How to compute historical volatility ?

Discussion in 'Options' started by Sekiyo, Jun 17, 2025.

  1. spy

    spy

    It was a joke... just like the results of your trading.
     
    #41     Jun 17, 2025

  2. my worst day last week beats your year.

    Just apply FFT to your realized vol estimator.
     
    #42     Jun 17, 2025
  3. spy

    spy

    Lies! Just take a look at my journal.
     
    #43     Jun 17, 2025
  4. Sekiyo

    Sekiyo

  5. Sekiyo

    Sekiyo

    Are high volatility stocks straddle underpriced ?

    Let's say the stock price is 100$,
    Option wise IV is 50% and DTE is 365 days.

    The ATF (forward) straddle IV is ~0.8 x IV or 40% (40$)

    The +1SD (compounded) is exp(+50%) or +65% (+65$)
    The -1SD (compounded) is exp(-50%) or -39% (-39$)

    The expected ±1SD range is therefore 104$
    Vs twice the ATF straddle (80$) .
    24$ or 30% difference.

    Or the compounding overstate the implied range ?
     
    Last edited: Jun 21, 2025
    #45     Jun 21, 2025
  6. Sekiyo

    Sekiyo

    [​IMG]
     
    #46     Jun 21, 2025
    spy likes this.
  7. Sekiyo

    Sekiyo

    I need to start tinkering around simulations.
     
    #47     Jun 21, 2025
    spy likes this.
  8. Interesting approach! I use a similar structure but with a slight twist to better adapt to recent regime shifts — especially when applying it within fuzzy or adaptive systems.

    Here's my HV formula setup:

    lookback = 21
    log_returns = log(close / close.shift(1))
    ema_returns = ema(log_returns, lookback)
    deviation = log_returns - ema_returns
    variance = ema(deviation ** 2, lookback)
    hv = sqrt(variance)

    What I’ve found helpful is:
    Comparing this HV to a longer-term average (say 63 days) to detect volatility compression/expansion.
    Feeding this into fuzzy rule sets or AI filters to adapt position sizing.

    Also, you're right — std deviation being centered on the mean can be misleading when returns are skewed. I sometimes adjust by anchoring around median absolute deviation (MAD) or using percentile ranges if I suspect fat tails.
    Have you tried normalizing your stdDev against rolling realized volatility bands or using entropy-based measures?

    Would love to hear how you're applying it — are you using it for entries, exits, or risk sizing?
     
    #48     Jun 21, 2025
    Sekiyo, taowave and SunTrader like this.
  9. Sekiyo

    Sekiyo

    Currently I am trying MAD such as ema(|log_returns - sma_returns|).
    It's less sensitive to outliers because we don't square the deviation.
    The lookback being DTE * 252 / 365.

    This MAD more or less correlates with the straddle.
    It gives me a 365DTE SPY straddle @ 79$
    The current SPY 362DTE 615$ straddle is ~84$
    Don't know how much volatility pts difference this is.

    We can convert the MAD to StdDev by dividing the MAD by 0.8 (± 17%)
    Actually the MAD to StdDev ratio isn't always 0.8 but that's a shortcut.

    365DTE SPY closed 49% outside its ±compounded volatility
    Same 49% without compounding volatility.
    Much less for shorter DTEs (1/3).
    Same for more volatile assets.

    You're right about comparing the measure about its history,
    I believe that's what matters in the end. Relative Value.
    But I try to get a good basis for pricing the straddle.

    However I have a 287$ straddle for 365DTE NFLX
    But the market prices it twice as much.
    Biggest deviation I've seen.
    Not seen much.

    I also try to compare the volatility against ROC 16th, 84th percentiles..
    But there is autocorrelation with the ROC so it's kind of biased right away.

    I've read probability wise ...
    The ATM straddle has 1/2% of ending OTM,
    1/3 chance of ending ITM & 1/6 of ending ATM <-> ITM

    Currently I am just tinkering.
    Simply buying Calls & Puts without delta hedging xD
    The goal would be to buy options and beat the above odds.

    Or try delta hedging.
    Trading relative volatility.

    You went farther than me. Thanks for sharing :D
     
    Last edited: Jun 22, 2025
    #49     Jun 22, 2025
  10. Sekiyo

    Sekiyo

    Well ... That's 362DTE NFLX

    Using StdDev * 0.8.
    No volatility compounding.
    upload_2025-6-22_20-43-38.png
    Volatility compounding
    upload_2025-6-22_20-43-57.png

    Using MeanAvgDeviation.
    No volatility compounding.
    upload_2025-6-22_20-44-19.png
    Volatility compounding
    upload_2025-6-22_20-44-39.png

    Still have to (re)check the code ...
    But obviously they don't compound ATForward since Put & call are symmetrical.
     
    Last edited: Jun 22, 2025
    #50     Jun 22, 2025