Understand survivorship bias — Hinglish notes

6.2.3 · Stock-Market › Backtesting Frameworks

Survivorship Bias Kya Hai?

Statistical distortion: Agar 2010 ke 30% stocks 2020 tak bankrupt ho gaye, lekin tumhare data mein sirf 70% survivors hain, toh tumhare backtest ka average return inflate ho jaata hai. Tum ek pre-filtered "winners" set par performance measure kar rahe ho.

Survivorship Bias Kyun Hota Hai

Data vendor shortcuts: Delisted stocks maintain karna expensive hai. Bahut se providers sirf active tickers serve karte hain.
Index rebalancing: S&P 500 poor performers ko kick out kar deta hai. Historical "S&P 500 data" aksar matlab hota hai "current members ki history," na ki "jo bhi us waqt index mein tha."
Merger/acquisition erasure: Company A kharidi jaati hai; uska ticker gayab ho jaata hai. Data providers uski poori history delete kar sakte hain.

Impact Ko Derive Karo (Quantitative)

Chalte hain expected return distortion ko first principles se derive karte hain.

Setup

Time $t_{0}$ par $N$ stocks ka universe
Time $t_{1}$ tak, fraction $f$ survive karta hai average return $r_{s}$ ke saath, fraction $(1 - f)$ fail hota hai return $r_{f}$ ke saath (aksar $r_{f} \approx - 100%$ ya $- 90%$ )
True portfolio return (equal-weighted):

r_{true} = f \cdot r_{s} + (1 - f) \cdot r_{f}

Numerical Example

Diagram: Survivorship Bias Visual

Diagram do parallel universes dikhata hai: full dataset (upar) jisme saare stocks hain including failures, versus survivor-biased dataset (neeche) jahan failed stocks gayab ho jaate hain. Dekho kaise biased dataset ka average return curve systematically zyada hai.

Survivorship Bias Backtests Ko Kaise Barbad Karta Hai

Alag-Alag Strategies Par Effect

Strategy Type	Bias Impact	Kyun
Value/distressed	Severe	Struggling companies ko target karta hai—bahut saari delisted
Momentum	Moderate	Winners ko chase karta hai, lekin winners mid-run bhi fail ho sakte hain
Index replication	Low (agar point-in-time index use karo)	Lekin HIGH agar "current S&P 500 history" use karo
Small-cap	Severe	Small caps mein failure rate zyada hota hai

Common Mistakes (Steel-manning)

Survivorship Bias Ko Fix Kaise Karo

Survivorship-bias-free data use karo:
- Vendors Norgate, Sharadar (Quandl), CRSP, Compustat (delisted flags ke saath)
- delisting_return ya delisting_date fields check karo
Point-in-time universe construction:
- Har backtest date par, sirf woh stocks include karo jo us date par exist karte the aur tradable the
- Forward mat dekho ki "kya yeh stock survive karega?"
Delisting explicitly model karo:
- Jab ek stock delist ho, realistic delisting return apply karo (aksar bankruptcy ke liye -90%, merger ke liye -5% discount par)
- Position silently drop mat karo; loss absorb karo
Failure rates se cross-check karo:
- Agar tumhare backtest universe mein 10 saal mein 0% delistings hain, toh yeh biased hai. Realistic: 2-5% annual delisting rate (small caps mein zyada)

Real-World Impact: Historical Studies

Brown et al. (1992): Paya ki survivorship bias ne mutual fund returns ko 0.5-1.5% annually inflate kiya.
Elton et al. (1996): S&P 500 survivor bias estimate kiya 0.3% per year par (chhota lagta hai, 10 saal mein 3% compound ho jaata hai).
Shumway (1997): NASDAQ delisting bias: small caps ke liye 1.1% per year.

Ek backtest ke liye jo 12% annual return dikhaata hai, 1% survivorship bias ka matlab hai tumhari real strategy 11% bana sakti hai—jo tumhara Sharpe ratio risk-free benchmark se neeche le ja sakta hai.

