Positional Option Trading

Positional Option Trading

Trading is a process: find edge, structure the trade, control the risk. Repeat.

When looking for ideas it is important to focus on phenomena rather than parametrizations or models

Options: A Summary

  • BSM Spits out implied vol. The best way to think of implied vol is not as a predictor but simply as changing the fast moving various option prices into something slower moving
    • It is a simplifying tool
  • Success is largely about finding situation where vol is mispriced
  • Hedging is intended to balance variance for transaction costs

The Efficient Market Hypothesis

  • Exceptions: Inefficiencies or Risk Premia
    • Inefficiencies: temporary phenomena that last only until enough people notice them, or poorly priced risk premia
    • Risk premia persists and can form the core of operations. But the profits will be less than the one off inefficiencies, which need aggressive action.

Forecasting Volatility

  • We can also distinguish trades at the tactical level (above is strategic): model driven or event driven
    • Model Driven: we have a fair value or edge and always have an opinion. relative value.
    • Event driven is a specific or unusual situation
  • Examples in blackjack: card counting is a model, ace tracking is event driven
  • Event driven are easily backtested in excel and can be very profitable. But may not often happen
  • GARCH: Many tweaks, but basically says the same thing as always:
    • best guess of vol for next N days is what it was the last N days
    • ...taking into account clustering of vol (by weighing the more recent moves)
    • ...and mean reverts in the long term (with a decay term)
    • Bottom Line: Vol tomorrow will probably be close to what it was today. Vol in long term will probably be whatever the long term average has been.
  • Best methods of predicting vol now use a variety of methods and weight htem
    • The simple 30 day historical vol is actually the best single method...

The Variance Premium

Really a combo of things

Finding Trades with Positive Expected Value


Confidence Level III

  • Implied vol term structure as a predictor of vol (contango = lower)
  • Options on fundamental factors (revisit if ever get into it)

Confidence Level II

  • Overnight effect: Index variance premium is (usually) realized overnight)
  • FOMC and volatility: VIX futures rise into and selloff after FOMC meetings
    • Buy futures two days before, sell before announcement
    • at some time sell VIX futures, cover 45 minutes after
  • Weekend effect: options decay more than expected over the weekend
  • Vol of Vol Risk Premia: High VVIX = predictor of lower VIX

Confidence Level I

Volatility Positions

  • It can be tempting to choose a shorter dated option when hedging. Don't do this.
    • two options usually cost more than one longer dated.
    • If the vol is in steep enough contango, MAY, make sense
  • Selling OTM option structures give higher median returns and higher win rates but that can make it hard to distinguish between luck and genuine positive edge
  • Highest vol premium in options that are short dated (1 week)
  • Out of the money puts most premium, out of the money calls almost none

Directional Option Trading

  • BSM is risk nuetral
  • You can add an estimate of drift and subjectively price options (better reflect the real world)
  • Still include an interest rate. Stock appreciates by one rate and discounted by risk free
  • "Best" or "optimal" is only with respect to a given criterion, trading decisions need to be based on more than just one criterion

Directional Strategy Selection

  • Overwriting: Strike determines how much edge is due to drift and how much to variance premium
  • Short put spread is the classic conservative position
  • Buying teenies makes sense
  • Bullish risk reversals with a teeny (5d) put is his favorite position to get long:
    • Potential for large wins
    • takes advantage of skew premium
  • Short implied skew trades are very difficult to make money on in practice
  • Few rules in trading but here's one: Don't buy ratios (buy 1 sell 2)
  • Variance and skew premia still most important even directionally
  • single options have best correlation between profit and prediction
  • spreads mitigate path dependency and create a stop - but too hard to do consistently (predict spot)

Trade Sizing

Kelly Criterion should form the basis of sizing
  • We can scale Kelly, and many do, but what if the edge is actually -ve?
    • need to know variance of edge, and probability we are wrong. adjust bet downward based on how uncertain
    • negative skewness also means we must adjust downward
  • Scaling is ok, but the real best way to do this is to use Kelly with a percentage stop
    • Keeps the best points of Kelly
  • Stops all have costs
    • Trailing stops worse than fixed stops. Both cost because small losers never get a chance to recover, and some would have
  • Optimal Method: Split account into "safe" and "risky" ā€”> fully kelly bet on the risky portion
    • Fixed split (not ideal)
    • Trailing percentage stop ā€”> keep safe amount constant percentage below peak
      • ex: start 20 risky 80 safe, double risky, now 40/80
      • rebalance 80 safe to be 20% below 120, or 96
      • new account is 96 safe 24 risky (77% and 23%, risky has grown as percentage)

Meta Risks

  • Real risks are things no one even has thought about (Madoff, theft, currency)

Adjustments to BSM

  • Stochastic interest rates not necessary (vol night high enough)
  • Pricing Real options using BSM framework (these are options on actual physical things, not always traded)
    • Find best correlated traded asset, and at least get an idea of value
    • It wont be good, but do the best you can, accept the variance, and ask for enough edge
  • Stochastic vols creates the convexity of the smile (vol gamma). Farther OTM options have convex payoffs to vol moves ā€”> worth more
  • Correlation between vol and spot creates slope of the smile

Rules of Thumb

Rule of Five

  • In a randomly sampled population, you can be 93% certain the median of a measure fall in between the range of 5 measurements (like vol periods)

Rule of Three

  • Probability something that never has been observed will happen: divide by length of time not seen and this gives you 95% upper limit
    • so havent seen a presidential upset in first 30 elections, chance of next time = 3/30 = 10


  • Balance between paying too much or not making any trades whatsoever
  • need to be aggressive when trading in direction of markets and patient when going against