Yesterday me an my colleague and co-author Andreas Binder had a Workout in Computational Finance seminar in Frankfurt, Germany. The seminar has been organised together with our German partners from Additive.
Four different topics have been covered:
Extreme Vasicek is not Enough - Mean reverting short-rate models. What are the pros and cons of trees, finite differences / elements, Monte Carlo techniques? Lognormal or normal models? What about higher dimensions?
Model Calibration and Spurious Precision - A general framework for stable and robust parameter identification. Even with analytic inversion formulae, noise in the data can lead to results which are pure nonsense. Can we trust our parameters?
When Monte Carlo is the Only Choice - More than 3 dimensions or severe path-dependence? Monte Carlo techniques. Monte Carlo or Quasi Monte Carlo? How can the variance of the result be decreased? What about early exercise?
Risk Management Cascades - The requirements posed by regulators become more and more stringent. How can we calculate the different VaRs? Expected shortfall? In reasonable time? And how can we build a CVA system?
During the breaks (off topic: the snacks provided by Roomers are excellent) a lot of lively discussions took place. I even managed to finish my sessions (almost) on time so that our German guests were able to watch and celebrate the win of their team against the team of the USA.
I hope the attendees of the seminar enjoyed it as much as Andreas and me have done.
StartupQuants 2014 - An Intimate Workout
This is a plan. We want to establish a new yearly event. And make it right for you. We just need a date and go.
StartupQuants 2014 - at the UnRisk Headquarter in Linz, Austria
StartupQuants 2014 - at the UnRisk Headquarter in Linz, Austria
We offer a first intimate Friday-Saturday workout with original thinkers on numerical maths and knowledge-based programming. It shall become an annual gathering to help startup quants leverage their work.
What you'll learn
You will walk away with tangible model-method-technology triples enabling you doing the difficult quant work and add value to valuation and risk management systems.
Who is moderating?
We're bringing the heads of the UnRisk Academy and authors of A Workout in Computational Finance and selected team members experienced in complex mathematical problem solving and hybrid programming. Those who write here.
How do you get in?
You are a founder or a quant who thinks like a startup in a financial institution committed to learning how to detect and avoid the risky horror of model-method traps. Interested in project organization and even promotion of your work.
You talk about yourself and what you are working on, stuck on, looking to improve.
Delegates will learn from each other and us.
To apply we will ask you to write us 3 lines about your position and work and what you expect from the workout.
It is FREE
It does not cost anything to attend. The workout is only open for 12 people. You need to come to Linz and stay two days, You need to pay lodging and food except two quick lunches.
Our Hot Topics include
Stabilization techniques for PDE solving
Overcoming the ill-posedness of the calibrations problem
Variance reduction MC and why QMC is the only choice
The valuation tsunami of MC VaR and xVA
The benefits of hybrid, multi-paradigm programming
Wolfram Language, Python, C++11, Java 8 or/and Javascript?
Why radical experimentation is indispensable
A simple but effective and efficient project management toolset
StartupQuants 2014 is a computational finance workout for quants who want to do the difficult work, avoiding the risky horror of computational traps, using motivational tools and transform their potential into stunning results.
For event pre-processing
If you are interested send me an e-mail (Herbert Exner) with a little work description, expectations, ideas, … Why, do I ask? We want to push the boundaries and make it a perfect event for and with you.
It is not a registration and will be kept confidential.
Germany versus USA: Arbitrage at the FIFA 2014 world cup?
At the FIFA 2014 world cup, group G consists of the teams of Germany, USA, Ghana, and Portugal. After two rounds, Germany and USA have 4 points each (one win, one draw), Ghana and Portugal one point each (one dra, one defeat). In the final round, Germany and the USA play against each other, and the match takes place at the same time as the Ghana vs. Portugal match.
The two teams leading in front after the final round will enter the stage of the final 16 teams.
From the arbitrage point of view, a draw between Germany and USA would make sure that both teams succeed in getting to the round of 16.
(There is the rumour that in the 1982 world cup, there has been an arbitrage match between Germany and Austria in Gijon with both teams coming to the next stage after a 1-0 win of Germany.)
Will similar things happen in 2014? I do not think so. Football is not only about results but also on passion.
Like UnRisk.
The two teams leading in front after the final round will enter the stage of the final 16 teams.
From the arbitrage point of view, a draw between Germany and USA would make sure that both teams succeed in getting to the round of 16.
(There is the rumour that in the 1982 world cup, there has been an arbitrage match between Germany and Austria in Gijon with both teams coming to the next stage after a 1-0 win of Germany.)
Will similar things happen in 2014? I do not think so. Football is not only about results but also on passion.
