Model-Based

Trading Strategies
Financial-Hacker.com
Johann Christian Lotter / [email protected]
oP group Germany GmbH
  All trading systems herein are for education only.
 No profits are guaranteed.
 Don‘t blame me for losses.

U.S. Government Required Disclaimer - Commodity Futures, Trading

Commission Futures, Derivatives and Options trading has large potential
rewards, but also large potential risk. You must be aware of the risks and be
Disclaimer willing to accept them in order to invest in the futures and options markets.
Don't trade with money you can't afford to lose. This website is neither a
solicitation nor an offer to Buy/Sell futures or options. The past performance of
any trading system or methodology is not necessarily indicative of future
results.
CFTC rule 4.41 - Hypothetical or simulated performance results have certain
limitations. Unlike an actual performance record, simulated results do not
represent actual trading. Also, since the trades have not been executed, the
results may have under-or-over compensated for the impact, if any, of certain
market factors, such as lack of 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
profit or losses similar to those shown.
 Modelling the Market

 Model based vs. data mining strategies

Agenda  The order book model
 The random walk model
 What we can learn from the random walk
 The general price curve model
 Model vs. Data

Modelling  Two general approaches to strategy building:

the Market model based - data mining
 Model based system development starts with a market theory
and attempts to find it reflected in the data.
 Data mining system development starts with the price data
and attempts to find predictive patterns or rules.
 Model = Simplified image of the
reality
Modelling
the Market  We describe the trader behavior with a market model.
 Problem: A model is NOT the reality. The reality is unknown.
 The same reality can be described with many different models.
 The best model must be selected by experiment.
 The „order book“ model
Order book:
 Ask 10k @ 1.03

Modelling  Ask 20k @ 1.02

 Ask 10k @ 1.01
the Market
 Bid 20k @ 0.99
 Bid 10k @ 0.98
 Bid 10k @ 0.97

Broker: Price = 1.01, Spread = 0.02

 You buy 10k at market
 Ask 10k @ 1.03
 Ask 20k @ 1.02

Modelling  (Ask 10k @ 1.01 <- order filled at 1.01)

 Bid 20k @ 0.99
the Market
 Bid 10k @ 0.98
 Bid 10k @ 0.97

Broker: New price = 1.02, Spread = 0.03

-> A buy order pushes the price up, a sell order pushes it down
 You buy 10k at market
Someone sells 10k at market
 Ask 10k @ 1.03
 Ask 20k @ 1.02
Modelling  Ask 10k @ 1.01 (your order filled at 1.01)
the Market  Bid 20k @ 0.99 (other order filled at 0.99)
 Bid 10k @ 0.98
 Bid 10k @ 0.97

Broker: New price = 1.01, Spread = 0.02

-> buy and sell orders cancel each other
 The Random Walk model

Modelling
the Market
Buyers
Sellers
Price = PreviousPrice + Buyers - Sellers
 Which price curve is real?

Modelling
the Market
 Two rules from the random walk
model
Modelling Rule 1: A pure random walk curve can not be traded
the Market (Rule of No Roulette System)

Rule 2: The volatility of a random walk curve is proportional to the

square root of its duration
(Rule of Square Root Volatility).
 Forces, pulling at the price curve
 Previous price (~ 99%)
 Random buyers and sellers
Modelling  Fundamental buyers and sellers
the Market  Technical buyers and sellers

Inefficiency = systematic
deviation from the Random Walk
 General price curve model

Modelling
the Market
𝑦𝑡 = 𝑦𝑡−1
+ 𝑓(𝑦𝑡−1 , … 𝑦𝑡−𝑛 )
+ 𝜀𝑓 + 𝜀𝑟
 Exploiting Market Inefficiencies
 Momentum
 Mean Reversion
 Cycles
 Stat Arb
Agenda  Constraints
 Clusters
 Patterns
 Gaps
 Seasonality
 Heteroskedasticity
 Momentum

𝑦𝑡 = 𝑦𝑡−1
Exploiting
Market
+ 𝑎1 (𝑦𝑡−1 − 𝑦𝑡−2 )
Inefficiencies + 𝑎2 (𝑦𝑡−2 − 𝑦𝑡−3 )
+…
+ 𝜀𝑓 + 𝜀𝑟
 A simple momentum strategy
Exploiting
 Detect the market regime: trend or mean reversion?
Market
 Get a trend line with a lowpass filter
Inefficiencies  When market regime is trending:
 Enter long on a trend line valley
 Enter short on a trend line peak
 Mean Reversion

