Real numbers are linked by the two equations:
Determine the largest value for .
First I tried to reason by analogy, if instead of a, b, c, d, e we only had c, d, e then we can think about it geometrically in R^3. We can have e as the vertical axis, then we have a plane (first equation) and a sphere (second equation) intersecting in some sort of rotated circle in R^3 and we're trying to find the highest point on the circle, which we visually see occurs when c=d. By analogy we may expect that equality occurs when a=b=c=d, and generally when dealing with symmetric inequalities this is a good idea.
If a=b=c=d then the first equation gives us that a = b = c = d = 10 - e/2. We try to analyse the effect of deviation of the values of a, b, c, d away from 10 - e/2 (our suspected best point). So we let a = 10 - e/2 + a', and similarly for b,c,d, and we think of a' as representing a small deviation. Now we know that a' + b' + c' + d' = 0 from the first equation, this is very convenient. Plugging these into the second equation results in massive cancellation, and we're left with
(assuming I did this right), which tells us our maximum occurs when 5e - 80 = 0, or e = 16.
It's interesting to think about why symmetry causes the massive cancellation, and why performing a change of coordinates (a,b,c,d) -> (a',b',c',d') to localize our variables at our suspected maximum is a good idea.