Like UnRisk.
UnRisk FACTORY 5.1 Is Released
From today, we take UnRisk FACTORY and UnRisk FACTORY Capital Manager version 5.1 to financial institutions.
We call it the "Bloomberg Release"
It integrates a market data solution for owners of a Bloomberg data license. The UnRisk FACTORY market data adapter for multiple market data sources was expanded by a Bloomberg instantiation that enables an immediate set up.
This is the 11th release of the compellingly comprehensive combo - a risk management solution and development system in one - since may 2008.
Order and set it up to run in 10 days.
We call it the "Bloomberg Release"
It integrates a market data solution for owners of a Bloomberg data license. The UnRisk FACTORY market data adapter for multiple market data sources was expanded by a Bloomberg instantiation that enables an immediate set up.
This is the 11th release of the compellingly comprehensive combo - a risk management solution and development system in one - since may 2008.
Order and set it up to run in 10 days.
What Will I Be Writing in 2024
This post has been motivated by a session of the WIRED UK June-14 issue.
The UnRisk team always looks a little into the future. What about new mathematical schemes, programming techniques, computing platforms and tools for better development and project management tools …. and media and communication platforms …. socio-economic and financial systems. In short, new environments for innovation and marketing?
Look a little further ahead?
The future as it was? It's hard to believe, but back in 2014 people propagated fragile, tightly coupled complex systems for front-to-back investment and risk management and not so few developers believed in one-technology-fits-all approaches.
It was not so difficult to predict that the financial security systems did not make the system safer - regulatory bodies forced banks do "go nuke" (do as the nuclear power plants needed to do, because of the technology)
Why the atifragility argument led to re-decentralization
For some it may be surprising, but ….
Big players providing information technologies did voluntarily split themselves into independent units. Buffers, diffusion and agility will fix what did not work: hierarchies, strategic plans and industrialized work in an economy of scales - flexibility, innovation and enthusiasm drive better offerings and consequently returns - and are safer for the economies.
I finance this led to a retraction of the strict central collateral management and clearing system. Market systems are too complicated to be centrally managed.
In combination with the extreme xVA treatment it broke to a marginal cost regime. Large banks have identified collateral transformation and central clearing as lucrative service, but that created the old problems of market dominance through another doorway.
How to program the money system
A system is universal, if it is solid enough to store and liquid enough to transform. A universal system is programmable. Our money can store values and it is a media for economic transactions, consequently, the money system is programmable. We now understand that this needed an operating system and development tools.
System architects, computer scientists and quants are now indispensable in new jobs dealing with money policy and creation as well as providing the operational semantics of the programs. Symbolic, declarative programs that needed a lot of clever algorithms to implement them.
Following the decentralization logic money creation became even more decentralized and for better programming we invented new money types, like derivative money - futures, options, …
Technologies Do It Yourself investors should use
Structure me this is not longer in disrepute. There are technologies that help to make complex
financial concepts to non experts. Decision support system will use new media and dynamic visualizations techniques. Systems will adapt to a financial "aura" of a market participant describing objectives, risk appetites, ….
High level, simple, interactive, knowledge-based, domain-specific programming will allow them to simulate deals with portfolios of various deal types.
The internet of finance is reality
We all, dealing with financial applications, knew that it was a joke to worry about big data, when still struggling with getting informative small (market)data with the possibility to turn this data into something that supports decisions.
Informations like debt and profit webs came into play to let agents understand feedback loop better ….
Those data are now available and widely accessible. They allow us technology providers to intelligently combine modeling with more data driven adaptability - making data meaningful.
Reasons, why quant work is now more polarized
10 years ago quants whee so overwhelmed by the operational requirements meeting the regulatory system. If they did xVA they did the same thing (valuations) 100 million times an hour instead of 5 million times an hour when doing portfolio-across.scenario simulation. This needed a lot of plumbing, especial if their banks suffered from the was-not-invented-here syndrome.
Now, the new technologies and tools give them time to optimize their business. They can design new financial systems. How they do it will not make a big difference, but what they do will.
Adding value still means: doing the hard work.
Programming is knowledge-based
Programming is symbolic, multi-paradigm and uses algorithmic knowledge-bases. There are development environments that support high level programming in multi-languages, that are platform agnostic. We do not longer care about versions and releases and underlying programming languages. All algorithms for high performance computing are inherently parallel and support all architectures of the new computing muscles.
The most complex applications go where the users go
Write once, deploy anywhere.
Clouds went private and crunch and store all the (risk) numbers blazingly fast.
Even smart devices will have enough power and display-quality to do massive pre-processing creating insight about incoming data flows and prepare them for processing.
