Aether Analytics Strategy Deck

ther Analytics
Available Quantitative Strategies

Long-Short Quantitative Machine Learning
MESA AMG
Adaptive Cycle Trend
Futures Trading System
MESA AMG Futures Strategy Mechanics

Comprised of two equity (ES & YM) and two US treasury futures (ZN & ZB) contracts
Multi strategy system that monitors for specific market regimes then applies the highest quality signal
Signals aim at ANTICIPATING market turns rather than reacting to market prices
Positions are held for profit according to the evolving market state. If the market begins to cycle it will take
profits at a cycle swing. If the market begins to trend it will hold for the maximum duration of that trend.
Signals generated at end of day for entry on the next day at the market open
There are no resting stop orders for this system but there are emergency stop loss orders that will cover all
positions if the market has a extreme price movement.
Strategy Description
Our algorithmic market model uses digital signal processing (DSP) techniques to separate futures contract
prices into three distinct components of cycle, trend, and noise. Proprietary DSP filters and signal processing
algorithms are used to extract, de-trend, amplify, and isolate each component using MESA techniques originally developed for the radar and aerospace industries by John F. Ehlers.
The dominant market cycle waveform is prepared from scientific measurements to determine the cyclic amplitude, frequency, and phase. From this, the waveform is phase-shifted into the future to anticipate the next
cyclic peak or trough. Cycle peaks and troughs are further analyzed and anticipated based on a component that incorporates market vigor as a precursor to short-term cyclic reversals.
MESA AMG Futures Strategy Mechanics

Market cycles vary in amplitude and duration, and are ephemeral. In short-term trading, it is essential to
quickly determine whether a tradable cycle is present and its characteristics to exploit any market inefficiencies that may be present. Traditional techniques such as Fourier analysis are inadequate in this environment,
primarily because they lack resolution. Proprietary MESA algorithms are used as they offer superior performance over classical technical analysis techniques. In a low signal to noise environment such as that produced by market data, MESA typically provides better accuracy with reduced lag resulting in a higher percentage of winning trades.
There are also periods of time when the trend or noise components effectively swamp the cyclical component. Thus our algorithm includes a cycle mode versus trend mode switch to a take different approach in
cases where the cycle is deemed to be ineffective. Cyclic entries are taken counter to the trend while trend
entries are based on momentum in the direction of the trend. In cases where the noise measurement is excessive, the system remains out of the market.
Many if not most algorithmic trading systems are trend-followers. In the past, trend-following systems have
been effective but due to their widespread usage, trend-followers have been rendered less effective in recent years. Trend-following algorithms typically use breakout detection in their approach. With breakout detection, the trading system samples a trade each time a presumed trend begins. The idea is to be willing to
take a larger number of small losses and that are made up for by catching and riding a big trend every once
in a while. Our trend detection algorithm is unique in that we use the cyclic component to determine a trend
to cycle ratio. Whenever this ratio is above a threshold, the system is in trend mode. The trading system thus
provides a unique strategy for trend detection coupled with a cycle detection algorithm and a switch to determine when to trade the cycle and when to trade the trend.
Hypothetical equity curve
Hypothetical Backtest & Walk Forward Out of Sample Returns
Jan
2016 12.63%
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
6.33%
Max DD
6.8%
1.93% -2.93%
1.65% 23.02%**
11.24%
5.94% -0.31%
5.74%
4.29% 39.60%**
10.88%
4.76%
3.27% -0.79% -0.35%
0.00% -5.84%
5.62%
6.88% -2.76%
7.19%
2014
4.36%
4.92%
0.33%
5.05%
1.27%
0.15% -1.91%
2013 -0.46%
0.10%
5.72%
0.53%
0.81% -2.58%
6.53%
2012
1.28% -3.36% -1.93%
0.95%
2.55%
3.37% -2.41%
7.18% 10.37% -9.36%
2011 -0.24%
4.80%
1.50%
2.82%
7.52% 11.21%
5.65%
2010 -1.58%
8.99% -0.19%
1.01% -2.07%
2.83%
2009 -1.33%
0.21%
8.88%
7.20%
2008
3.27% -2.63%
5.69%
2007
5.13%
1.65% -0.55%
2006
0.00%
0.36%
7.82%
Return
20%**
2015
4.80%
Dec
3.03% -0.17%
7.34%
27.04%
13.10%
3.40%
16.48%
8.54%
7.36%
64.99%
16.24%
3.19% -0.87% 11.34% -0.04%
1.91% -1.34%
24.66%
9.81%
3.70% -7.08%
6.02%
0.51%
5.82% -6.01%
22.36%
12.67%
4.79%
0.91%
9.34%
4.59%
3.16% -0.21% 13.51% 10.56%
2.04%
69.42%
18.12%
0.73%
0.45% -3.76%
5.25%
2.00%
2.82%
31.66%
11.53%
0.82%
5.38%
2.39% -0.55% -0.57%
1.33% -2.02%
7.73%
7.85%
32%
11.12%
3.23%
2.75% -2.56%
0.40%
7.19%
4.41%
2.43% -2.20%
4.13% -0.30%
5.58%
5.53%
3.43%
average expected monthly re-
2.38%
Aver-
average up month
4.24%
Max
69.99%
18.12%
-2.07%
Min
7.73%
6.8%
average down month
** 20142015 return are 100% out of sample
Risk Management
Assumptions of Risk Calculation
Through Monte Carlo Analysis the most likely profit is divided by five times the two sigma drawdown (worst
case scenario max loss), to produce expected percent gain. The reason for five times the drawdown is that
5 * 20 = 100 and we don't want to exceed a 20% drawdown at any point. Note our method of calculating
risk is EXTRODINARILY conservative and does not even include compounding in the returns. Essentially the
system is designed to maximize return without ever breaching the 20-25% max drawdown limit.
50,000 Monte Carlo Run Summary

