Forums

Sorting an Array

kannan — Fri, 23 Dec 2011 11:16:01 GMT

My objective is to find the 3 largest maximum values from the observation levels. I think this can be scripted by sorting an array (in descending order) and finding the first 3 values of an array. How can I go about in doing this in Numerix scripting? Does Numerix provide this array capabilities?

Forum: Numerix CrossAsset XL

Generating Multivariate Normal variables using Cholesky decomposition

husky123 — Sun, 28 Aug 2011 10:46:23 GMT

I am trying to calculate the VaR measure of a portfolio using Monte Carlo method. Using some sample history, I was able to obtain the pairwise correlation of returns between all the assets. I then used the Cholesky decomposition to obtain the lower triangular matrix from my correlation matrix.

Can someone please indicate how I can obtain the multivariate stochastic normal variables from here to proceed to my asset price calculation?

Forum: Quantitative Analysis

Calculating Option Adjusted Spread in Numerix given the Bond Data

husky123 — Mon, 25 Jul 2011 08:00:36 GMT

Is there a function/procedure/example to calculate the Option Adjusted Spread (OAS) of an Option-Embedded Bond in the Numerix Cross Asset XL version? I noticed there is a function to calculate the Zero Volatility Spread (ZSpread) and we know that OAS=(ZSpread-Option Value). We know that the Option value changes with time, please suggest a way to implement this.

Forum: Numerix CrossAsset XL

Calculating OAS in Numerix Given the Bond data

husky123 — Mon, 25 Jul 2011 07:56:26 GMT

Is there a function/procedure/example to calculate the Option Adjusted spread (OAS) in Numerix, I see there is a function to compute the Zero-Volatility Spread (Z-spread) and We know that OAS=(ZSpread-Option Cost). Please suggest a way to implement this.

Forum: Quantitative Analysis

Calculation of CMS-Spreads in EUR

Michael_T — Thur, 26 May 2011 09:15:44 GMT

Hi everybody,

I'm looking to someone who has experience with calibrating and pricing CMS-Spreads.

We worked with this example in EUR:
maturity: 5y
coupon: 2 * (CMS10y - CMS2y), floored at 2.25% and capped at 5.50%

At first we used the Hull/White-Model but the calculated Numerix-price was to far away from counterparty prices
(pls note, that I know, that CMS-Spreads have higher differences between calculation and offered prices than pv-swaps).

After that experience we switched to Libor-Market-Model.
And now we have two problems:
- for calibration Numerix takes time between 30 min. and more than one hour (we use a PC with 3 CPUs, but only one is in progress) or sometimes Excel 2007 crashed.
- if the calibration was done than the Numerix-price was better as with H/W-Model but also to far away.

Who has an idea or can give us some more infos?

Thanking you in anticipation for your help and spent time.

Forum: Numerix CrossAsset XL

VALUEDATE in pricing

Graham — Fri, 01 Apr 2011 20:35:32 GMT

I have several questions about the input VALUEDATE which is available in Kernel Price and some other pricers/analytics. I need to clarify my understanding of VALUEDATE. The application guide (s. 11.2) says this:

Quote:

The dirty price only includes payments on or after the value date.... The PVs are computed by reverse discounting the model-calculated PV; e.g., outputPV = modelPV / DF(nowDate, valueDate).

Let's let t0 = NOWDATE, t1 = VALUEDATE, and assume we're using a Kernel Price with t0 < t1.

As I understand it from the Application Guide, in a deterministic interest rate model, a discounting pricing product X is calculated by Kernel Price at t0, and then future valued to t1 effectively using

X(t1) = / DF(t0,t1),

where X(tn) is the present value of X at time tn according to Kernel Price, and DF(tj,tk) is the discount factor from the model between time tj and tk. Here C(t0) is the sum of any discounting values that were accumulated into X at paydates t < t1. I.e. a Cash statement accumulating into X, such as X += Cash(...) at time t < t1, has 0 present value at t1, so X(t1) is actually the sum of discounting values which occurred at (paydates) >= t1, present valued to t1.

So here are my questions.

1) Say we were pricing a callable bond with a single call fix date at t with pay date at t + 2, with t0 < t < t + 2 < t1, using an assignment statement in script such as X = MIN(X,Cash(CallStrike,...,ThisPay at t+2)), where CallStrike is not a discounting product - maybe it's a temporary calculated from an INDEX input. If the bond called at t, so that X = Cash(CallStrike,...,ThisPay at t+2), then we would find X(t1) = 0, correct? This makes sense to me because if the bond called at t < t1, it should no longer have a value at t1. However, is this the way it works?

