Hi, Im doing a dissertation about gas storage valuation. I have some problems, so if anybode believes that can help me, please answer

Ask your question and I might help you.

Hi,Currently Im doing about Spot based model valuation. But I want try to do a rolling intrinsic forward valuation, and compare both models.Unfortunately, I dont have time to write a new code, and Im searching Matlab code to do both the multifactor forward curve simulations and the optmization.Concretely, I want to replicate Maragos (2002), or Gray and Kandewhal (2004) valuation method.Thank yoy very much

Hello rafaelvivo,Apologies if my comments are naive or uninformed, as they are probably both. I used to do a bit of valuation of gas storage. Typically, I used a variety of approaches, but none were really very "good" in my opinion. A company called Finance Engineering Associates has a software package that does exactly what you are talking about. No idea in what language it's written, but I think C++ and the modules are used in excel. Approaches I used were:(1) Spot based - pretty crappy imo due to not being able to specify the spot process well (natgas is weird) and that, again, imo, the values it spits out mean nothing. If you believe the process, I guess it spits out an expectation, but no manager will want to hear you explain how your purchased a gas storage asset with a great edge only to show a large red pnl at the end of the year.(2) Rolling intrinsic - again, another crappy approach for pretty much the same reasons. Also, it requires forward month correlations, where the stability of these parameters are dubious at best. Each maturity on the curve was modeled as a separate GBM with correlated random factors. Good thing is that the forward curve is pretty transparent (at least a couple of years out). I'm pretty sure that this is the way most traders trade the asset though.(3) Spread option - probably the best of the three, but (a) requires a real implied volatility surface and (b) uses the prices of calendar spread options between maturities (which require correlation for pricing) which don't trade regularly (as far as I know). I liked this best because it is the only one out of the three that gives you any solid hedging advice in terms of your second order risks, vega/gamma etc. The third gives you bucketed delta, vega, and gamma expsosures as well as your cross gammas and correlation risks (essentially the greeks from "owning" a bunch of calendar spreads on gas). This, to me, however, is the most intuitive approach because, in the presence of a deep and liquid market for calendar spread options, you can essentially hedge away all your risks (for the most part. If it blows up or leaks gas, you're still kinda screwed).FEA has some white papers that show the expectation of PnL from (2) and (3) are equivalent under some assumptions, but the variance in PnL is higher for (2). I would suspect that (3) < (2) < (1) in terms of PnL variance, but I don't know this for a fact. Furthermore, all are pretty weak because of what I see as mis-specified stochastic processes for natty (is a standard GBM a good process to model natty? Does it exhibit mean reversion? etc etc). All three also use some form of optimization to figure out what the best "plan" is based on the curve. Though I am far from being a quant, I have numerous reservations about nat gas storage valuation. I have since moved on to another group at my company, so I luckily don't have to deal with these. If you give FEA a shout, they may share some info and/or even let you "borrow" their code. Being cited in an academic paper may be good for business. I have not read the papers you're referring to, but those guys probably treat the subject with much more mathematical rigor than I ever have.Hope that helps. Would be happy to chat about this topic a bit more offline if you wish.

I cannot say I am fan of FEA’s software, but I think jd is correct about the C++ base.However, they do have a few ads that detail some of the process modeling, but I do not think that they layout a storage valuation model.The coding does not have to be as hard as it might seem; so if having to code it yourself is a worst case scenario you will be fine. Also, there is a paper titled “Commodity Storage Valuation” that appeared in Energy Economics and is a bit messy to implement, but it could be useful to you.

I totally disagree with jd1123.Spread Option is the worst method to price a storage as it does not take into account the dynamicity of the problem. The only good idea is to be able to statically replicate your portfolio with american spread options that you can find on the market. Pb : the market does not trade these options and you would be lucky to find vamilla calls and puts (both in Europe or in the US). It seriously undervalues the storage and I wouldn't hedge on it.Rolling Intrinsic is not bad and that what most traders do in practise. The problem is that it is a strategy that you cannot hedge.Eventually, Spot optimisation is the best because it derives from the SDE which is the proper mathematical way to consider the storage (like american option pricing). You can derive deltas and gammas (though not done in FEA) and hedge on it.The problem is that prices and sensitivities are often long to compute (so you need good quants and IT quants to optimize the code) and traders don't like it because they don't understand the sophistication of the model.