Financial objects and their risk spectra can be post-processed and insightfully arranged on the smart mobile devices.
Program everything
The blog posts are interactive, some stories are created by and act like programs. 70% of the news are created by computers. Ask a question and you will receive a little program that answers all questions of that type.
More quants do more meaningful things
and turned them into businesses
And
it is late June, the sky has scattered clouds, the birds are singing, …..
Basic Math - Big Impact
In today’s blog I am going to give you some insight on how
the development of the UnRisk pricing routines works.
Some years ago one of our customers asked for an UnRisk
function to price a bond having the following coupon structure (Example 1):
If the EURCMS10Y is
bigger than the EURCMS2Y, then 8*(EURCMS10Y-EURCMS2Y) is paid (with a floor of
0%), otherwise the instrument pays a fix coupon of 4%
We developed functions to price such a product (under LMM
and under Hull & White 2 factor) by implementing the following coupon logic
(in the example RefRate1 is the EURCMS10Y and RefRate2 is the EURCMS2Y):
(1) Condition:
RefRate1 –
Refrate2 > 0
(2) Branch
1 (if Condition is fulfilled):
Coupon Rate = 8*(RefRate1 - RefRate2),
Floor = 0%
(3) Branch
2 (if Condition is not fulfilled):
Coupon Rate = 4%
Some weeks after we had finished these developments, another
customer came up with the following coupon structure (Example 2):
The coupon rate is
given by 8*(EURCMS10Y-EURCMS2Y) capped with the EURCMS10Y (again, with a floor
of 0%)
This can be written as
(4) Min(8*(EURCMS10Y-EURCMS2Y),EURCMS10Y),
Floor = 0%
So we were thinking on how to incorporate this instrument
into our framwork. We came to the following conclusion (again, we use the variables
RefRate1 and RefRate2 for the two EURCMS rates):
(5) Condition:
RefRate1 > 8*(RefRate1-RefRate2)
Which can be written as (this is the simple math part):
(6) Condition:
-7*RefRate1 + 8*RefRate2 > 0
(7) Branch
1 (if Condition is fulfilled):
Coupon Rate = 8*(RefRate1 –
RefRate2) , Floor = 0%
(8) Branch
2 (if Condition is not fulfilled):
Coupon Rate = RefRate1 , Floor =
0%
In order to enable us to cover both instruments by the same
pricing function and to prevent us from
new implementations if another customer comes with a similar problem, we
extended our first pricing function to cover the following coupon structure (x1,
y1, z1, x2, y2, c2, f2, x3, y3, c3 and f3 are numbers):
(9) Condition:
x1*RefRate1 + y1*RefRate2 > z1
(10)
Branch 1 (if Condition is fulfilled):
Coupon Rate = x2*RefRate1 + y2 *
RefRate2 , with Cap c2 and Floor f2
(11)
Branch 2 (if Condition is not fulfilled):
Coupon Rate = x3*RefRate1 +
y3*RefRate2, with Cap c3 and Floor f3
Within the UnRisk world we gave an instrument having such a
coupon structure the name „Steepener Type 2“.
At the end I am going to explain how the mentioned examples
may be set up as Steepener Type 2.
Example 1 (x3 and y3 may be set to any number, e.g. , 1):
(12)
Condition: x1 = 1 , y1 = 1 , z1 =
0
(13)
Branch 1: x2 = 8 , y2 = -8 , c2
= 1000 , f2 = 0
(14)
Branch 2: x3 = 1 , y3 = 1 , c3 =
4% , f3 = 4%
Example 2:
(15)
Condition: x1 = -7 , y1 = 8 , z1 =
0
(16)
Branch 1: x2 = 8 , y2 = -8 , c2
= 1000 , f2 = 0
(17)
Branch 2: x3 = 1 , y3 = 0 , c3 =
1000 , f3 = 0
By using some variables and applying basic math we have
developed a very mighty function for pricing instruments fitting into our „Steepener
Type 2 framework“. Many of the UnRisk users price instruments fitting (and
there are many of them) into this framework by the use of one single pricing
function.
Quantum Physics and Options
Coming from the field of quantum many body physics I have always been interested in books and publications combining quantum physics with quantitative finance. Therefore I am going to write some blog posts on this topic. As a starting point I want to introduce the concept of the Hamiltonian in the pricing of options. In quantum mechanics, the Hamiltonian is the operator corresponding to the total energy of the system. By analogy with classical mechanics, the Hamiltonian is commonly expressed as the sum of operators corresponding to the kinetic and potential energies of a system in the form
H=T+V
Can a Hamiltonian formulation provide new tools for obtaining solutions for op- tion pricing ? Two key concepts related to Hamiltonians are
- Eigenfunctions
- Potentials
It turns out, that the knowledge of all the eigenfunctions of a Hamiltonian yields an exact solution for a large class of path-dependent and path-independent options. If we take, for an example, barrier options - they can modelled by placing constraints on the eigen- functions of the Hamiltonian. In our next blog post we will review those aspects of quantum mechanics that are relevant for the analysis of option pricing.
