How to estimate pixels by eye.
Pixel estimation sounds more mysterious than it is. You look at a space, make a guess, measure it, and let the miss teach your eye a little. Pixactly makes that loop hard to ignore: target size first, drawing second, result right after. No inspector panel to lean on.
Keep a few sizes in your head
I would not try to memorize every possible rectangle. That gets boring fast, and it does not really transfer. It is better to keep a few rough sizes in your head. For example, 24px, 48px, 100px, 200px, and 320px are enough to start.
Those numbers have a feel. 24px is close to a small icon. 48px is around the size of many comfortable touch targets. 100px is a small chunk of UI. 320px is not far from an old narrow phone width. These are not rules. They are anchors.
When Pixactly asks for a new rectangle, do not guess from nothing. Compare the target against the closest reference in your mental set. If the width is 180px, picture two 100px blocks with a little taken away. If the height is 64px, picture a 48px control plus a little breathing room.
Do width and height separately
Most misses happen because the player judges area instead of dimensions. A rectangle can feel visually similar while being too wide and too short. Before drawing, say the width and height to yourself as two separate problems. Place the width first, pause, then decide the height. This small habit prevents your eye from averaging the whole shape into a single vague size.
This is especially true on wider screens. Horizontal space feels cheap, so it is easy to overshoot width. Vertical space feels more limited, so some people draw heights too short. If your results keep showing the same pattern, trust the pattern. It is usually telling you more than one lucky good round.
Use the screen around the shape
Pixel estimation improves when you use surrounding space. Notice the distance from your starting point to the edge of the canvas, the center of the screen, and the game HUD. These landmarks help you scale the target. Designers do the same thing in interfaces: they judge a button by its relationship to nearby labels, margins, and containers.
After each round, look at the error pattern rather than only the final score. If you are often 20px too wide, your internal reference for horizontal distance is inflated. If your heights are consistently short, you may be compressing vertical space because screen height feels scarcer than width.
Try one boring drill
Play one five-round set and write down only two things: average width miss and average height miss. Ignore the total score for a minute. If width is the problem, spend the next set starting every rectangle a little narrower than your first instinct. If height is the problem, do the same thing vertically.
It sounds too simple, but it works because you are correcting one habit at a time. Trying to fix every miss at once usually turns into random guessing with more concentration.
Practice in short sets
Five focused rounds are usually better than a long distracted session. Play once, review whether width or height caused more error, then play again with one correction in mind. The goal is not to memorize Pixactly targets. The goal is to make your next visual guess slightly more calibrated than the last one.
I would stop after two or three sets. After that, most people start chasing the score instead of reading the feedback. Come back later and see whether the same bias is still there.