What this simulation shows
This simulation builds up accuracy and precision in three steps. First you fire at an archery target whose centre stands for the true value. Next the same ideas become real repeat readings, taken with a ruler, a thermometer or a stopwatch. Last, one steel ball is measured on three instruments with very different resolutions. Every number on screen comes from the same model as the picture, so the cards, the dots and the shots always agree.
The target view
Two sliders control the shooter. Sight offset is a systematic error: a knocked sight moves every shot the same distance in the same direction. Scatter is a random error: shaky hands send each shot a different way. The orange cross marks the mean position of the shots, and the shaded circle shows their average distance from it.
Here a set counts as accurate when its mean lies inside the gold, within 4 cm of the centre, and as precise when the shots sit on average within 2 cm of their own mean. Drag the offset and the whole group slides without changing shape. Drag the scatter and the group swells or shrinks around the point the sight aims at.
Real readings
In the Readings view, each press of Read records one measurement, rounded to the step the instrument can be read to. The ruler has lost its first 2 mm, so a pencil laid against its end always reads long. The thermometer reads 1.5 °C too high. The stopwatch is stopped a moment after the toy car hits the end of its ramp.
Watch the Mean card as the readings pile up. Random scatter largely cancels, so the mean steadies, but it steadies on the dashed line, the true value plus the systematic error, and never on the green true-value line. Only finding the cause and correcting it closes that gap. The range and half the range describe the spread, and half the range gives an estimate of the uncertainty.
Resolution is not accuracy
In the Resolution view, press Measure to read a ball 12.374 mm across with a ruler (1 mm), vernier callipers (0.1 mm) and a micrometer (0.01 mm). The shaded band on each row holds every length that would produce that reading, so a finer instrument has a narrower band. Give all three the same zero error and each band moves in jumps of its own size. Every band still holds the true value plus the error, but once the error is big enough, none of them holds the true value itself. Extra digits never cancel a systematic error.
When one reading does not fit
Switch on the odd reading and reading 4 lands far from the rest, with a note saying what went wrong. Keeping it drags the mean towards it and stretches the range. Leaving it out is fine when you can state the reason, and repeating the measurement is better still. A reading should never disappear just because it spoils a result.
Your course's words
AQA and Cambridge use precise to mean that repeat readings agree closely, although Cambridge also speaks of an instrument's precision when it tells you how many decimal places to record. The Punjab Class 9 book uses the first meaning for its target picture, but it also ties precision to the instrument, so a screw gauge reading counts as more precise than a metre rule reading. It links accuracy to fractional uncertainty too. Cambridge asks you to judge a reading between two marks to the nearest half division, which this sim does for the thermometer while the ruler is read to the nearest millimetre, and it gives a range as its lowest and highest values. With the Punjab course chosen, a mean that ends exactly in 5 is rounded to the even digit, as the book does in section 1.13. Choose a course in the symbol switcher to see its own terms in the Key words box, along with where they come from.