Understand simple vs exponential moving averages

3.4.1 · Stock-Market › Indicators & Oscillators

Overview

Moving averages price data ko smooth karte hain taaki trend direction reveal ho sake, short-term noise ko filter karke. Do fundamental types—Simple Moving Average (SMA) aur Exponential Moving Average (EMA)—is baat mein alag hain ki woh historical data ko weight kaise karte hain, jisse alag responsiveness aur lag characteristics aati hain.

Core Concepts

Derivation from First Principles

SMA: Arithmetic Mean kyun?

Shuruaati sawaal: " $n$ days mein typical price" ko hum kaise represent karen?

Sabse simple unbiased estimator arithmetic mean hai. Koi bhi din preference nahi pata:

$SMA_{n} = \frac{1}{n} \sum_{i = 1}^{n} P_{i}$

Yeh step kyun? Equal weighting ka matlab hai ki kisi bhi din ki special predictive value nahi hai. Day 1 aur Day 10 ko identically treat kiya jaata hai.

Rolling mechanism: Jaise jaise time aage badhta hai, hum sabse purana price drop karte hain aur naya add karte hain: $SMA_{new} = SMA_{old} + \frac{P _{new} - P _{old}}{n}$

Yeh step kyun? Yeh computationally efficient hai—hum poori sum dobara calculate nahi karte, sirf difference se adjust karte hain.

EMA: Exponential Weights kyun?

SMA ki problem: Jab window se ek purana price drop hota hai, toh SMA "jump" karta hai kyunki poora $\frac{1}{n}$ weight achanak gayab ho jaata hai. Isse step discontinuities banti hain.

Solution concept: Fixed windows ki jagah, ek recursive formula use karo jahan har past price contribute karta hai, lekin exponentially decaying weight ke saath.

Derivation:

Desired property: Recent prices dominate karni chahiye. Maano aaj ke price ko weight $α$ milta hai, aur puri prior history (jo kal ke EMA ke roop mein summarize hai) ko weight $(1 - α)$ milta hai.

$EMA_{t} = α P_{t} + (1 - α) EMA_{t - 1}$

Yeh step kyun? Yeh aaj ke price aur kal ke smoothed value ka weighted average hai. Parameter $α \in (0, 1)$ responsiveness control karta hai.

$α$ choose karna: EMA ki "effective window" ko $n$ -period SMA se comparable banane ke liye, hum set karte hain:

$α = \frac{2}{n + 1}$

Yeh step kyun? Yeh formula exponential decay ke center of mass ko $n$ -period uniform window se match karne se aata hai. Proof:

EMA ki infinite expansion hai: $EMA_{t} = α P_{t} + α (1 - α) P_{t - 1} + α (1 - α)^{2} P_{t - 2} + \dots$

$k$ days pehle ke price par weight $α (1 - α)^{k}$ hai. Yeh weights 1 tak sum hote hain (geometric series): $\sum_{k = 0}^{\infty} α (1 - α)^{k} = \frac{α}{1 - ( 1 - α )} = \frac{α}{=} 1 ✓$

Expected age (center of mass) hai: $Age = \sum_{k = 0}^{\infty} k \cdot α (1 - α)^{k} = \frac{1 - α}{α}$

$n$ -period SMA ke liye, average age $\frac{n - 1}{2}$ hai. Inhe equal set karte hain: $\frac{1 - α}{α} = \frac{n - 1}{2}$

$α$ ke liye solve karte hain: $2 (1 - α) = α (n - 1)$ $2 - 2 α = α n - α$ $2 = α n + α$ $α = \frac{2}{n + 1}$

Yeh step kyun? Yeh ensure karta hai ki EMA $(n)$ aur SMA $(n)$ ke similar "lookback periods" hon, lekin EMA faster respond karta hai kyunki woh purana data kabhi poori tarah discard nahi karta—bas diminish karta hai.

Visual Comparison

Figure — Understand simple vs exponential moving averages

Key observations:

Sharp moves ke dauran SMA zyada lag karta hai (SMA trace mein straight line segments)
EMA smoothly curve karta hai aur current price ko zyada closely track karta hai
Stable trends ke dauran dono converge ho jaate hain

Worked Examples

Worked example Example 2: 3-Period EMA Calculate Karna

Given: Same prices. Pehle $α$ nikalo: $α = \frac{2}{3 + 1} = 0.5$ Initialization: Shuruwaat ke EMA ke roop mein Day 1-3 SMA use karo: $EMA_{3} = 102.33 (Example 1 se)$ Yeh step kyun? EMA ko ek seed value chahiye. Common practice pehla SMA use karna hai.

Day 4 calculation: $EMA_{4} = 0.5 \times 103 + 0.5 \times 102.33 = 51.5 + 51.165 = 102.665$ Yeh step kyun? Hum recursive formula apply kar rahe hain: naye price par aadha weight, prior EMA par aadha.

Compare karo: SMA 103.33 par jump kiya, lekin EMA sirf 102.67 tak gaya. Kyun? EMA abhi bhi Day 1 ka $100 "yaad rakhta hai" prior EMA value ke through, jabki SMA ne usse completely bhula diya.

**Day 5 price = $106:** $$\text{EMA}_5 = 0.5 \times 106 + 0.5 \times 102.665 = 53 + 51.33 = 104.33$$ **Yeh step kyun?** Same recursion. Dekho kaise Day 1 ka$ 100 abhi bhi influence kar raha hai (102.665 term mein chhupa hua), effective weight $(1 - α)^{4} = 0. 5^{4} = 0.0625$ ke saath.