Exploiting 𝑦𝑡 = 𝑦𝑡−1
Market 1
Inefficiencies − (𝑦𝑡−1 − 𝑦)
ො
𝜆
+ 𝜀𝑓 + 𝜀𝑟
 A simple mean reversion strategy
Exploiting
 Detect the market regime: trend or mean reversion?
Market
 Remove trend with a highpass filter
Inefficiencies  When market regime is mean reverting:
 Enter short when the price exceeds a high threshold
 Enter long when the price falls below a low threshold
 Cycles

𝑦𝑡 = 𝑦𝑡−1
2𝜋
+ 𝑎1 𝑠𝑖𝑛( 𝑡 + 𝑑1 )
𝑐1
Exploiting
2𝜋
Market + 𝑎2 𝑠𝑖𝑛( 𝑡
𝑐2
+ 𝑑2 )
Inefficiencies +…
+ 𝜀𝑓 + 𝜀𝑟

𝑎1 = Amplitude of the first cycle

𝑐1 = Bar period of the first cycle
 Frequency spectrum of a price
curve

Exploiting
Market
Inefficiencies
 A simple cycle strategy
Exploiting
 Detect the dominant cycle 𝑐1 and phase 𝑑1 .
Market
 Get the current amplitude of the dominant cycle.
Inefficiencies  When amplitude is above a threshold:
 Enter short when the phase 𝑑1 is approaching a sine peak.
 Enter long when the phase 𝑑1 is approaching a sine valley.
 Statistical Arbitrage

Exploiting
Market 𝑦𝑡 = ℎ1 𝑦1 − ℎ2 𝑦2
Inefficiencies
𝑦𝑡 = price difference (mean reverting)
ℎ1 , ℎ2 = hedge factors
Typically ℎ1 = 1, ℎ2 by linear regression of 𝑦1 , 𝑦2
 Price constraints
Exploiting
Market  Price is restriced by an upper or lower hard boundary
Inefficiencies  Or price is strongly mean reverting outside a soft boundary
 Classical example: The Swiss Franc cap 2011-2015
 But distant price constraints exist for most assets
 A simple price constraint strategy
(„Grid Trader“)
 Place lines at equal or increasing distances from a mid price.
 Whenever the price crosses a line:
Exploiting  Close all open trades that are in profit.
Market  Open a new long and short trade if there isn‘t already one open at that line.
 Use a hedging method for avoiding open long and short positions at the same
Inefficiencies time.

Possible problems:
 Low short-term volatility
 Trading costs – especially rollover
 Exceeding boundaries -> margin call
 Price clusters
Exploiting
Market
 Where do prices concentrate?
Inefficiencies  „Support and Resistance“ -> two clusters
 „Fair price“ -> one cluster
 Curve patterns

Exploiting  Not to be confused with „Candle Patterns“

Market  Some famous patterns – such as „Head and Shoulders“ – have no

significance in real price curves and are probably myths.
Inefficiencies  Other patterns – such as „Cups“ and „Half-Cups“ – really exist and
can be explained by a behavior model („breakout“).
 Several algorithms for detecting curve patterns, f.i. the Frechèt
algorithm.
 Gaps

Exploiting
Market
Inefficiencies  Overnight and weekend gaps can „amplify“ and „synchronize“
trader behavior patterns
 Trend and mean reversion before the gap reappears „with a
revenge“
 A simple gap trading strategy

Exploiting
 On an upwards trend, buy long on Friday when a 10-days high is
Market reached.
Inefficiencies  On a downwards trend, buy short on Friday when a 10-days low is
reached.
 Close the position on Monday morning.
 Live trading can be followed on the Zorro forum.
 Seasonality
 Trader behavior depends on time of day, day of week, day of month,
month of year
 Seasonal effects in a price curve can be detected by simple statistical
methods
Exploiting
Market
Inefficiencies
 Heterosketasticity

Exploiting 𝑦𝑡 = 𝑦𝑡−1 +
Market
Inefficiencies 2
𝜀𝑡 𝑎 + 𝑏(𝑦𝑡−1 − 𝑦𝑡−2 )
GARCH model (Generalised Autoregressive Conditional
Heteroskedasticity)
 The Development Process
1) Selecting the model. Confirming it with price data
2) Developing the trade algorithm
Agenda 3) Developing the filter algorithm
4) Parameter adaption („optimizing“)
5) Test
6) Reality check
7) Implementing risk and money management
 Step 1: Model selection
The
Development The three prerequisites for a financial model:
Process 1) Has a rational basis in market structure / trader behavior
2) Can be expressed in a program flow or formula
3) Has statistical significance in real price curves
 Confirming the model

The
Development  Find an algorithm that detects the inefficiency in price curves.
Process  Do a statistic. Plot a histogram.
 Compare with random walk curves or shuffled price curves.
Difference should be significant.
 Do NOT rely on other people‘s research! Scam is ubiquitous
(-> „Elliott Waves“, Rich Swannell)
 Example: Frequency spectrum of
a price curve

