What a thermal is shaped like.
Forget the drawing from the textbook — the neat cone rising from a ploughed field. A working thermal is a column of air whose vertical speed falls off from a core toward its edges, roughly like a bell curve, surrounded by a ring of compensating sink. It is tilted, because the wind at 1500 m is not the wind at 300 m. It is lumpy, because it is fed by pulses from the ground rather than a steady pipe. And it moves, drifting downwind at very nearly the speed of the air it lives in.
Two consequences follow, and they are the whole reason the assistant exists. First, a circle flown around a fixed point over the ground leaves the thermal within a lap or two: you must drift with it. Second, the difference between a circle centred on the core and one displaced by a hundred metres is not decoration — with a typical column it is the difference between three metres a second and two, repeated on every climb of the day. Nothing else in cross-country flying pays that well for so little effort.
The instrument has one sensor, and it is behind you.
A variometer measures the air where the glider is, now — one number, no direction. That single number is the only raw material available. So the computer does what a navigator does: it remembers. Every second it stores the vertical speed together with the position it was measured at, and it removes the wind so that a drifting thermal stays still in the picture. After one lap it holds a ring of forty or fifty lift samples arranged around your circle, and the strongest arc of that ring is the best available estimate of where the core lies.
That is what the coloured circles on the screen are: the last turn, plotted. It is why the display is empty when you arrive and useful after twenty seconds, why it is only ever as good as the circle you flew to build it, and why a sloppy, varying-bank turn produces a smeared, useless picture. You are the sensor. The instrument is the memory and the arithmetic.
One lap around an off-centre core.
Now push the delay slider. The blue ring is where the strongest lift physically is; the red ring is where it arrives in your ears and your seat, rotated 36° further round the circle by the vario filter and your own reaction time. Correct toward the red one and you will chase the thermal around the sky all afternoon. This rotation is exactly what a good assistant removes for you — it timestamps each sample and plots it where it was measured, not where you were when you noticed it.
Why the correction has to be a bank angle.
Knowing the core is 120 m to the north-west is not the same as getting there. A circling glider cannot sidestep; it can only change the radius and the position of the circle it is already flying. The tool every pilot ends up using is simple: open the bank for a few seconds when heading toward the good side, tighten it on the bad side. A shallower turn has a bigger radius, so that half of the circle bulges outward — and the centre of the whole pattern walks in the direction you wanted.
The arithmetic under it is one line: radius equals the square of your true airspeed divided by g times the tangent of the bank angle. At 100 km/h a 30° bank circles 136 m across the radius; 45° brings that to 79 m. That is the entire steering budget of a glider in a thermal, and it explains the standard advice to bank up when the lift is strong and narrow: the core of a punchy thermal can be smaller than the circle a lazy 30° turn draws around it, and you will average the whole ring rather than sit in the middle.
The average, not the needle.
The instantaneous vario in a thermal is mostly noise: gusts, stick movements, the entry and exit of every lump. The number that decides whether your correction worked is the averager — typically 20 or 30 seconds, which is to say roughly one lap. Judge a change of circle on the average two laps later, not on the needle two seconds later, and thermalling becomes a calm business of small experiments instead of a fight.
The related trap is comparing averages between different thermals while you are still inside one. Your climb average tells you how this thermal is going; MacCready theory tells you what that average is worth against the day. Both matter, and they answer different questions: centring is a local optimisation, leaving is a global one. The pilot who centres beautifully and stays too long lands out at the same field as the one who never centres at all.
The wind correction that makes it all work.
One last piece, and it is the one that separates a real assistant from a pretty animation. Every lift sample is recorded at a position over the ground, but the thermal is moving with the air mass. Plotted raw, a drifting column smears into a banana and the picture is worthless within two laps. So the computer subtracts the wind it has calculated — the same wind the wind article explains — before it plots anything, which is why the display holds still even though you and the thermal are both travelling downwind at 30 km/h.
It also quietly tells you something useful about your own flying: if the assistant looks stable and the correction it suggests keeps improving your average, your circles are good and your wind estimate is good. If it wanders, one of the two is not. Either way you have learned something the vario alone could never have told you.
The thermal assistant appears the moment you start circling — bigger red circles mark the core, with wind, airspeed and altitude alongside.
The whole climb on the biggest 80 mm display: assistant, averager and wind together, in the panel your old system already occupies.
Inertial variometer and instant wind computation from the Sense unit — the fastest, cleanest input a centring display can have.