Serving the Quantitative Finance Community

 
User avatar
davidhigh
Topic Author
Posts: 0
Joined: March 2nd, 2015, 9:09 pm

Trading with pricing models

March 5th, 2015, 9:43 am

This is probably some old stuff for some, but it is a question which I couldn't definitely answer by myself in the last two years, so your comments are welcome.Derivatives of any kind are usually described via stochastic models and managed in practice by standard Monte Carlo and PDE methods (my base is in the topic of Interest Rate Derivatives, but no need to focus on these). These methods have their difficulties but work well for many models ... no big deal, as most of the models used in practice are one- or two-factor models anyways.Now, let's assume you had available an astronomical solution method, which gives you the exact solution of any stochastic model (for an arbitrary number of factors) and that in milli-seconds. Would there be any way to use this god-solver for trading? How would that be and in which topics?Thanks in advance, David
Last edited by davidhigh on March 4th, 2015, 11:00 pm, edited 1 time in total.
 
User avatar
Alan
Posts: 2958
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Trading with pricing models

March 5th, 2015, 2:53 pm

Become a market maker at the CBOE.
 
User avatar
savr
Posts: 6
Joined: January 21st, 2013, 3:28 pm

Trading with pricing models

March 5th, 2015, 5:00 pm

Quote?What good are computers? They can only give you answers.? Tell me what your model describes, and perhaps someone can tell you how to use it.On the most absurdly abstract level, I suppose you'd be able to back out what model each participant is using (assuming they could reply to your rapid queries with similar speed). That would still not make you able to exploit them unless you can predict their changes in modelling, not to mention their once-in-a-while big fat mistakes.Now stop procrastinating and get back to work on that solver!
 
User avatar
davidhigh
Topic Author
Posts: 0
Joined: March 2nd, 2015, 9:09 pm

Trading with pricing models

March 5th, 2015, 8:04 pm

QuoteOriginally posted by: AlanBecome a market maker at the CBOE.This is anyways the next plan. Do you have any other ideas? :-)QuoteOriginally posted by: savrTell me what your model describes, and perhaps someone can tell you how to use it.I was talking about a solver for stochastic models. Let's again say one could solve any stochastic model fast and with high-accuracy. Example: the accurate evaluation prices or sensitivities in the Libor-Market-Model with Monte-Carlo is not fast and not very accurate. Say now you, and exclusively you, can solve it very fast and very accurate, or include another 10 parameters and do the same. Is there some way to make money (by trading, not by selling the solver ;-)?I basically know how banks determine their prices. If each of them uses a simple one-factor model for highly structured options, well, then the price of this option is what they get out of their calculations (of course plus some further costs, but leave them aside). Let us assume this simple one-factor model is all they have available. Now comes me and uses the god-solver to price this same option by some "heavy" model. Is there any advantage which I can get of this? For instance, could I find the option is sold under value?Another similar example: let's assume each bank uses a "heavy" model (--sorry for this distinction in "heavy" and "simple" models, I hope you know what I mean ... I hope I do). But now, let's assume they can solve it only with a small accuracy. Entrance the god-solver. He's able to calculates it all much more accurately. Same question: is there any advantage that could be drawn from this?These questions probably show I have little idea of the market or economy ... I'm basically a physicist and only slowly get a grip on finance. So thanks again in advance for your help.QuoteOriginally posted by: savr Now stop procrastinating and get back to work on that solver!Yes, sir!
 
User avatar
Alan
Posts: 2958
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Trading with pricing models

March 6th, 2015, 1:08 pm

QuoteOriginally posted by: davidhighQuoteOriginally posted by: AlanBecome a market maker at the CBOE.This is anyways the next plan. Do you have any other ideas? :-)Learn about high frequency trading.Solving a model quickly is only helpful when- the model contains a genuine insight, an "edge", and- speed mattersThat's why I mention market making and HFT. It is up to you to develop the edge.
 
