Cassandra V2.24.0 (No longer trading)
Cassandra is a trading system that employs guerilla trading techniques with a very strong emphasis on risk management facilitating a controlled exposure to the foreign exchange market
Cassandra uses eighteen diverse FX pairs and trades them in a single account. This leverages up the performance exponentially producing what we consider to be excellent returns. Below are the results of the backtesting of this system, an independent analysis by Adam Landes and finally the results of the individual tests.
This is version 2.24 of Cassandra, she has been in forward test since November 2012 and is now upgraded to this latest version of the system.
Guerilla Trading Rules
- Employ a small controlled percentage of available capital for each trade
- Reduce risk to a minimum by trading as little as possible (typical trades last a few hours to a few weeks)
- Spread risk and magnify gains by using multiple instruments packaged together in a single "fund"
- Employ infinite patience and iron discipline to wait for the correct trading opportunity
- Take profits early but keep a runner and give it "wiggle" room
- Any system tuning to be done on instrument character and kept to an absolute minimum
- A maximum loss level to be used for each instrument, if breached trading stops for that season
- To maximize the return runners should use a trailing stop with no target
- Risk on each individual currency kept to a maximum level
- If the monthly Loss parameter is exceed for a currency it is not traded again, until the condition desists
Cassandra Mission Statement
Enable fund managers to have good exposure to foreign exchange, keeping drawdown to a maximum of 15% whilst not limiting profit potential and having a minimum performance of 35% PA.
Combined Backtest Performance(logarithmic) Curve 2007-2013
Average Monthly performance
The maximum drawdown of 13.3% occurred in September 2007
Various methods are used in the algorithms to control risk but during periods of low volume price tends to chop and drawdowns are inevitable. Like all breakout systems if price breaks out then reverses there are hits to equity. Currency pairs will trend for long periods of time but this trading environment could change. The system is designed to perform well in both bull and bear markets but will suffer if it stagnates.
Drawdown is calculated from banked profits only, The system sets targets when a position is opened and takes profits when the targets are hit closing 75% of a full size position, this acts as insurance against reversals, the remaining 25% position is a runner managed by means of a wide trailing stop. This being the case the drawdown from un-banked profits will be much larger, the system considers un-banked profits to be working capital and is designed to maximize the return on these remaining positions not minimize drawdown. It can be uncomfortable watching a runner hit a stop for a small loss after achieving a good profit. You have been warned don't count your chickens until they are hatched!
The system will still be profitable if runners hit their stops early but the lions share of profits come from allowing the runners to shine and giving them "wiggle" room.
Risk to equity being taken V drawdown sustained.
Winning and losing streak duration
The Fibarchie Master Slave infrastructure enables money managers to set their own risk factor increasing or decreasing the trade size by a percentage. Cassandra puts 2% of available equity at risk for each trade. In the testing this was found to give a drawdown of just over 13% and a profit of +600B from 100k (64,000,000%) over the 6.5 years backtested, this is the recommended setting which in the Fibarchie Slave equates to a setting of 100%.
To demonstrate the effect of profit and drawdown the position size was multiplied by 0.25, 0.5, 1, 2 and 3 ( This equates to Slave settings of 25%,50%, 100%, 200% & 300%) Above 7% risk/trade the system exhibits a 100% drawdown. The center line equates to the recommended setting of 100%. (2% risk/trade)
Using this logarithmic scale the drawdowns are only perceptible at the higher risk setting. At three times the recommended setting (6% risk/trade) the drawdown went up to 41%! The system has been designed to return equity to clients first and foremost and give a return on equity as a secondary principle. Far more important than increasing risk is encouraging clients to allow the fund to compound the return.
Unlike traditional funds the Fibarchie system actively trades instruments and can profit from a rising or falling market. This being the case end clients do not need to time their entry and exit from a fund.
