It was a long time I wanted to start a thread about this topic. Diagonals are really one of the most interesting combinations available to options traders, as they combine elements of both time spreads and directional spreads. I will start by writing what I know (not a lot) and then I will proceed to write what I don't know (a lot more), hoping that the good people of ET will chime in with some words of wisdom. What I know: 1) usually you go long a diagonal, short the front month and long the back month, rarely the opposite 3)if not ratioed or with the long side too much OTM, diagonals have positive vega and benefit from an increase in volatility. One caveat: if we divide volatility into ambient vol and event vol, it should be mostly the ambient vol to increase. Since event vol is usually in the front month, that would only hurt the short side. 4)diagonals have slightly positive theta, but it's usually overshadowed by changes in spot and vol, you don't really feel that much 5)call diagonals and put diagonals with same strikes and expiration are equivalent in terms of P&L graph What I don't know: 1)what are the differences in structuring the long ATM/ITM/OTM and the short ATM/ITM/OTM? Personally I would keep the long ATM and the short OTM in order to have as much initial gamma as possible, is this a correct way of thinking? 2)considering that vol behaves differently with spot going up vs spot going down, is it really true that put diagonal = call diagonal? I would expect on average a put diagonal to perform better than a call diagonal 3) what would be the rationale for ratioing a diagonal? Going vega neutral/negative and increasing theta? Creating a pitchfork of sorts? I hope this thread can rival in length the one about naked calls, please make me proud
lol, I need to add the third category: "Things I didn't even suspect existed" @jamesbp please do share
No, wait, actually I take that back. I found a white paper on this website: www.academiavox/finance/98834_43/apex_lookback
For calendar / time spreads you cannot simply add together the raw Vega for each expiry to calculate the total Vega Volatility changes by a different amount in each expiry ... say using an oversimplified example Expiry#1 / 30 DTE / Vol 16% / Position Vega ($1) Expiry#2 / 120 DTE / Vol 12% / Position Vega +$2 The nett raw Vega of position is +$1 ... you would assume that the position is Vega positive If Vol in Expiry#1 changes by +4% to 20% ... Vega loss = 4% x $1 = ($4) If Vol in Expiry#2 changes by a smaller amount ... say +2% to 14% ... Vega gain = 2% x $2 = +$4 The nett gain from Vega will be zero You should try and calculate the Vega exposure for each expiry separately ... and adjust by some factor such as square root of time ... vega buckets In the above example ... the factor I would use would Sqrt(30/120) = 0.50 Expiry#1 / Position Vega ($1) / Weighted Vega ($1) x 1.0 = ($1) Expiry#2 / Position Vega +$2 / Weighted Vega +$2 x 0.5 = +$1 The nett Weighted Vega = $0 There are some other threads on ET that also discuss this ...
Here is a study on the days to expiration, deltas and price for a diagonal put spread in SPY. The diagonal put spread buys a long term put and sells a short term put. I tested various days to expiration (DTE) combinations, delta combinations of leg 1 & leg 2, and spread yield (options spread price divided by stock price). In all, there were 126 backtests run from combining the parameters, 7 DTE parameters, 6 deltas, and 3 spread yield amounts 126=7*6*3. The Optimizer function in our backtester allows the automatic combinations of parameters of trading strategies. I tested the following days to expiration, Ideal | Minimum | Maximum with Leg2 being the short put and Leg1 the long, for example, the best average returns and Sharpe Ratio was short the 60 day long the 555 day put: I tested the following delta combinations, Ideal | Minimum | Maximum, for example, the best average returns and Sharpe Ratio was long the 10 delta short the 40 delta tied with short the 5 delta long the 40 delta: I tested the following spread yields, Ideal | Minimum | Maximum, for example, the best Sharpe Ratio was the zero priced spread (vs a credit or debit) however the best annual return was receiving a credit of 1%: Here are the results of the top strategies shorted by Sharpe in our Optimizer: The best single overall backtest in terms of Sharpe based on notional (stock price) not margin returns was: Short 45 day, 30 delta put. Long 555 day, 20 delta put. Paying 1% for the spread. Here are the monthly and yearly stats back to 2007: https://gyazo.com/07ac7fead05c603e5a31d03473a584e3 The best single overall backtest in terms of annual return was: Short 60 day, 40 delta put. Long 555 day, 10 delta put. Receiving 1% for the spread. Here are the monthly and yearly stats: https://gyazo.com/0e6b97886263203a413c11f0eadbca37 Let me know if this analysis helps.
@Matt_ORATS I don't know, the backtesting is cool, however the spread selection is a bit strange.. 60DTE vs 555DTE I would not even call that a calendar, it's much like uncorrelated positions.. I think there should a constraint on the DTE of the long related to the short. Also, does your platform allow for backtesting under specific conditions or is it simply mechanically opening a spread every X days?