In the November issue of RINA Systems' Performance Update:
Feature Highlight – Monte Carlo Simulation for Portfolios
New Information – Performance Update Strategies
Offers - Customer Survey
Feature Highlight -
PortfolioStream version 4.5 which will be released on December 10th will contain a new state-of the art portfolio Monte Carlo simulator that seamlessly interfaces to PortfolioStream’s performance databases. Now you can
conduct what-if analysis with portfolios and systems generated in PortfolioStream with just a few clicks of the button. You will be able to estimate the possible returns and drawdowns that could have been realized
from the historical distribution, based on various random sampling methodologies. This helps the users gain insight into alternative results that could have occurred with a reordering of returns or trades (assuming the same underlying return/trade distribution). Industry
experts recognize Monte Carlo simulation as a valid method for better understanding the risks associated with trading a portfolio of systems and setting reasonable expectations for and possible variations of performance. Our Monte Carlo software offers a broad set of
portfolio simulation capabilities. Here are several capabilities that this new software has to offer:
Multiple Types Of Historical Data To Resample From Your Portfolios Including:
Returns at daily, weekly, and monthly intervals expressed as both a percent or $ change.
Historical trades in the portfolio or individual system.
Sampling of both the portfolio composite (combined) return, or sampling from the individual market returns to generate the portfolio return.
All of the features above give users maximum flexibility in setting up your simulations.
Multiple Re-Sampling Methods Including:
Standard Monte Carlo simulation using Normal Distribution.
BootStrapping with Replacement.
BootStrapping without Replacement.
Users are not limited to a single simulation and can choose the method they feel is most appropriate.
Several Key Statistics In Output Tables And Graphs:
Total Return.
Average Return.
Maximum Drawdown.
Longest Time Between Equity Peaks.
Maximum Consecutive losing trades (for trades based resampling).
Users are not limited to a single simulation and can choose the method they feel is most appropriate.
Fully Integrated:
Making use of these powerful features just takes the click of a button, because our Monte Carlo Simulator is fully integrated to
PortfolioStream and interfaces to the same databases of portfolio performance.For an example of Monte Carlo simulation in use check out our
updated study of the trading strategies below, where we use the simulator to better understand system performance.
New Information - Performance Update Strategies
We wanted to update the trend following models we introduced and tested in the previous issue of
Performance update. We had selected some common strategies from the book “Trading Systems and Methods” by Perry J. Kaufman and felt it would be interesting to track their performance.
For a complete review of the system logic and concomitant TradeStation code click here, which references the previous issue of Performance Update.
Readers will recall that we tested the strategies on 29 domestic futures markets. In this issue
we tested the same parameters, as we again wanted to avoid retrospection and “curve fitting”. The first two tables below show both the performance since inception, while the following graphs
shows the return since the last running the strategies for the two most profitable (retrospectively) of the six variations. The full portfolio reports for all variations of the strategies are
available. For more information click on the following links:
In updating the entire period and returns only
for October we reviewed the returns for the standard deviation band system - PerUp_BBandCBO(60,2,0.01) - and the Channel Breakout - PerUp_CBO(80,32,0.01)
(which the graph below depicts ) and we were surprised to find how correlated the returns for the 2 portfolios appeared. We then decided to run
the correlation coefficient for the daily changes in equity for these two variations of the basic system we were testing and found it to be .97 over
the month of October. This finding peaked our interest to see the correlation for the historical period between the two portfolios. From the period where
both variations of the systems were trading (5/26/1995) until the end of October, 2003 we see that the correlation of daily changes ($ returns) in equity
were .27.