Performance of the 8 currencies traded in the combined back-test
The Cassandra system has been independently analyzed by one of the Fibarchie partners they took the raw data from the Metatrader backtests and produced his own set of data based on the combined and leveraged performance of the eighteen pairs that constitute the Cassandra system. Fibarchie analysis is geared towards checking the functionality and profitability of the system, managing risk and drawdown.
CASSANDRA BACK-TEST ANALYSIS 2007-2013
For all the instruments, the strategy closed (including partials) 2,459 trades over the back-test period, winning 63.8% of the time. 64.8% of days were winners, with the average daily return (when trades were closed) 1.31%. Positive weekly, monthly, quarterly, and yearly instances were all very high, including 27/28 winning quarters and 7/7 winning years.
Modeled maximum drawdown was capped at 13.3%. There were 345 draw-down phases, averaging 2.6% (one standard deviation 2.3%). Drawdown skew was very negative (-1.64), i.e.. there were few relatively large drawdowns, although their impact was high. Daily (calendar) compound return was 0.63% on back-test, while weekly compound return was 4.50%, monthly compound return was 20.80%, quarterly compound return was 76.29% and finally, yearly compound return was 865.82%.
All years on back-test were positive but the performance range quite large, from a high of +3.087.65% (2008) to a low of +486.4% (2011). During the back-test period, a 1% return was achieved every 1.6 days, approximately. The strategy performs best with pronounced volatility.
|CASSANDRA BACK-TEST ANALYSIS 2007-2013|
|Daily Compound Return||0.630%|
|Weekly Compound Return||4.498%|
|Monthly Compound Return||20.802%|
|Quarterly Compound Return||76.288%|
|Yearly Compound Return||865.819%|
|Absolute Drawdown (Balance)||-$4,297.48|
|Maximum Drawdown (Balance)||-$51,909,756,406.51|
|Maximum Relative Drawdown (Balance)||-$53,897.73|
|Days at Maximum Balance||1,153|
|Days in Drawdown||1,338|
|Percentage (of periods with closed trades)||64.84%||78.75%||95.12%||96.43%||100.00%|
|* 2013 is annualized based on results through 25th October, 2013.|
Constituent Back tests 2007-2013
Every effort has been made to make the backtests as realistic as possible, the tests were conducted using a standard retail Alpari UK account and trading CFD instruments. The account was live and the spread used was whatever Alpari had set at the time. If you would like to see the raw data from these tests then please let us know.
Show me the employer who would tolerate a trader doing nothing for weeks on end and the trader who is still sharp after so much inactivity. Guerilla trading is unique, potentially very profitable and not always the easiest system to live with. Risk is controlled to such an extent that boring is probably the best word to describe Cassandra.
Each of the Cassandra instruments was back-tested using a starting equity of 100k. These tests were conducted using Metatrader 4.00 Build 509. the individual back-test drawdown ranged between 6% and 15%. When the instruments are traded together in the same account they tended to offset each others drawdowns and leverage gains. Conditions could change whereby drawdowns occur simultaneously, this would have a detrimental effect on any fund performance. The algorithms have failsafe mechanisms whereby each instrument will discontinue trading for a season if 3 simultaneous losses occur in that instrument. As previously mentioned the risk for each trade is limited to 2% but the results may not reflect this as any profits made by a trade are not included in this maximum risk factor. IE if a trade makes a 2% profit without reaching a target then goes on to hit a 'stop' the drawdown on equity and unbanked profits will be 4%.
The following four charts detail the individual back-test PL and the effect on the "fund" as each is added.
Fund performance as each security is added, starting with AUDJPY then adding AUDUSD and so on.
The drawdown effect of each security is a measure of how much hedging was derived from the instrument. Of particular note is gbpusd where drawdown Increased from -16% to -10% when it was added.
The charts below are the PL curves for each of the eighteen Cassandra instruments, as in the combined test the starting equity for each individual test was 100k.
ultra low frequency trading
the most boring trading system ever developed