally of submitted wrong pairs — and treating selection changes as a separate, free channel — is the right mental model for understanding the rules.
That separation is the single most useful fact a player can carry into a session: edits are free, submissions are paid, and only a wrong submission opens a recoverable miss.
How to change selections safely during the Answer phase
- Memorize the two ringed discs during the Preview phase. Both targets are marked with accent rings while the disc field is still.
- Press Space or choose Start motion to leave Preview. The rings unmount, every tracking disc shares the same fill, stroke, size, label, and accessible description, and no target data, class, or label follows either object.
- Follow both identical discs through the fixed frames. Each frame is a complete permutation of disc identities zero through five, guaranteeing exactly one disc in every location and preventing overlaps or missing objects. The two targets are identical discs among six, indistinguishable once the rings disappear.
- Wait for the motion to finish on its own — frames advance every 650 milliseconds with a declared CSS transition — or choose Finish motion now to jump to the same final frame for deterministic short play.
- During Answer, the moving SVG is replaced by six numbered position buttons. Click, press keys 1 through 6, or tap to toggle selections. Clicking an active position deselects it; clicking an inactive position adds it until two are selected; attempting a third selection does nothing. None of these toggles counts as an error.
- Press Enter or the Submit control when exactly two positions are selected. A correct unordered pair awards exactly 200 points and returns to Preview for the next target set; an incorrect unordered pair records one miss.
Why only submitted wrong pairs record misses
The validator looks at one thing at the end of the chain: the sorted two-position set the player chose to submit. The internal implementation takes the two target identities declared for the round, finds them in the last permutation frame, sorts the two final positions into a canonical unordered set, and compares that with the sorted set the player submitted. Selection edits that happen before Submit never reach the validator, which is why the miss counter stays at zero no matter how many times a player toggles the six buttons.
A side effect of this rule is that the Submit control enforces its own safety net. The button is disabled until the pair contains exactly two positions, so it is impossible to commit to a half-formed pair. Pressing Enter with zero or one selected position is also a no-op, and pressing Enter with three positions is impossible because the third toggle never registers. The only frames in which a miss can be recorded are the frames in which the validator receives a sorted two-position set and finds that it does not match the canonical target pair.
A correct unordered pair awards exactly 200 points. There is no speed bonus, no click penalty, no adaptive motion, no hidden multiplier, and no scoring curve. Five correct rounds therefore produce exactly 1,000 points, and a single wrong pair does not reduce that potential — it only opens the door to a recoverable mistake.
Recoverable miss and the two-miss deadlock
Because selection edits do not count as errors, a single wrong pair is recoverable in a strict sense: the round is not closed, and the player can still submit a corrected pair. The exact rule is that the validator recognises an identical repeat submission as a repeat and does not consume another miss. Only changing at least one of the two selected positions and submitting a different wrong unordered pair records the second distinct miss and closes the run.
In practical terms, after submitting a wrong unordered pair the player has the following free options without recording a second miss:
- Change both selected positions to the correct ones and submit.
- Change one or both positions without pressing Submit, leaving the board in an "edited but uncommitted" state.
- Press Submit again with the exact same wrong unordered pair, which the validator recognises as a repeat and treats as a non-error.
The only path that ends the run is submitting a second, distinct wrong unordered pair. The first wrong pair opens that door but does not walk through it. This is also why changing a selection before submitting carries no penalty at all: until Submit is pressed, the miss counter is still whatever it was when the round started. Wrong signatures use round and sorted positions for order-independent dedupe, so the dedupe itself is what makes the repeat-on-identical rule work without an error flag.
Selection edits, invalid positions, incomplete submits, out-of-phase actions, and terminal actions are all atomic or non-errors in the implementation. None of them raise a miss, none of them raise a warning, and none of them end the run on their own.
| Action | Records a miss? |
|---|---|
| Pressing Space to start motion from Preview | No |
| Pressing Space again to finish motion during Tracking | No |
| Toggling a position with key 1, 2, 3, 4, 5, or 6 | No |
| Clicking an active position to deselect it | No |
| Trying to add a third position | No (the toggle does nothing) |
| Pressing Enter or Submit with zero or one selected position | No (the control is disabled) |
| Submitting a wrong unordered pair for the first time | Yes (one miss) |
| Submitting the identical wrong unordered pair again | No (repeat, no new miss) |
| Submitting a different wrong unordered pair | Yes (second miss, run ends) |
| Submitting a correct unordered pair | No (awards 200 points) |
| Restarting the game | No (resets score and selected positions) |
The rows above are not the result of one playthrough but the published product rule, with each row mapped to a specific action a player can perform during a session.
Score ledger, keyboard play, and reduced-motion support
The on-screen ledger reports frame count, selected count, score, and misses. None of those four numbers change while the player edits selections; all four only update on phase boundaries. The frame count advances while motion plays. The selected count follows the toggle behavior — it goes up to two and stops, it goes down to one or zero on a deselect, and a third toggle attempt does nothing. The score updates only on a successful submission. The miss counter updates only on a distinct wrong submission.
The display is also deliberately minimal. During Answer, the buttons announce only final position numbers and their selected state, so there is no visual hint about which disc was the original target. Target rings exist only during Preview; once Tracking begins, the target-ring elements are unmounted, and no target label, target data attribute, special class, color, or number follows either object. Reduced-motion styling removes the CSS transition between frames while preserving each declared state, so a player who prefers no animation still sees every fixed permutation frame in sequence.
Keyboard play covers the complete game: Space starts motion from Preview and finishes it during Tracking, keys 1 through 6 toggle final positions, and Enter submits a complete pair. Pointer and touch users receive the same start, finish, position, submit, and restart controls with large targets and visible selected states. Everything runs locally with no camera, no eye tracker, no file input, no account, no model, no external asset, no random service, and no personal data. Only the shared best completed score can remain in this browser, and Restart returns the field to the first target preview, frame zero, zero score, and no selected positions or errors. For the practical game flow, the Moving Object Memory Game guide walks through the same route from a different angle.
How to apply this rule during a real attempt
Knowing that selection edits are free changes how a player should approach the Answer phase. The natural temptation is to lock in the first guess to avoid any cost of changing, but the actual cost of changing is zero. The more useful pattern is to land on the most likely pair as soon as the motion finishes, then refine if a closer look suggests one of the two positions is wrong. There is no rush to submit, because Finish motion now moves directly to the same final frame and the automatic timer route ends at the same final frame, so the result is the same whether a player waits or speeds up.
For readers who want to verify the rules with a different mental model, the product's independent test restates every target identity, every frame permutation, and every expected final target position as literal fixtures, advances timer boundaries, checks Finish motion now, selects correct positions in reverse order to prove unordered validation, executes the full 200-through-1,000 route, repeats an identical wrong pair, changes one position to trigger deadlock, rejects invalid locations, and verifies phase and terminal atomicity. The game itself uses those fixtures, so any player can recover from one wrong pair with confidence and treat selection edits as the free tool they were always meant to be.
For other fixed visual-memory formats that follow a similar structure, the Position Memory Game, Visual Memory Test, Pattern Memory Game, and Face Memory Test each use static layouts or recognition rather than this two-target motion loop. Their tasks are different, but the same idea applies: free selection, single validator, no penalty for second-guessing. To play the actual two-target motion loop covered by these rules, head to the Multiple Object Tracking Test page and use the keyboard, pointer, or touch controls to track the two ringed discs through five deterministic six-object motion fields.
For a deeper look, see Do the Numbers Change After Restart in Number Memory Test.