← All guides Home

Sharp Eye Tips: Judging Midpoints, Angles and Centres by Eye

Updated October 2026

Sharp Eye gives you ten shapes and asks for one point on each, with no ruler and no grid. It feels like pure talent, but the scoring is simple and very predictable, and once you know exactly what is being measured you can spend your attention where the points are. This guide explains how each round is marked, which of the six tasks reward which technique, and a short routine for finding and correcting your own habitual errors.

How a round is scored

Every round is worth between 0 and 100 points, and a full game is ten rounds, so the ceiling is 1,000. The six task types are dealt so that each one appears at least once per game; the remaining four rounds are picked at random from the same six, which means some types come up twice or even three times.

The score falls in a straight line with your error. On the four "place a point" tasks the error is the distance between your mark and the true answer, divided by a reference length taken from the shape itself. You lose about 8.3 points for every 1 percent of that reference length, and you reach zero at 12 percent. On the two angle tasks only direction counts: you lose about 8.3 points per degree and reach zero at 12 degrees. A mark that is 3 percent or 3 degrees off is therefore worth 75, and one that is 6 off is worth 50.

The reference lengths matter, because they tell you which mistakes are cheap:

Because everything is measured relative to the shape, and the shapes are sized from the shorter side of your board, a small phone and a tablet produce the same score for the same proportional error. Screen size is not an advantage.

Aiming without your finger in the way

Nothing is locked in when you touch the board. While your finger is down you are only moving a crosshair; the point is committed the moment you lift. On a touch screen the crosshair sits 46 pixels above your fingertip, so the target stays visible. With a mouse there is no offset and the cross sits right under the pointer.

Two things follow from this. First, there is no clock: you can hold, slide, look again and adjust for as long as you like before releasing, and nothing in the code rewards speed. Second, the offset needs a moment to get used to. Put your finger down a little below where you want the point, watch the cross rather than your finger, then make small slides. Most early misses come from lifting too soon, not from bad judgement. If the system interrupts your touch (a notification gesture, for example), the aim is simply dropped and you can start again; a cancelled touch is not scored.

After you release, the true answer is drawn in green next to your yellow mark along with the points earned, coloured green for 80 or more, yellow for 40 to 79 and red below 40. The next shape appears about 1.3 seconds later, so that is your window to study the miss.

The four point tasks

Midpoint of a line. The line is always at least 45 percent as long as the shorter side of the board, and both ends are marked with dots. Rather than staring at the middle, compare the two halves of your proposed split: slide the cross until the left part and the right part look like the same length, then check once along the line's direction for any sideways drift. On a typical phone in portrait the shortest possible line is about 160 pixels, so 80 points on it means landing within roughly 4 pixels.

Centre of a circle. Circles have a radius between 20 and 36 percent of the shorter side of the board. Judge one axis at a time. Find the horizontal position where the gaps to the left and right edges of the ring look equal, hold that, then slide vertically until the top and bottom gaps match. Because the score uses the diameter, a large circle is the most forgiving task in the game; aim for 90 or more on it.

Fourth corner of a parallelogram. You see three corners joined by two sides, with the middle corner shared by both sides. The missing corner sits opposite that middle corner. The cleanest construction is to imagine the side from the middle corner to one end, then picture an identical copy of it starting from the other end, same length, same tilt. The far end of that copy is the answer. Check it the other way round as well: the second drawn side, copied from the first end, should arrive at the same spot. If your two mental copies disagree, split the difference. The game never deals very flat shapes; the angle at the middle corner is always between 30 and 150 degrees, and both sides are at least 30 percent of the board's shorter side.

Centre of a triangle. The target is the centroid, the average of the three corners. The construction to remember is the median: a line from a corner to the midpoint of the opposite side. The centroid lies on every median, one third of the way up from the side and two thirds of the way from the corner. Pick the median that looks easiest, usually the one from the sharpest corner, locate the middle of the opposite side first, then come back one third of the distance toward the corner. A common mistake is to aim for the centre of the largest circle that fits inside, which drifts toward the widest part of the shape. The game avoids skinny triangles: every side is at least 30 percent of the board's shorter side and the area has a minimum too.

The two angle tasks

Both angle tasks are scored only by direction from the corner dot, measured from the corner to your crosshair. How far out you place the cross makes no difference to the score, but it makes a big difference to your control. A one-pixel wobble close to the corner swings the direction by several degrees, while the same wobble far out barely moves it. Always aim near or beyond the end of the drawn arms. Releasing within 4 pixels of the corner itself scores zero, because no direction can be read from that.

Splitting an angle in half. Two arms leave the corner with an opening somewhere between 40 and 150 degrees. A dashed arm follows your aim. Look at the two wedges either side of the dashed arm and make them look equal in area, not just at the tips. Wide openings tend to be harder than narrow ones, because the arms end far apart and the two wedges are harder to compare at a glance; slow down on anything wider than a right angle.

Making a right angle. One arm is drawn and you aim a second, dashed arm to stand square to it. The surest reference is the edge of your screen. If the drawn arm is close to horizontal, the answer is close to vertical, and the difference between the arm and the screen's horizontal edge is exactly the tilt you need to copy onto the answer. For a steeply tilted arm, picture a small square sitting in the corner of the angle and aim along its other edge.

Either perpendicular counts. With only one arm drawn, a quarter turn to the left and a quarter turn to the right are both right angles, and the round is scored against whichever of the two your dashed arm is closer to; the green answer line is drawn on that side. So pick whichever direction has more room on the board, or whichever side your eye finds easier to square up, and spend your attention on the tilt rather than on guessing a side.

Finding your personal bias

Your total over a game is not decided by your best rounds but by consistent small errors repeated across all ten. Most people have a systematic lean: midpoints a touch toward one end, triangle centres too high, bisectors pulled toward the longer-looking arm. You can only fix a bias once you have seen it, and the 1.3-second reveal after each round shows you the green answer right next to your yellow mark.

  1. Play three games purely to observe. After each release, note in which direction the green dot lies from your mark, and say it to yourself, for example "too far right" or "rotated too far".
  2. Look at the breakdown on the final card. It lists each round with its task symbol and score, so you can see whether one task type is dragging your total down.
  3. For the next few games, deliberately nudge every mark of your weakest type a little against your usual lean, then check whether the gap closes.
  4. Once a correction feels natural, move on to the next weakest task.

A realistic target

Since a 6 percent or 6 degree error already costs half the round, steady 70s across the board add up to around 700 and are a strong result. Getting beyond that is about removing the occasional disaster round, typically a hurried triangle or a wide bisector, rather than polishing the circle you already hit well. If you rotate the device while aiming, the current shape is kept and simply rescaled to the new board, so rotating never deals you a different shape and costs nothing in score. Your best total is saved on the device, and any score above zero can be sent to the Hall of Fame from the final card.

▶ Play Sharp Eye Ten shapes, one point each, judged by eye alone.