Stop Overtrading With This RISK Model | EWMA Volatility

Finance-specific summary

The video presents an EWMA (Exponentially Weighted Moving Average) volatility risk model designed to prevent overtrading by dynamically adjusting position sizing based on whether the market is in a low / neutral / high volatility regime.

The key idea is to predict the trading environment (risk level), not direction, using a strict no-look-ahead rule.

Instruments / tickers mentioned

• SPY (used as an example to illustrate how the regime model behaves)

Mentions (no specific tickers):

• Equities
• Gold
• Silver
• Indices
• Crypto

Methodology / step-by-step framework (EWMA → regimes → risk action)

Goal

Produce a daily stress level computed in real time (one-pass, left-to-right).

1) EWMA variance update (“memory knob”)

Track an estimate of variance σ̂² and update each day using:

• σ̂²(new) = λ · σ̂²(old) + (1 − λ) · (return_today)²

Then take the square root for volatility:

• σ̂ = sqrt(σ̂²)

Interpretation of λ:

• High λ → slow-moving estimate (trusts past more)
• Low λ → reactive/jumpy estimate

2) Convert volatility into regimes (relative thresholds)

Use a trailing window of 252 trading days and compute percentile thresholds over that window:

• Low threshold: 20th percentile of volatility
• High threshold: 80th percentile of volatility

Regime mapping:

• Below 20th percentile → low volatility regime
• Above 80th percentile → high volatility regime
• Between → neutral

Thresholds adapt over time, since the volatility baseline changes by asset type.

3) Strict no-look-ahead / honesty rule (one-day lag)

When selecting an action for day t, the system may use information only up to day t − 1.

Clean implementation: compute the regime first, then choose the action based on that regime, using the appropriate lag.

The video emphasizes that the one-day lag is essential to avoid backtest “magic.”

Action / trading rule (risk dial)

Adjust position size by volatility regime:

• High volatility: trade smaller
• Low volatility: trade normal or slightly larger
• Neutral: intermediate sizing

The recommendation framing is to adjust position size automatically based on regime, without predicting up/down direction.

Key numbers / parameters / timelines

• EWMA memory parameter: λ (no specific numeric value given)
• Trailing window length for regime thresholds: 252 trading days
• Percentile cutoffs:
  • 20th percentile = low regime threshold
  • 80th percentile = high regime threshold
• Qualitative update behavior:
  • Volatility estimate should cool down slowly but spike fast after shocks
• Timeline disclosure:
  • Next video will test the relationship across asset classes using the same model

Research question / caution addressed

The video challenges the common equity narrative that high volatility implies downside stress, and asks whether this holds for:

• Gold, Silver, Indices, Crypto

It cautions—implicitly through design choices—by using:

• No ML / no black box
• Real-time computability
• A no look-ahead rule to prevent overfitting/backtest cheating

Disclaimers

The subtitles do not explicitly include a “not financial advice” or similar disclaimer.

Presenters / sources

No presenter name or external source is explicitly stated in the subtitles.