Crossroad Multitasker

How it works

Two things are always on screen: a duck facing one of four cardinal directions on roads painted with flow arrows, and four corner math pods (simple multiplication, addition, or subtraction), one of which is highlighted as live. Each trial asks one of two questions - motion congruence ("does the duck's facing match the road arrows?") or quadrant math ("is the live corner's result even?") - and you answer YES or NO.

Which stream is asked switches unpredictably from trial to trial, including sometimes staying on the same stream, with a run capped at 3 in a row - so you can never settle into one task or anticipate the switch. A run is 35 trials, and the trial ends the moment you respond; your reaction time is measured from trial start.

Controls: Z or the left arrow answers YES, M or the right arrow answers NO, and there are on-screen buttons labelled to match the live question (YES MATCH / NO DIFFERENT for duck trials, YES EVEN / NO ODD for math trials).

How scoring works

A correct answer scores 100 minus your reaction time in milliseconds divided by 15, with a floor of 10 - so a 300 ms response earns 80 points, a 900 ms response 40, and anything slower than about 1.35 seconds the floor of 10. A wrong answer subtracts 40, and the running total never drops below 0.

The results screen also reports accuracy, average reaction time, and your switch cost: the average reaction time on correct switch trials minus correct repeat trials - a direct measure of how much changing tasks slows you. Your best score is stored locally.

Read the question first, always

The trap in this game is answering the stream you expect instead of the stream being asked. Because the same stimulus is always fully visible - duck and math pods together - your eyes will happily solve the wrong task. Build a rigid order: read which question is live, then look at the relevant stimulus, then answer. The half-second that ordering costs is far cheaper than a -40 for a fast answer to the wrong question.

The cap of 3 repeats in a row is worth internalizing: after three trials on one stream, the next trial is guaranteed to switch. That is the one free prediction the design allows - use it to pre-orient to the other stimulus on the fourth trial. Every other trial, assume nothing.

Precompute the cheap stream during dead time

Both judgments can be made from the standing display before the question arrives. The duck check is a single visual comparison - facing versus arrows - that takes a fraction of a second. The live corner's parity is a small computation with shortcuts: for addition and subtraction, only the parity of the operands matters (even plus even and odd plus odd are even; mixed is odd), and a product is even unless both factors are odd. You never need the actual result, only its parity.

So the strong loop is: the instant a trial resolves, refresh both reads - glance at the duck match, take the parity of the highlighted corner - and hold both one-bit answers in mind. When the next question lands, you are retrieving a prepared bit, not starting a computation. This converts a dual-task game into a lookup game and is where sub-500 ms correct responses come from.

Parity shortcuts deserve drilling on their own: "7 x 6, even" (one even factor), "13 + 9, even" (odd plus odd), "21 - 8, odd" (mixed). Saying the parity without computing 42, 22, or 13 is the whole trick - the operands' last digits tell you everything.

Manage switch cost like the metric it is

The game measures your switch cost explicitly, and the score formula makes its price concrete: every extra 150 ms of reaction time costs 10 points on a correct answer. The classic result in task-switching research is that switch trials are slower because the old task set lingers; the countermeasure is to drop the old set deliberately - after answering, do not linger on the stream you just answered. Return to neutral, refresh both reads.

Respect the asymmetry between the streams. The duck judgment is nearly free; the math judgment has real content. So when you must triage under pressure, keep the math parity fresh and let the duck read happen on demand - a stale duck read costs a quarter second to redo, a stale math read costs a full computation.

The -40 penalty against a 10-to-100 point gain sets the speed-accuracy dial: a wrong answer erases roughly half a fast correct answer plus the trial itself. Across 35 trials, 90-plus percent accuracy at a moderate pace outscores blazing speed at 80 percent. Push speed only on trials where your prepared bit is fresh and certain.

A worked example

Trial 14: the previous trial was quadrant math, and during the gap you refreshed both reads - duck facing NORTH with arrows pointing EAST (not congruent: prepared answer NO), live corner showing 7 x 6 (one even factor, so even: prepared answer YES).

The question lands: motion congruence. Retrieve the prepared bit - NO - and press M. Response in about 400 ms scores 100 - 400/15, roughly 73 points. Had you instead answered the math stream's YES out of habit, that would have been -40 and a reset of your rhythm; reading the question first is what made the prepared bit safe to fire.

Common mistakes

• Answering the stream you expect instead of reading which question is live - the fast wrong answer costs 40 points.

• Computing the corner's full result instead of just its parity - even/odd follows from the operands' parities alone.

• Lingering on the task you just answered, which is precisely the residue that inflates measured switch cost.

• Forgetting the 3-in-a-row cap - after three repeats the switch is guaranteed, and that free prediction is wasted by players not counting.

• Chasing pure speed at the cost of accuracy - the penalty structure makes 90 percent accuracy at a moderate pace beat 80 percent at speed.

Why interviews test this

Task switching is the paradigm cognitive psychology uses to measure executive control, and the switch cost this game reports is its standard metric - the reaction-time price of changing task sets. Trading firms' assessment batteries include dual-stream and switching tasks for a direct reason: desk work is exactly this, monitoring several simple things whose demands interleave unpredictably.

The skill that transfers is not the duck or the parity - it is the discipline of reading what is actually being asked before answering, and keeping cheap state refreshed so a switch retrieves an answer instead of starting a computation. That habit shows up in fast-paced interview games and in the job they screen for.