Hi to all,First, sorry if my comments are unfortunated, but I´m just a beginner in this world, and I dont have work experience.Commoquant, I don´t understand when you say that rolling intrinsic is an unhedge strategy. Precisely, during the life of the storage contract, you create value winding and unwinding forward positions, so you are always covered. The rolling intrinsic is only an extension of the initial intrinsic value, which is only the value of taking forward postions at the beginning of the contract taking into account the restrictions, a totally hedge position. Sorry if Im wrong.Second place, and very important for me because is the central part of my dissertation. Spot based models are based on day ahead gas price. For example, a MonteCarlo model is based on the evolution of the Spot price, so every day you take the decision of inject, withdraw or do nothing depending of the daily Spot evolution, so you are totally unhedge. If Im wrong, How can I create the hedge?Thanks

If you compute the price of the storage via MTC simulations by applying Rolling intrinsic, you are not able to hedge this price. The only thing you are sure to hedge with RI is the initial profile.You can buy DA, BOM and futures contract on the market (at least in Europe) so you can hedge tomorrow's injection / withdrawal price and the sensitivity of your instrument wrt the fwd curve (BOM and the other remaining month). That way you can hedge the greeks of the Spot Optimisation.View it like a bermudan option on IR : you will be using the whole forward curve to hedge incoming exercises.

QuoteOriginally posted by: commoquantI totally disagree with jd1123.Spread Option is the worst method to price a storage as it does not take into account the dynamicity of the problem. The only good idea is to be able to statically replicate your portfolio with american spread options that you can find on the market. Pb : the market does not trade these options and you would be lucky to find vamilla calls and puts (both in Europe or in the US). It seriously undervalues the storage and I wouldn't hedge on it.Rolling Intrinsic is not bad and that what most traders do in practise. The problem is that it is a strategy that you cannot hedge.Eventually, Spot optimisation is the best because it derives from the SDE which is the proper mathematical way to consider the storage (like american option pricing). You can derive deltas and gammas (though not done in FEA) and hedge on it.The problem is that prices and sensitivities are often long to compute (so you need good quants and IT quants to optimize the code) and traders don't like it because they don't understand the sophistication of the model.commoquant,I like your comments, but what specifically do you disagree with? My choice of the spread option model as my favorite?The reason I like the spread option model the best is that it is the only one that is remotely hedgeable. Yes, the inability to capture the dyamicity of the storage field is a HUGE problem, one of the main complaints I have about it. But the other approaches suppose a distribution on the price of natural gas that is not at all hedgeable. This is why I don't trust spot models at all, regardless of how scientific, proprietary and mathematically sound your spot process may be (unless you have a model that gives you reliable hedge ratios). The sad part, I guess, is that these all better approximate the "true value" of the lease than the spread option model does. But, using the (and pardon my french) "oh shit" approach to risk management, if spread volatility goes to very low levels (perhaps to those that were more realistic when I was in grade school), you're stuck long volatility without a hedge (unless of course you just sell volatility against it in some way). Another problem with the spread option model is that it does not capture value related to any spot operations (selling expensive gas due to supply problems, or buying cheap gas during OFO's etc etc) without tinkering with the vol. If you do that, you don't have any remotely reliable hedge. This is all non-hedgeable "probabilistic" value that is certainly worth more than zero. When I was valuing storage, I had huge problems with ALL the standard approaches. At the time, I was very junior in my company, and so nothing I said was taken too seriously. But I guess, this is a spot where someone like rafaelvivo can make some serious contributions to the field. I know that there are LOTS of proprietary approaches to valuing storage, all of those that I know of I cannot comment about. I find this topic incredibly interesting, but no more babbling

No, please continue to babble

QuoteOriginally posted by: jd1123But the other approaches suppose a distribution on the price of natural gas that is not at all hedgeable. This is why I don't trust spot models at all, regardless of how scientific, proprietary and mathematically sound your spot process may beunless you have a model that gives you reliable hedge ratios This is precisely the issue. The problem (which has not yet been solved yet) is how to compute good greeks (and fast?) with american options via Monte Carlo? (Monte Carlo beacuse you may want a Levy-driven spot model and multi factor can also be useful).

Good morning from Spain!!Today I´ve understood one thing, I need to work at least a couple years in a energy company to understand your comments about hedge. Unfortunately, the literature about gas storage valuation is quite poor, in the sense that is not a very academic task. Most of the papers are only suggestions about the valuation, but 70% of the papers have 4 ó 5 pages.commoquant, you have the reason. Dejong and Walet (2004) "To store or not to store" calculate the delta greek of the strategy. Like others papers, they only comment just two things about it, so, despite I use their method to calculate the storage value, I dont know how calculate the delta.On the other hand, If Im not wrong, Gray and Kandewhal make the hedge under the rolling intrinsic and basket spread. please have a look.I´ve heard FEA. Also, I contacted with an important ex-employee of this company, but he didn´t aswer me...thanks for this comments, its a nice conversation, at least the things I understand!

hi, if you've got the papers in pdf would you mind sending them to me if i give you my e-mail?

the article is at:http://www.maycroft.com/resources/Artic ... 202003.pdf

hi,gray and khandelwal link is: http://www.fea.com/resources/articles.asp