2) This leads to another question: if at t it just happened that X(t) == Cash(CallStrike,...,ThisPay at t+2) before we encountered the assignment statement X = MIN(X, Cash(CallStrike,...,ThisPay at t+2)), how is X assigned in the statement X = MIN(X, Cash(CallStrike,...,ThisPay at t+2))? It seems to me we could have two different values for X(t1) depending on which value is chosen by the MIN() in the assignment. How is this handled?

3) What happens in a deterministic IR, deterministic CR model when t0 < t1? (X is now a CREDIT product.) Is the calculation of X(t1) from X(t0) influenced at all by survival probabilities? I guess not, so that X(t1) = / DF(t0,t1) as in the deterministic IR model, where in this case C(t0) is the sum of Cash() and CashOnDefault() values accumulated into X at paydates before t1.

4) Finally, what happens in a stochastic IR model, for example in the case of the callable bond considered above? Say we're pricing with BackwardTree for example, and on one branch the call at t < t1 occurs but on the other it doesn't. Now X(t0) and C(t0) are expected values. Do we form X(t1) = / DF(t0,t1) as before, where in this case DF(t0,t1) is the discount factor taken from the yield curve? Or is X(t1) computed in some other way? For example, we might have

X(t1) = E / DF(t0,t1,p) ]

where the p in X(t0,p), C(t0,p) and DF(t0,t1,p) indicates the value along a particular path/branch, and then X(t1) is the expected value across paths/branches.

Sorry about the many questions. I'm curious about how VALUEDATE functions in these different scenarios because I'd like to know to what extent it's possible to use it for periods longer than "some small number of days", as the application guide describes its typical use.

Forum: Numerix CrossAsset XL

Numerix 8.2 Excel Addin issue

vancovc — Sun, 20 Mar 2011 17:38:29 GMT

Hi Numerix,

I am currently running Numerix 8.2 addin on MS Excel 2007 on Windows 7 platform. The problem I am experiencing is that Excel takes a very long time to load and unload the addin. Sometimes it takes as long as 5mins to open a fresh instance of Excel with Numerix 8.2 addin. I am wondering if you have received any report related to this problem. If so, what solution do you recommend to reduce the addin load time?

Thank you,
Vanco

Forum: Derivatives Markets

MFE at University of Hawaii at Manoa

mfeuhm — Fri, 10 Dec 2010 20:10:16 GMT

The MFE program is a new program at University of Hawaii at Manoa.

* Small class size (35 student max)
* Laptops are provided for the duration of the program
* We are one of the few programs which offers a course on weather derivatives and environmental finance
* We have a Bloomberg terminal and the latest books and journals
* Lectures are recorded in our hight tech class room for students to review
* Job interview training is provided as well as aggressive job placement
* Students have an opportunity to take part in applying what they learn to advise a hedge fund with over $100,000 under management

Tuition Estimate for the 2010/2011 academic year, 30 credit hours:
Tuition - $32,000 (Residents - $20,000)

If you want to learn more about this program, see here http://mfe.shidler.hawaii.edu/

Forum: Quantitative Analysis

JPY Basis Swap curve

dwu3954 — Wed, 29 Sep 2010 10:41:14 GMT

Hi,

As we know the JPY libor curve is very low and the Xccy basis spread is negative (big in absolute value). So the zero coupon rate for the basis curve is negative. This situation is causing valuation issue in Numerix, (especially with Numerix integration in Summit version).

So to avoid this situation I have to add 3 or 4 bps to the Basis curve to keep the zero rates positive.

I want to ask if someone else has the same problem and how you solve it.

Thank you.

Forum: Derivatives Markets

Local Volatility along "Most Likely Path"

fab10ab — Wed, 17 Mar 2010 13:27:51 GMT

This method of estimating local volatility is explained in Jim Gatheral's volatility book and also in a Calyon white paper by Adil Reghai.

Has anyone on this forum implemented this method?

Thanks!

Forum: Quantitative Analysis

time slice observer values

Graham — Tue, 16 Mar 2010 17:06:47 GMT

The values reported by a time slice observer of a single script product may be multiple on a single fix date.

For example, when using a Backward Analytic pricer and deterministic EQ model, I get 3 slightly different values of the product for every (non-past) fix date. I imagine the three values correspond to bumps in order to calculate deltas and gammas for the script products.

When using a Backward Tree pricer and a Black EQ model, I get what looks like a tree of values for the observed product, with possibly many values (~40) for a given fix date, the number of values reported non-decreasing as a function of fix date. I get similar behaviour with a Hull White 1F model and a Backward Tree pricer.