The
Development
Process
 Step 2: Determining the
algorithm
The
Development Example: Cycle strategy
 Detect the dominant cycle and phase.
Process
 Generate a forerunning sine curve.
 Enter short at a sine peak.
 Enter long at a sine valley.
 Exit on reversal or after a half-period.
 Step 3: The filter

The A market inefficiency normally does not exist all the time.
Therefore, we need a filter for determining if the inefficiency is
Development present or not. In most cases the filter is more important than the
algorithm.
Process
Example: Cycle strategy
 Measure the amplitude of the dominant cycle.
 Trade only when the amplitude is above a threshold.
 Step 4: Parameter adaption
(„Training“)
If the model has „free parameters“:
The  Find out how the strategy reacts on parameter changes.
Development  Find the most robust parameter range („sweet spot“).
Process  Adapt the strategy to different assets.
 Adapt it to different market situations (even while live trading).

Bad ideas:
- Optimizing too many parameters.
- Optimizing for peaks (= brute force or genetic optimization).
 Example (Cycles system)
Adapted parameters: sine phase and threshold
Training results:
The
Development
Process
 Step 5: Test

The
Development
Process  Test should cover all significant market periods (5-10 years)
 Any parameter adaption introduce bias to the test result.
 The bias renders backtests completely useless.
 The solution: Testing the system with data not used for the
adaption.
 Walk-Forward Analysis
 Roll a window over the simulation period
 Separate the window in a training and test section.
 Good: The test is out-of-sample and still covers most of the data.
The  Bad: The system depends on two more parameters.
Development
Process
 Analyzing test results

The Main performance parameters:

Development  Wins divided by losses (Profit Factor)
Process  Annual profit in relation to drawdown (Calmar ratio)
(Drawdown must be normalized -> square root rule!)
 Annual return in relation to sigma (Sharpe ratio)
 Linearity of returns (R2 coefficient)
 Monte Carlo method

For eliminating „randomness“ from the test results:

The  Split the equity curve in small sections
Development  „Randomize“ the curve by shuffling without replacement
Process  Repeat 1000 times.
 Calculate test results from every shuffled curve.
 Sort test results by confidence intervals.
 Step 6: Reality check
Even with walk forward and Monte Carlo analysis, test results still
suffer from bias. Bias is introduced by the mere development
process. Several methods to detect bias:

The 1) White‘s Reality Check: gives a quantitative measure of bias

Development 2) Monte Carlo Reality Check: run the system with price curves
randomized by shuffling with replacement. Plot a result
Process distribution
3) Variants: run the system with inverted, detrended, or
oversampled price curves
4) Real-out-of-sample test: Set aside a part of the data and only
use it for this test.
 White‘s Reality check

The
Development
Process

Details under:
http://www.financial-hacker.com/whites-reality-check/
 Step 7: Risk and Money
Management
• Use a stop loss for eliminating negative outliers.
The • Do not use profit targets. (If you really want to, use a profit-lock
Development mechanism instead).
Process • Use an algorithm for calculating the optimal investment per
portfolio component (Kelly, OptimalF, Markowitz).
• Re-invest only the square root of your profits.
• Supervise your system permanently and compare live results with
backtest results (-> „Cold Blood Index“).
 No Reinvestment

The
Development
Process

CAGR: 15%
 1% Reinvestment

The
Development
Process

CAGR: 16%
 0.5 OptimalF Reinvestment

The
Development
Process

CAGR: 48%
 Sqrt(P) OptimalF Reinvestment

The
Development
Process

CAGR: 43%
 So far the theory…
Here‘s the real development process
Step 1. Visit trader forums. Look for the thread with new fabulous indicator.
Step 2. Implement the indicator with a long coding session. Ugh, the backtest
The does not look this good. Some coding mistake? Debug. Debug some more.
Step 3. Still no good result, but you have more tricks up your sleeve. Add a
Development trailing stop. Run a week analysis. Tuesday is a bad day for this indicator? Add a
filter for not trading on Tuesday. Add more filters for not trading between 12:00
Process and 14:00 and on any full moon except on Thursday. Wow, now we see some
backtest profit!!
Step 4. Of course you’re not fooled by in-sample results. After optimizing all 23
parameters, run a walk forward analysis. Ugh, the result does not look this good.
Try different WFA cycles. Try different bar periods. Optimize more parameters.
Finally, a sensational test result! And this completely out of sample!
Step 5. Trade the system live.
Step 6. Ugh, the result does not look this good.
Step 7. Hold many trading seminars for recovering your bank account.