Probable Annual Drawdown : $12,100 / 12.1%
Reward to Risk : 2.8 : 1
1 Sigma Drawdown : $18,260 / 18.26%
2 Sigma Drawdown : $27,000 / 27.5%
Profit Expectancy
Assumptions of Profit Expectancy
Through Monte Carlo Analysis the most likely profit is divided by 5 times the 2 sigma drawdown (worst case
scenario max loss), to produce expected percent gain. The reason for 5 times the drawdown is that 5 * 20 =
100 and we don't want to exceed a 20% drawdown at any point. Note our method of calculating risk is
EXTRODINARILY conservative and does not even include compounding in the returns. Essentially the system is
designed to maximize return without ever breaching the 20-25% max drawdown limit. Position sizing is
assumed to be 1 contract each : ES, TY, US, YM Per $100,000 starting capital
50,000 Monte Carlo Run Summary

Probable Annual Profit : $33,670 / 33.67%%
Reward to Risk : 2.8 : 1
1 Sigma Profit : $52,300 / 52.3%
2 Sigma Drawdown : $73,000 / 73%
Sector Rhythm
Smart Sector Rotation
LongShort SPDR ETFs
Sector Rhythm - Strategy Mechanics

Quantitative model that predicts cyclical turning points in the 9 Sector SPDR Etfs
Most market models are formed from linear system theory. Our market model is one of a deterministic signal
waveform with additive white noise. The technical analysis problem using this model is simple and straightforward, removing the noise by filtering or smoothing. What was left is free of noise and therefore must be the
true signal. Then, having the true signal, all one has to do is to properly interpret it to create algorithmic rules
for a profitable trading system.
A casual examination of daily and weekly charts shows they appear the same if the labels are removed. The
resulting implication is that the amplitude of the cyclic swings in the markets are inversely proportional to frequency. The generalized 1/F , or Pink Noise is ubiquitous across all physical phenomena. This observation in
market data is hardly novel.; Mandelbrot described it as self-replicating fractals Fibonaccians describe the
growth rate of the logarithmic spiral as 1.618.; the Hurst coefficient attempts to measure the growth rate.
Most generally, Pink Noise amplitude double every time the wavelength is doubled. The shorthand notation
for this growth rate is that the noise spectral power grows 6 dB per octave.
While the core of the strategy relies on a directional model to determine the short term price movement of
securities, we employ additional models to optimize returns while managing risk.
Three distinctly different models are employed

Each direction model is optimized for performance on each sector ETF
Directional Models are dynamically selected, based on the slope of there recent equity curve divided by
there drawdown
Sectors are traded independently, but net long and net short positions are limited.
Sector Rhythm - Strategy Mechanics