Common Mistakes

Practical Applications

SMA kab use karein:

Major support/resistance identify karna: 200-day SMA widely dekha jaata hai; iska lag self-fulfilling prophecies create karta hai kyunki institutions ise use karte hain.
Noise filter karna: Volatile stocks mein, SMA ki smoothness overtrading se bachati hai.
Long-term trend confirmation: Lag acceptable hai jab tum confirm karna chahte ho ki trend establish ho gaya hai.

EMA kab use karein:

Short-term trading: MACD indicator mein 12/26-day EMAs momentum shifts jaldi pakad lete hain.
Trending markets: EMA price action ke saath closely rehta hai, winning trades mein zyada der tak rakho.
Dynamic support/resistance: EMAs "moving floors/ceilings" ki tarah kaam karte hain jo faster adapt karte hain.

Combinations:

Golden Cross: 50-day SMA ka 200-day MA ke upar cross karna (bullish). False signals se bachne ke liye SMA use karta hai.
MACD (Moving Average Convergence Divergence): Momentum changes spot karne ke liye 12-day aur 26-day EMAs use karta hai.

Connections 3.4.02-Moving-average-crossovers-and-signals – Buy/sell signals kaise generate karein

3.4.05-MACD-indicator – Momentum ke liye EMA difference use karta hai
3.4.08-Bollinger-Bands – SMA ± standard deviations par built hai
2.3.03-Support-and-resistance-levels – Moving averages dynamic S/R ki tarah kaam karte hain
3.2.04-Lag-vs-responsiveness-tradeoff – Saari smoothing mein fundamental tension

Feynman Recall

Recall Ek 12-saal ke bachche ko samjhao

Socho tum figure out karna chahte ho ki bahar garmi ho rahi hai ya thandi, lekin temperature minute-to-minute kaafi jump karta hai. Tum yeh kar sakte ho: Method 1 (SMA): Har ghante, pichle 10 minutes average karo. 1:00 baje, 12:50-1:00 average karo. 1:01 baje, 12:50 drop karo, 1:01 add karo, aur naye 10 minutes average karo. Har ek equal count karta hai.

Method 2 (EMA): Purane minutes completely bhulne ki jagah, tum kaho "is minute ka temperature important hai, lekin jo average maine already calculate ki hai woh bhi hai." Tum ek naya average banate ho jo aadha aaj ka temperature hai aur aadha kal ka average. Kal ke average mein secretly pichle din bhi shaamil hai, aur yeh silsila chalta rehta hai. Toh purane temperatures dheere dheere fade hote hain jaise ek echo, achanak gayab hone ki jagah.

Difference: Method 1 (SMA) dheere react karta hai lekin zyada stable hai. Method 2 (EMA) tezi se react karta hai lekin jumpy ho sakta hai. Agar sach mein garmi ho rahi hai, toh EMA jaldi figure out kar leta hai. Agar koi weird minute hai, toh SMA use better ignore karta hai.

Stocks ke liye, hum jaanna chahte hain: "Kya price sach mein upar ja raha hai, ya sirf random noise hai?" Dono methods help karte hain, lekin tum choose karte ho based on iske ki tum fast (EMA) rehna chahte ho ya careful (SMA).

Mnemonic

Formulas at a Glance

Active Recall Flashcards

#flashcards/stock-market

n-period Simple Moving Average ka formula kya hai? :: SMA_n = (P_1 + P_2 + ... + P_n) / n, jahan har price ka equal weight 1/n hota hai.

Exponential Moving Average ka recursive formula kya hai?

EMA_today = α·P_today + (1-α)·EMA_yesterday, jahan α = 2/(n+1).

SMA mein "step discontinuities" kyun aati hain jab ek price drop hoti hai?

Kyunki jab ek purani price window se bahar jaati hai, uska poora 1/n weight achanak gayab ho jaata hai, jisse average mein sudden jump aata hai.

SMA aur EMA mein purane data ko treat karne ka key difference kya hai?

SMA saare n prices ko equal weight deta hai aur purana data completely discard karta hai. EMA saare historical data ko exponentially decaying weight deta hai—purane prices kabhi poori tarah gayab nahi hote.

EMA smoothing factor α ka formula kya hai?

α = 2/(n+1). Yeh EMA ka effective lookback period n-period SMA se comparable banata hai.

EMA ko pehle price ki jagah pehle n-period SMA se initialize kyun karte hain?

P_1 se shuru karne par massive initialization bias aati hai. Agar early prices unrepresentative hoon, toh EMA ko converge hone mein kai periods lagte hain (1-α) ke compounding effect ki wajah se.

Kis market condition mein EMA, SMA se zyada false signals deta hai?

Sideways ya choppy markets mein, EMA ki higher responsiveness zyada whipsaws (false breakout signals) produce karti hai kyunki woh short-term noise par react karta hai.

n-period EMA ka "effective age" ya center of mass kya hai?

(1-α)/α = (n-1)/2, jo n-period SMA ki average age se match karta hai. Yeh α = 2/(n+1) formula justify karta hai.

Agar EMA ke liye α = 0.1 hai, toh aaj ke price ko kitna percentage weight milta hai?

10%. Baaki 90% prior EMA value hai, jo recursively saare historical prices embed karti hai.

Kya hota hai SMA mein jab window ka sabse purana price naye price ke barabar hota hai?

SMA unchanged rehta hai kyunki drop aur add cancel out ho jaate hain: SMA_new = SMA_old + (P_new - P_old)/n = SMA_old jab P_new = P_old.

20-period EMA ke liye, 20 din pehle ke price ka approximately kitna weight hota hai?

α(1-α)^20 ≈ 0.095 × (0.905)^20 ≈ 0.095 × 0.134 ≈ 1.3%. Abhi bhi present hai lekin heavily diminished.