My question is, what am I looking at?

In the BackwardAnalytic/deterministic EQ model example (where the product I was observing was the EQ price), it looks as though the value I really want to extract is in the Value2 column of the observer results. Value1 and Value3 are the bumped values.

In the BackwardTree examples it would be nice if the values reported by the time slice observer correspond more-or-less 1-1 with the tree used by the pricer. This would allow visualization of the values considered during the pricing for debugging or validation uses. (I mean debugging/validation of our work, not the Nx models.)

Forum: Numerix CrossAsset XL

temporary products in script

Graham — Tue, 16 Mar 2010 16:41:24 GMT

Suppose I have the following script fragment:

PRODUCTS

DISCOUNTING Bond

TEMPORARY CallStrike

...

END PRODUCTS

PAYOFFSCRIPT

...

IF ISACTIVE(CallDates) THEN

...

CallStrike = CASH(Notional * Something, CallDates, THISPAY)

Bond = MIN(Bond, CallStrike)

...

END IF

...

END PAYOFFSCRIPT

Is there any difference between the following statement:

Bond = MIN(Bond, CASH(Notional * Something, CallDates, THISPAY))

and the way it is done in the script fragment, assigning the CASH statement value to a temporary product first?

This is an overly simplified example, but I have many uses where assigning a CASH statement to an intermediate variable is helpful. Should I be using discounting products as intermediate variables for the results of (functions of) CASH statements, rather than temporary products?

--Graham

Forum: Numerix CrossAsset XL

Tip for Filtering Trades

sanket_nx — Mon, 15 Mar 2010 17:40:26 GMT

One of the most powerful features of Portfolio is ability to filter in almost every imaginable way. Trades can be filtered based on anything from trade date and counterparty name to funding leg basis and everything in between.

There may occasionally be instances when you need to filter trades using criteria not provided "out-of-the-box." I've found 2 ways to get around this situation.

1) Say for example you want the ability to filter trades based upon the trader who put on the trade. First off, there is no 'Trader' or 'Trader Name' field in any of the trade templates in Portfolio. The question now is: "how do you filter on a condition that doesn't exist?" The problem can be easily solved creating a 'Custom Property' using the Portfolio Administration Tool: Portfolio Administration Tool > Trade > trade type (See attachment 1). Once this field has been created, you'll now be able use it as a filtering condition (See attachment 2). If the out-of-the-box filtering conditions don't provide you coverage you need, you now have ability to filter using custom properties!

2) If you're not in the mood for creating custom properties, you can also use the 'Notes' section of a trade as a tool for filtering. Continuing with the example above, to set up a filter for trades by John Smith, simply add this (or any variant) to the 'Notes' section of relevant trades "Trader : John Smith." (See Attachment 3). Filtering by the trader, John Smith, is now as easy as using the filter, "Notes contains string 'John Smith'" (See attachment 4). The advantage of method 2 is that you enter multiple pieces of information in the Notes field and filter each separately using the 'Contains String' as the comparison method.

I'd be more than happy provide a more detailed explanation/clarification if anyone needs it. In the meantime, happy filtering!

-Sanket

Forum: Numerix Portfolio

Portfolio Scripting

dlang — Mon, 15 Mar 2010 15:55:22 GMT

Here is something I find useful when using Portfolio scripting (usually when trying to create batch files to automatically run reports).

Use the Help() function to output a list and description of all the methods of a given class.

For instance if I was working on a project which required me to use Portfolio's reporting API I could use the following; "ReportResult.Help()"

Hope this helps someone, I know I really rely on this.

Dan

Forum: Numerix Portfolio

Structured Product Webinar: Trends for 2010

mark_adams — Mon, 15 Mar 2010 15:32:41 GMT

Last week, we hosted webinar on the state of the global structured products with Keith Styrcula, Chairman of the Structured Products Association, and Joe Burris, Head of Americas StructuredRetailProducts.com.

Read the recap or listen to the replay.

A couple highlights:

Tranche sales appear to have stabilized in 2H2009, and are expected to grow 9% in 2010. (source: StructuredRetailProducts.com)
Structured product vehicles are trending toward simplicity, but the underlyings are growing more complex.
Market-Linked Products with 100% capital protection (especially FDIC-insured), leveraged returns, reverse convertibles and auto-callable features are growing in popularity in 2010.
The big hidden development is that mutual funds are substantially increasing their use of SPs to enhance investor returns.
RIAs are the next significant growth market for structured products.

So, I put it to the forum: What do you think are the key areas to watch for in structured products in 2010?

Forum: Derivatives Markets