User avatar
Traden4Alpha
Posts: 3300
Joined: September 20th, 2002, 8:30 pm

Trading with pricing models

March 6th, 2015, 3:10 pm

QuoteOriginally posted by: AlanQuoteOriginally posted by: davidhighQuoteOriginally posted by: AlanBecome a market maker at the CBOE.This is anyways the next plan. Do you have any other ideas? :-)Learn about high frequency trading.Solving a model quickly is only helpful when- the model contains a genuine insight, an "edge", and- speed mattersThat's why I mention market making and HFT. It is up to you to develop the edge.Alan's approach is the right way to think about this.Yet there is another wrinkle to consider. The value of a zero-latency solution is modulated by the timeliness of the data going into the model and the timeliness of actions arising from the model. If you use old data in the model, you'll get a result that is identical to that produced by using new data in a slow model. Likewise if there are delays between the estimate of price and transactions using that price, then the transaction would be old. A god-solver gets rid of one type of latency (the data->price latency) but does not eliminate other types of latency such as (counterparty -> data) and (price->trade). Yes, there is an edge to being the lowest latency trader, but that edge might be attenuated by the presence of other latencies in the system.I would also start pondering whether the stochastic model used by the god-solver is valid at all time scales or in markets with high-frequency traders. If one is not careful, one might find that one is quickly calculating wrong answers or that other players in the system have learned that they can manipulate a god-solver trader by inducing very short-term shifts in the data that induce the god-solver trader to buy or sell in predictable ways.
 
User avatar
Cuchulainn
Posts: 20254
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Trading with pricing models

March 7th, 2015, 9:21 am

QuoteNow, let's assume you had available an astronomical solution method, which gives you the exact solution of any stochastic model (for an arbitrary number of factors) and that in milli-seconds. Like a closed solution to the stochastic Navier-Stokes equation?
Last edited by Cuchulainn on March 6th, 2015, 11:00 pm, edited 1 time in total.
 
User avatar
davidhigh
Topic Author
Posts: 0
Joined: March 2nd, 2015, 9:09 pm

Trading with pricing models

March 8th, 2015, 1:20 pm

Thank you for your answers, I took some time to think about it and now have the feeling I'm beginning to understand what you mean.(i) When I asked the question, I had more traditional markets in mind (like for swaptions, where HFT isn't important afaik). Here, I guess "making money" always corresponds in a way to "seeing better in the future". That is, it doesn't matter whether a model is much more complicated, but only whether it better describes the conditions into which the market is going to evolve (... and the same holds for the set of free parameters you need to adjust in any model). This is what Alan means by "insight", I guess ... and this seems not directly related to a "god-solver", as it's not about the solution but rather about the model. (ii) The benefits of the speed component of the "god-solver" is more or less obvious -- when speed matters (as in HFT), being the fastest is a good thing. However, as before, this doesn't mean you can better look into the future. So again one needs a valid strategy, and there seems to be no cases where you directly gain just because you have an astronomical solution method (at least unless you supply it with an "astronomical trading strategy").When I asked the question my hope was more or less that there are such situations in the market where computational power -- and only computational power --matters. Now, after the input by you and some thoughts, I'd rather think that this is never the case -- great computational power doesn't relieve you from using a good trading strategy. Conversely, good strategies might give you an edge, and possibly these strategies might require heavy computations.QuoteOriginally posted by: CuchulainnQuoteNow, let's assume you had available an astronomical solution method, which gives you the exact solution of any stochastic model (for an arbitrary number of factors) and that in milli-seconds. Like a closed solution to the stochastic Navier-Stokes equation?I was thinking more in numerical terms, but why not? The strategy to make money here is also simple: just get you the millenium price million ;-)
 
User avatar
gelfand
Posts: 14
Joined: July 14th, 2002, 3:00 am

Trading with pricing models

March 13th, 2015, 2:11 pm

QuoteOriginally posted by: AlanBecome a market maker at the CBOE.When I was at the CBOE many market markers just fit some polynomial or slightly more complicatedfunction to the Black Scholes implied vol surface. None of them knew what a "Heston model" was. Has that changed?
 
User avatar
Alan
Posts: 2958
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Trading with pricing models

March 14th, 2015, 2:17 am

QuoteOriginally posted by: gelfandQuoteOriginally posted by: AlanBecome a market maker at the CBOE.When I was at the CBOE many market markers just fit some polynomial or slightly more complicatedfunction to the Black Scholes implied vol surface. None of them knew what a "Heston model" was. Has that changed?Well, from some limited experience I can say that some are indeed interested in modeling, although not necessarilythe Heston model.