Noise With Memory
1/F noise is often called Noise With Memory. This characterization certainly fits market data because intraday
traders remember the opening price, yesterdays daily range, etc. It works on every scale. Most traders remember what happened to prices in 2008. Noise with memory perspective applies to other fields as well.
Stealth aircraft are derived from the work of Peter Swerling who described classes of fluctuating target scattering models to characterize the performance of the pulsed radar systems. The noise was explained as independent reflections from different parts of the airplane that added together in a noise-like way as the airplane changed aspect angle relative to the radar transmitter. Ehlers has built a deception radar jammer that
could look like any aircraft from an F-117 to a B-52 using a white noise source followed by a filter that is the analog equivalent of an Exponential Moving Average (EMA).
Application to Trading Strategies
Two key design requirements arise from the Pink Noise structure of market data: First, a filter must be employed to equalize the 6 dB per octave growth of the cyclic component amplitudes so that all cycles can be
processed equally. Second, the DSP signal processing of the data must be conducted in the frequency domain to avoid distortions that are introduced by common technical indicators. Trading signals are statistical
in nature, providing the optimum timing for predicting a reversion to the mean.
Hypothetical Backtest Equity Curve
Hypothetical Backtest Statistics
* Includes 1.5% annual Management fee and 15% performance (High Water Mark)
IncubationLive Account Performance
Aether Equities
Long Only Mean Reversion
US and EU Stocks
Aether Equities - Strategy Mechanics

Aether Equities is an algorithmic portfolio of quantitative strategies for short-term trading of best of breed
large-cap equities, ETFS or Equity Options. The strategy utilizes proprietary digital signal processing (DSP) techniques to measure each stocks dominant market cycle and generate analytical waveforms to forecast shortterm turning points on a ticker-by-ticker basis. The strategy produces end-of-day long position buy/sell trading
signals for the next market-on-open. The maximum number of open positions and new positions per day are
configurable. During times of heightened volatility trades can be hedged to mitigate risk. In addition the strategy has strict risk management controls that monitor over all macro market conditions, and when conditions
are deemed to risk to trade it automatically reduces risk or intermittently goes into a 100% cash position. Aether Equities is also scalable to European and Asian Equites.
Holds a Maximum of 32 open positions at any 1 time, but can be scaled up to unlimited positions
Delays the entry of those positions over multiple days to prevent clumping
Attempts to utilize capital at 80% or higher at all times
Trades are entered at the open of the day, to ensure maximum liquidity and minimum slippage
Strategy was built 100% on OUT OF SAMPLE / WALKFORWARD methodology.
This prevents the possibility of Curve Fitting or Lookback Bias
Performance results over a ten year time frame shows robust performance of the MESA Strat93 trading system under various market conditions. The trading system is identical for all ticker symbols. The system made
very few trades during the market collapse of late 2008 through early 2009.
System cannot predict Black Swan events with brief unexpected volatility but it has internal RISK aversion
components that will shut off signals if it has data showing possible high volatility down moves (Prolonged
Bear Markets)
Hypothetical Backtest Equity Curve
Hypothetical Backtest Statistics
Reflects Reg T 2:1 leverage
Performance includes 1.5% management fee and 15 % performance fee ( high water mark)
With 2:1 leverage max drawdown is still less than 25%
For more information contact

Address : Montecito, California
Telephone: +1.252.805.7161
Email: +alex@aetheranalytics.com
Website : Gaviota Capital Management

Aether Analytics Strategy Deck

Uploaded by

Copyright:

Available Formats

Aether Analytics Strategy Deck

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aether Analytics Strategy Deck

Uploaded by

Copyright:

Available Formats

ther Analytics

Available Quantitative Strategies

MESA AMG Futures Strategy Mechanics

MESA AMG Futures Strategy Mechanics

Hypothetical equity curve

Hypothetical Backtest & Walk Forward Out of Sample Returns

3.27% -0.79% -0.35%

1.28% -3.36% -1.93%

7.18% 10.37% -9.36%

3.19% -0.87% 11.34% -0.04%

3.16% -0.21% 13.51% 10.56%

2.39% -0.55% -0.57%

average expected monthly re-

average down month

** 20142015 return are 100% out of sample

50,000 Monte Carlo Run Summary

50,000 Monte Carlo Run Summary

Sector Rhythm - Strategy Mechanics

Three distinctly different models are employed

Sector Rhythm - Strategy Mechanics

Hypothetical Backtest Equity Curve

Hypothetical Backtest Statistics

IncubationLive Account Performance

Aether Equities - Strategy Mechanics

Attempts to utilize capital at 80% or higher at all times

Strategy was built 100% on OUT OF SAMPLE / WALKFORWARD methodology.

This prevents the possibility of Curve Fitting or Lookback Bias

Aether Equities - Strategy Mechanics

Aether Equities - Strategy Mechanics

Hypothetical Backtest Equity Curve

Hypothetical Backtest Statistics

Reflects Reg T 2:1 leverage

With 2:1 leverage max drawdown is still less than 25%

For more information contact

Website : Gaviota Capital Management

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.