For those not familiar, the correlation
coefficient between 2 sets of values (in this case daily returns) will tell us if the data tends to move together. In our case it indicates
whether large returns of one portfolio are associated with large returns of the other (positive correlation) or conversely whether small
(or negative) returns of portfolio are associated with large returns in the other (negative correlation). If the returns between portfolios
were unrelated the value would be near zero. If the results we saw in October had been indicative of entire historical performance, a valid
question would be whether tracking both models prospectively would yield any additional information (as their returns would have been nearly
identical). As it turns out this is not the case. Though this observation may suggest that trading systems with similar approaches (in this
case trying to exploit long term trends) can over short periods (a month in this case) become very aligned in both returns and positions held
(which a cursory review revealed much overlap). Further work might involve looking at the rolling correlation, or the correlation for various
shorter time periods, and possibly the composition (markets traded) of the portfolios and the degree to which positions overlapped over time.
It’s also worth noting that the graph below represents the first month of “out of sample” results for the trading strategies.
Simulated Statistics
We also decided to take a look
at some alternative paths (equity curves) that could have occurred given the same historical return distributions. The graphs below
are generated using the new Monte Carlo capability that will be included in PortfolioStream version 4.5. We can infer from the
graph below that the “channel break out” system with the 80 period lookback (PerUp_CBO 80,32,.01) realized a return
historically of $5,827,840.65 in the test. Yet simulations of 1000 paths (possible equity curves) of monthly returns (in $)
show that simulated returns using the bootstrap with replacement method were as high as 9,095,831 at the 90-th percentile (top
10% of simulated returns) and as low as $2,598,844.88 (in the bottom 10% of simulated returns), the distribution being obtained
from running 1000 hypothetical simulations. Perhaps more interesting and displayed in the second image, the historic return
(also known as realized) achieved in our test had a longest period between equity peaks was 18 months. However, at least 10%
of the resampled returns displayed a maximum period between equity peaks of 49 months or more!
Notes on studies:
The average trade value is
on a per trade basis. Nearly all of the trades simulated here were multiple contract trades due to the fact they were geared
to risk 1% based on a hypothetical constant funding of 1 million dollars. The per-contract average trade would have
obviously been much smaller. Finally, the maximum drawdown is the largest peak to trough dollar drawdown for the
portfolio during the tested period.
Offers - Customer Survey
In our efforts to continually meet the needs of quantitative traders,
we request feedback from our customers and readers regarding functionality that they would like to see addressed in trading and investment software. Please take the
time to fill out the following survey by clicking here. As we schedule enhancements to existing applications and possible new development projects we are
eager to hear what traders think about technologies that would be valuable. To show our appreciation, anyone filling out this survey and leaving a valid name, email,
and phone number will receive a 10% discount on their next purchase of software or services. This survey will run through April 31, 2004 and the discount will be
good for any purchase (or lease) made before June 1, 2004. If you would like to contact RINA Systems, Inc. you may do so via phone at (513)-469-7462, via fax at
(513)-469-2223, or via e-mail at info@rinasystems.com
The purpose of the Performance Update newsletter is to
offer ideas and educational samples of strategies through the use of RINA Systems software applications. The strategies described in the Performance
Update newsletter are not intended to be used as actual trading strategies, but are included for educational purposes only. No offer or solicitation
to buy or sell securities, commodities or securities derivative products of any kind, or any type of trading or investment advice, recommendation or
strategy, is made, given or in any manner provided by RINA Systems or any of its affiliates. If such advice is sought, the services of a licensed
professional should be obtained. Past performance, whether actual or indicated by historical, simulated tests of strategies, is no guarantee of
future performance or success. Hypothetical or simulated performance results have certain limitations. Unlike an actual performance record simulations
do not represent actual trading. Also since the trades have not actually been executed the results may have under - or over - compensated for the
impact if any, of certain market factors such as liquidity. Simulated trading programs in general are also subject to the fact that they are designed
with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profits or losses similar to those shown.
There is a significant risk of loss involved in futures and securities trading. RINA Systems, Inc. will not be responsible for any losses or lost
profits resulting from investment decisions based on any RINA Systems product or service information obtained through use of any RINA Systems product
or service or otherwise. TradeStation® and EasyLanguage® are registered trademarks of TradeStation Technologies, Inc.
