Step 1 — Sample the function at n evenly spaced points.
Chop [a,b] into n pieces, each of width
Δx=nb−a.Why this step? We can only average finitely many numbers, so we approximate by sampling n values f(x1),f(x2),…,f(xn).
Step 2 — Average those n sampled values.fˉn=n1∑i=1nf(xi)Why this step? This is just the ordinary average — the thing we already trust.
Step 3 — Sneak Δx into the formula.
From Step 1, Δx=nb−a, so n1=b−aΔx. Substitute:
fˉn=b−aΔx∑i=1nf(xi)=b−a1∑i=1nf(xi)ΔxWhy this step? That sum ∑f(xi)Δx is a Riemann sum — it's begging to become an integral.
Step 4 — Take the limit n→∞ (more and more samples = the true continuous average):
favg=n→∞limfˉn=b−a1∫abf(x)dxWhy this step? As n→∞, Δx→0 and the Riemann sum becomes the definite integral, by the definition of the integral.
Recall Try before peeking: state the formula and where it comes from
favg=b−a1∫abf(x)dx, derived as the limit of the ordinary average n1∑f(xi), which becomes a Riemann sum b−a1∑f(xi)Δx as n→∞.
Recall Feynman: explain to a 12-year-old
Imagine the bumpy water level in a wavy bathtub. The "average level" is where the water would settle if you let all the bumps flatten out — the high bits fill the low bits. To find it, you measure the total amount of water (the area/integral) and spread it evenly across the tub's length (b−a). Total water ÷ length = flat level = the average.
Dekho, finite numbers ka average toh easy hai — saare jodo aur count se divide kar do. Par ek function jaise sinx ke toh [a,b] ke beech infinite values hoti hain. Inko average kaise karein? Trick yeh hai ki integral ek "continuous sum" hota hai. Toh saari function values ko jod do (matlab ∫abfdx) aur interval ki "lambai" b−a se divide kar do. Bas, mil gaya average value: favg=b−a1∫abfdx.
Geometry mein iska matlab — ek flat rectangle jiska base [a,b] hai aur height favg hai, uska area bilkul curve ke neeche wale area ke barabar hota hai. Curve ke jo upar nikle hue hisse hain, woh neeche ke khaali gaps ko exactly bhar dete hain. Isliye yaad rakho: Average = Area Over Width.
Ek important baat: average kabhi bhi maximum nahi hota. sinx ka peak toh 1 hai par [0,π] par average sirf 2/π≈0.64 hai, kyunki function zyada time kam values par rehta hai. Aur ek common galti — sirf endpoints ka average mat nikalo (2f(a)+f(b)); yeh sirf straight line ke liye kaam karta hai, curve ke liye nahi. Hamesha integrate karo phir width se divide karo.
Physics mein yeh seedha kaam aata hai: agar velocity v(t) hai, toh average velocity =b−a1∫vdt, jo actually displacement divided by time hi hai. Aur MVT for Integrals kehta hai ki agar f continuous hai, toh kahin na kahin ek point c zaroor hoga jahan f(c) exactly average ke barabar hai — IVT ki wajah se.