Recall Ek 12-saal ke bacche ko samjhao

Socho tum test karna chahte ho ki lucky socks pehenne se tum video games mein achhe ho jaate ho. Toh tum 10 pro gamers se poochhhte ho, "Kya tum lucky socks pehente ho?" 8 haan kehte hain. Tum conclude karte ho: "Lucky socks kaam karta hai—80% success rate!"

Lekin ruko. Tumne sirf pros se baat ki—woh log jo sach mein bahut achhe ho gaye. Tumne un 1,000 bacchon se nahi poocha jinohne lucky socks pehne aur phir bhi har match haare. Woh bacche gaming chod gaye, toh tum unse kabhi mile hi nahin. Yahi hai survivorship bias—tum sirf winners ko dekhte ho, toh tumhe lagta hai trick hamesha kaam karti hai.

Stocks mein: agar tum apni strategy sirf un companies par test karo jo aaj bhi exist karti hain, toh tum un saari companies miss kar dete ho jo bankrupt ho gayi. Tumhari strategy amazing lagti hai kyunki usne "avoid" kar liya saare failures ko—lekin sirf isliye kyunki woh failures tumhare data se erase ho gaye the. Real life mein, tum un landmines par zaroor kadam rakhte.

Connections

6.2.01-Choose-backtesting-platform – Kuch platforms survivorship bias automatically handle karte hain (QuantConnect), kuch nahi karte (basic pandas)
6.2.04-Account-for-lookahead-bias – Survivorship bias ek tarah ka lookahead hai: tum future ki info (kaun survive kiya) past mein use kar rahe ho
6.2.05-Transaction-cost-modeling – Delisting mein aksar extra costs aate hain (illiquidity, forced sales)
4.1.02-Historical-price-data – Data quality directly bias ki presence determine karti hai
6.3.01-Sharpe-ratio – Survivorship bias Sharpe ko inflate karta hai returns badhakar aur failures ke volatility spikes chhupake

#flashcards/stock-market

Backtesting mein survivorship bias kya hota hai? :: Jab tumhare dataset mein sirf woh stocks hote hain jo "survive" kar gaye (abhi bhi trading mein hain), delisted/bankrupt companies exclude hoti hain, jisse tumhara backtest artificially achha dikhta hai.

Survivorship bias backtest returns kyun inflate karta hai?

Kyunki failed stocks (jinke paas bade losses the) data mein absent hain, toh tum sirf winners ka average return measure karte ho, na ki true portfolio average.

Survivorship bias magnitude ka formula?

Δ r = (1 - f) (r_{s} - r_{f})

jahan

f

survival rate hai,

r_{s}

survivor return hai,

r_{f}

failure return hai. Bias failure rate aur return gap ke saath scale karta hai.

Apne data mein survivorship bias detect kaise karo?

Check karo ki 10 saal mein delisting rate 0% hai (unrealistic), ya vendor docs mein "delisted securities" mention nahi. Stock count ko known historical index size se compare karo.

Point-in-time universe construction kya hai?

Har backtest date par, sirf woh stocks include karo jo us exact date par exist karte the aur tradable the—future mein index mein add hone wale ya already delist ho chuke stocks use mat karo.

Small-cap stocks ke liye realistic annual delisting rate?

3-5% per year (large caps ke ~1-2% se zyada). Agar tumhara backtest 0% dikhata hai, data biased hai.

Backtest mein delisting ko model kaise karo?

Ek realistic delisting return apply karo (e.g., bankruptcy ke liye -80%, merger ke liye -10%) delisting date par, position silently drop mat karo.

Survivorship bias se kaun si strategy types sabse zyada suffer karti hain?

Value/distressed aur small-cap strategies (high failure rate wali struggling companies ko target karti hain). Index replication suffer karta hai agar "current members' history" use karo.

Real-world survivorship bias ki magnitude?

Academic studies: funds ke liye 0.5-1.5% annual return inflation, NASDAQ small caps ke liye 1.1%/year tak. 10 saal mein 5-15% compound ho jaata hai.