Would I Drive An Open Source Car?
A few days ago Tesla Motors' CEO, Elon Musk announced: All Our Patents are Belong To You.
It is very natural thinking: balance the tremendous inflow of valuable in depth information that you cannot transform a fraction of into margins with a free outflow of great results we cannot transform into values alone? No doubt, open source software was one of the drivers of globalization. Open innovation is less radical, but still offer out licensing.
As a reviewer and evaluator of innovation projects of the European Commission I evaluated about 500 projects. All of them were performed by international partners and the consortium agreement controlling the exploitation and intellectual property rights were seen as important part of the project. They often complicated the project and I argued for a more liberal treatment of intellectual property rights. Technology leadership is nit defined by patents (says, Elon Musk) - I agree.
But I ask myself: Would I drive and open source car? I would definitely drive an open innovation car!
It is very natural thinking: balance the tremendous inflow of valuable in depth information that you cannot transform a fraction of into margins with a free outflow of great results we cannot transform into values alone? No doubt, open source software was one of the drivers of globalization. Open innovation is less radical, but still offer out licensing.
As a reviewer and evaluator of innovation projects of the European Commission I evaluated about 500 projects. All of them were performed by international partners and the consortium agreement controlling the exploitation and intellectual property rights were seen as important part of the project. They often complicated the project and I argued for a more liberal treatment of intellectual property rights. Technology leadership is nit defined by patents (says, Elon Musk) - I agree.
But I ask myself: Would I drive and open source car? I would definitely drive an open innovation car!
Hot Topics at ECMI 2014
Last week, at the ECMI 2014 conference, the Friday was devoted to Finance. Plenary speakers were Claudio Albanese from Global Valuation and Jörg Kienitz, head of quantitaive analysis of the German Postbank.
Not too surprisingly, they saw future challenges for quants in xVA, and they also saw a shift of human ressources in quantitative analytics from valuation and trading to risk.
Not too surprisingly, they saw future challenges for quants in xVA, and they also saw a shift of human ressources in quantitative analytics from valuation and trading to risk.
Young Quants - Experience is Overrated
It is all about radical instead of incremental innovation.
We are sitting on enormous amount of cash, but business grows slowly. It seems the problem arises from the flawed assumption capital must be conserved at all cost.
Radical innovation needs talent spotting
We need more radical innovation at the Main Street and the Wall Street. And this needs talents.
Potential beats experience and competencies
Future businesses will be too volatile and complex to rely on experience and competencies (only) - we need potential, the ability to adapt to the changes swiftly.
To innovate is hard work
This is an enormous opportunity for young quants. Innovate by doing the difficult work. With motivation, curiosity, insight and engagement. Quants develop the tools that shall guide investments to the best opportunities.
To do that they use tools themselves. Quants engineer, model and program.
Tools shall help them to explore new things and verify them by doing multiple experiments. We at UnRisk provide solutions, but we also provide what we call motivational tools for quants - UnRisk-Q, providing the UnRisk Financial Language implemented in the UnRisk Engine with a broad coverage of deal types and risk scenarios.
We decided to help (young) quants leverage their work
This post has been motivated by the June 2014 issue of HBR Magazine - Experience is Overrated.
We are sitting on enormous amount of cash, but business grows slowly. It seems the problem arises from the flawed assumption capital must be conserved at all cost.
Radical innovation needs talent spotting
We need more radical innovation at the Main Street and the Wall Street. And this needs talents.
Potential beats experience and competencies
Future businesses will be too volatile and complex to rely on experience and competencies (only) - we need potential, the ability to adapt to the changes swiftly.
To innovate is hard work
This is an enormous opportunity for young quants. Innovate by doing the difficult work. With motivation, curiosity, insight and engagement. Quants develop the tools that shall guide investments to the best opportunities.
To do that they use tools themselves. Quants engineer, model and program.
Tools shall help them to explore new things and verify them by doing multiple experiments. We at UnRisk provide solutions, but we also provide what we call motivational tools for quants - UnRisk-Q, providing the UnRisk Financial Language implemented in the UnRisk Engine with a broad coverage of deal types and risk scenarios.
We decided to help (young) quants leverage their work
This post has been motivated by the June 2014 issue of HBR Magazine - Experience is Overrated.
Subscribe to:
Posts (Atom)