Forums  > Trading  > Automated Delta Hedging  
     
Page 1 of 1
Display using:  

izo


Total Posts: 5
Joined: Mar 2012
 
Posted: 2016-12-01 00:56
Hi folks,

I am wondering what option market makers like Optiver et al. do to hedge their books. While they probably try to balance it across strikes and maturities/underlyings, they are probably overall short gamma.

Being short gamma in oil (which is what I mainly trade) was OK today, because the market was trending towards where the Saudis wanted it to be, but I also got it in my face sometimes.

One possibility I thought about is using trailing stop orders to avoid these random spikes on news. But it's really hard to formalize, because sometimes the market really moves and sometimes it's just bullocks on Reuters and the market moves 1 dollar and comes back. So putting any stop there will just cost you money. We had these days a couple of weeks ago, where the market moved 1.5 up and down in 30 minutes and most of the algos dropped out because they probably hit their limits and are generally not made for news driven markets.

So what do they do? Of course, I don't have access to that type of infrastructure or research resources.

sigma


Total Posts: 105
Joined: Mar 2009
 
Posted: 2016-12-01 15:37
While I do relative value vol trading with relatively small frequency, here are my thoughts

1. Execution of delta-hedge should be triggered by changes in the delta and, as a result, by changes in the gamma. There is a trade-off between wider bounds for the delta change, transaction costs and the volatility of the P&L. The best thing is to have a versatile back-tester where you can investigate different approaches and choose an optimal one. I apply the forecast of the historical volatility in this type of computations rather than using implied vols. Also, I prefer to use the statistical model for my risk as they are more stable than those computed using implied vols.

2. The position sizing matters a lot. The implied ATM one month vol on the WTI pre meeting was high at 58% but so high was the risk of being short gamma. It is prudent to have a light positioning before such events. For market makers, if they end up with a sizable inventory, they can either adjust the bid-ask spreads to balance the exposure or do cross-asset hedging where gamma is cheaper (for an example, the oil vs the long-dated USTs).

3. Strategies being short gamma and those being short vega are not equivalent. Being short vega on the WTI (say, shorting a VIX-like futures on the WTI vol) was profitable as the one month ATM vol gaped down from 58% to 40%. For vega strategies you need to look at longer dated options with a smoother gamma profile and smaller vega decay. I separate from the beginning whether a strategy is related to the gamma or vega risks and adjust the execution accordingly.

4. Being short gamma inevitably exposes you to the gap risk which is unavoidable. The two things which can mitigate the risk are the position sizing (use the long-term forecast of the volatility to mitigate the cyclicality of volatility) and the diversification (look at the diversification of the gap risk, not the conventional risk specified by the covariance matrix). In addition, it is great to have a tool to estimate your historical/stress var, where you can evaluate your positions over specific past periods and check your exposures. For multi-asset exposures, the changing dynamics of cross-moves is always an issue which is the best to analyze case by case.

5. The biggest risk for short gamma occurs during the transition period, like yesterday in the oil market, where the range-bound regime over the past few months has changed (well, most likely changed) to a trending behavior. In the range-bound regime, the best strategy for a short gamma position is to hedge infrequently while, in the trending regime, the best strategy is to over-hedge. As a result, some identification of the price/volatility regimes is necessary for position sizing and hedge execution.

6. A strategy with short gamma is negatively correlated to a momentum/trend-following strategy. The short gamma strategy with infrequent hedging benefits from the reversion to the mean which also means that the long-term volatility of the T-period returns is smaller than the short-term realized volatility of returns up to time T. In opposite, the momentum strategy benefits from the negative mean-reversion, in which case the realized volatility of T-period return is higher than the short-term realized volatility. As an alternative explanation, the short gamma strategy is long negative autocorrelation while the trend-following strategy is long positive autocorrelation. As a result, you can get better risk-adjusted returns on a portfolio with a mix of these two strategies.

ronin


Total Posts: 183
Joined: May 2006
 
Posted: 2016-12-02 15:59
@izo,

The only magic bullet for short gamma is diversification. Make theta on each symbol, pay gamma on the basket. That's it. Anybody tells you anything else, they are trying to sell something.

It's a bit hard when you just trade oil.

How to tell in real time if a move is a random spike or a real move - search for "market making" or "adverse selection". There is no magic bullet. Know your market, know what a "real move" looks like. Is your move happening because traded volume is high or because orderbook volume is low? Are some other things moving, what are they and what way are they going? Is there news causing a move? Are there trigger happy traders overtrading each piece of news? Etc.



"People say nothing's impossible, but I do nothing every day" --Winnie The Pooh

izo


Total Posts: 5
Joined: Mar 2012
 
Posted: 2016-12-02 23:47
Very interesting thoughts. Much appreciated!

Definitely had issues on 3) as I was stuck on a mostly short term vega position and when the market became skeptical about OPEC's ability to deliver and vols spiked it went against me. Decided to stick to it though doing some worst case hedges. Got out of it ok (finally cashed in today) with a few scratches here and there, but in hindsight, I'd definitely spread it differently across the time.

How to implement 6). A lot of algos and the way the market digests news leads to significant positive autocorrelation. One obvious thing that comes to my mind is buying the wings, so if you're running too far from your desired mean, you gain some long exposure on autocorrelation. Did you have anything specific in mind?

Thanks very much for your reply.

izo


Total Posts: 5
Joined: Mar 2012
 
Posted: 2016-12-03 00:18
Did you ever do some number crunching on this? I wonder if sifting through this data will lead to anything substantial, given the resources I have.

Diversification & risk mitigation is probably the more realistic option to choose.


sigma


Total Posts: 105
Joined: Mar 2009
 
Posted: 2016-12-03 01:00
>> Did you ever do some number crunching on this? I wonder if sifting through this data will lead to anything substantial, given the resources I have.

Well, the number crunching is crucial - you do get the insight about optimal (back-test) executions, risk profiles, and diversification for a portfolio of different strategies.

I have been working on these topics for a long few years now and I find that I have handled only about 50% of potential improvements for vol trading. I am really aiming at a fully automated execution starting from trade generation to risk management. There is so much things to consider starting from choosing the best model for the forecast of the volatility, the robust model for options risk, the optimal way to execute the delta hedge, the robust way to handle the dividends, the best way to estimate the expected alpha, etc.

I primarily use Bloomberg for my data feed now, while I plan to work with OptionMetrics next year as they offer more detailed data. I also keep and peruse the database of my own trades.
Previous Thread :: Next Thread 
Page 1 of 1