Two students get the same problem: a block on a ramp, a string over a pulley, a hanging weight. One of them fills three pages with sines and cosines, makes a sign error on page two, and gets a number that is wrong by a factor of two. The other writes four lines and is done. They know the same physics. The difference happened before either of them wrote an equation.
The last lesson gave you a way to make hard expressions small and manageable once you have them. This one is about the step before that: how to set a problem up so that the expressions are never hard in the first place. Three decisions do most of the work: what you draw a boundary around, which directions you call and , and which law you reach for. Get those three right and most olympiad mechanics becomes bookkeeping.
Draw the boundary first
Every equation in mechanics is about a system: a specific collection of stuff that you have mentally cut out of the world. Everything inside the boundary is the system. Everything outside can only touch the system through external forces, and only those appear in .
That word only is the whole trick. Forces between two parts of the same system come in equal and opposite pairs, so they cancel inside the sum and vanish from your equation. Move the boundary and a force that was giving you trouble can disappear.
Take two blocks touching each other on a frictionless table, kg in front and kg behind it, with a horizontal force N pushing on the front block so that both slide together. The blocks push on each other with some contact force that you do not know. If you draw the boundary around both blocks, is internal and never appears:
One line. Now suppose the contact force is what you actually want. Move the boundary so it cuts between the blocks, keeping only the front block inside. The cut turns from internal into external, which is exactly why you make it. The front block feels forward and backward from the block behind it, and it has the acceleration you just found:
Two systems, two lines, no simultaneous equations. The general rule: put the boundary around everything when you want the motion, cut through the thing you want to know when you want a force.
A force belongs on your diagram only if something outside the boundary is touching the system, or it is gravity. If you cannot name the object doing the pushing, the arrow is imaginary and it is costing you the problem.
Quick check
A rope connects two carts that are pulled along a track. You want the tension in the rope. Which system should you analyse?
Axes are free, so choose selfishly
Nothing in physics says must be horizontal. Axes are a bookkeeping choice you make, and the only rule is that they be perpendicular. So choose the pair that makes the most vectors lie flat along an axis.
A block of mass kg on a frictionless incline of angle makes the point. With horizontal and vertical axes, the normal force and the acceleration both have two components, and you end up solving a pair of coupled equations. Tilt the axes instead: let point down along the surface and point out perpendicular to it. Now the normal force is entirely along , and the acceleration is entirely along , because the block cannot burrow into the ramp or lift off it. Only the weight needs splitting.
The weight points straight down, at angle from your new axis. So its component along the surface is and its component into the surface is . Watch which one gets the sine: the piece that pushes the block along the ramp has to vanish when , and , so it is the sine. That check takes a second and saves the most common error in mechanics.
Along : , so
The mass cancelled, which it had to: every object falls the same way on a frictionless slope. Along the block does not accelerate, so the components must balance:
Less than the N weight, as it should be: the ramp only has to support the part of the weight pressing into it.
When you tilt the axes, tilt them for everything. A very common wreck is resolving the weight along the slope, then writing the normal force as anyway because that is what it was on flat ground. Every vector in the problem gets resolved in the same frame, or none do.
Let symmetry answer the question for you
If the whole setup is unchanged by some operation (a rotation, a reflection, a swap of two labels) then the answer must be unchanged by it too. That sounds abstract until you see how much arithmetic it deletes.
A point charge sits at one corner of a cube. What is the electric flux through one of the three faces that do not touch that corner?
The honest way is a surface integral over a square with a charge at an awkward angle, which is unpleasant. Instead, build symmetry into the picture. Surround the charge with eight identical cubes, stacked to make one big cube with the charge at its centre. By Gauss's law the total flux out of that big cube is . The eight little cubes are interchangeable by reflection, so each takes an eighth: through your cube.
Now look at one little cube. Three of its faces touch the charge, and the flux through those is zero because the field skims along them. The other three faces are interchangeable with each other by rotating the cube about its long diagonal, so they must split the equally, a third each:
No integral. The same reflex answers a whole family of questions in one step. Two identical charges, field at the midpoint between them: reflect the picture through that midpoint and it is unchanged, but every field vector flips, so the field must equal minus itself, so it is zero. A Wheatstone bridge whose left and right branches have the same ratio: the two midpoints sit at the same potential, so the current through the bridge resistor is zero and you can delete it from the circuit.
Quick check
A bridge has 10 ohm and 20 ohm in series on the left, the same 10 ohm and 20 ohm in series on the right, and a 50 ohm resistor joining the two midpoints. What current flows in the 50 ohm resistor?
When the ground moves under you
Some problems hide their difficulty in the choice of frame. A block rests on the frictionless face of a wedge, and the wedge is pushed horizontally so that the block rides along without sliding. What acceleration does that take?
The block has only two forces on it: gravity down and the normal force perpendicular to the face. Here tilted axes are the wrong choice, because the acceleration is horizontal, not along the slope. Use horizontal and vertical axes and let be the one that splits. Vertically the block does not accelerate, so the vertical part of carries the weight: . Horizontally, the leftover part of is the only force, so it supplies the acceleration: . Divide the second by the first and cancels:
And if the block has mass kg, the normal force follows from the vertical equation:
larger than the N weight, because the surface is doing two jobs at once.
Now the harder cousin, and this one takes practice. A block of kg slides down the frictionless face of a wedge of kg that is itself free to slide on a frictionless floor. Find the wedge's acceleration.
Choose the system generously first: block plus wedge together. The floor is frictionless and gravity is vertical, so there is no external horizontal force at all, and the horizontal momentum of the pair is conserved. Starting from rest, that gives at every instant, and differentiating, where is the wedge's acceleration. One unknown expressed in terms of the other, for free.
For the second equation you need the block alone, and the key is that the block's acceleration is measured in the ground frame while the constraint (that it stays on the face) lives in the wedge's frame. Write the block's acceleration as the wedge's acceleration plus a sliding acceleration along the face. Doing the two component equations for the block and eliminating gives the standard result
Before plugging in, test it. If the denominator blows up and : a wedge nailed to the earth does not move. If the face is flat, nothing slides, and kills . If the block falls vertically past a vertical face, pushing on nothing, and kills again. Three checks passed, so plug in:
Check the formula before you trust the number
Those three limit checks were not decoration. A symbolic answer is a machine that makes predictions, and you can test it in ten seconds against cases whose answers you already know.
Run the checks in this order. Dimensions: both sides must carry the same units, and if they do not, stop. Limits: send each variable to zero, to infinity, or to equality with another, and ask whether what comes out is what physics demands. Symmetry: if swapping two labels leaves the problem unchanged, it must leave the answer unchanged. Sign: does a positive answer point where you said positive points? Magnitude: a kg block moving at m/s after a nudge, an efficiency of , a negative kinetic energy: each one means go back.
The Atwood machine is the cleanest place to practise. Two masses on a string over an ideal pulley give and . Set : the acceleration vanishes, correct. Let : , free fall, correct. Swap the labels: changes sign, which is right because "down on the left" becomes "up on the left". For kg and kg,
and sits between the two weights, N and N, exactly as it must: the string has to hold back the heavy one and drag up the light one.
Plugging numbers in at the start throws away every one of these checks. Once has been multiplied into three other numbers you cannot send the mass to infinity and see what survives. Carry symbols to the last line, then substitute once.
Recognising which move a problem wants
The problem statement usually tells you, if you read it for structure rather than for numbers.
If two objects are joined and forced to move together (touching blocks, a string, a pulley) do the pair first, then cut. If the question asks for a tension, a contact force, or a force in a hinge, that is the signal to cut exactly there.
If the motion is constrained to a surface, tilt the axes to that surface. If instead the acceleration is along some other fixed direction, as with the driven wedge, keep the axes along the acceleration and resolve the normal force instead. The axis you refuse to tilt is the one the acceleration lies on.
If the setup has a repeated or interchangeable piece (identical charges, identical resistors on the edges of a cube, an infinite lattice that looks the same from every node) look for a symmetry or superposition argument before computing anything. Those problems are usually unfeasible by brute force and one line by symmetry, which is a strong hint about what was intended.
If the floor or the pivot is frictionless and nothing external pushes sideways, horizontal momentum is conserved and that is a free equation. If the question names a "before" and an "after" and asks nothing about the time in between, reach for a conservation law rather than for .
And when the answer is a formula, spend the ten seconds on limits. That habit catches more mistakes than any amount of care in the algebra, and it is the one thing that separates the four-line solution from the three-page one.
Next you will turn measurements into answers, which needs its own discipline: graphs and uncertainty.