-
in this course we will study both investment and speculation
The main difference between investment and speculation lies in the time horizon.
Investment is concerned with capturing maximum returns in the long-run with lower risk, while speculation is concerned with achieving returns over a short period of time.
Speculation attempts to switch between investments to achieve the best return versus risk.
Investment attempts to choose the best investments and hold onto them for longer periods, to achieve the best return versus risk over the long term.
http://blogs.wsj.com/moneybeat/2015/12/24/this-simple-way-is-the-best-way-to-predict-the-market/ -
get Andrew Ang book Asset Management: A Systematic Approach to Factor Investing
The exclamation !!! marks signify very good papers
!!! good course bullet points:
http://www.londonfs.com/programmes/Modern-Asset-Allocation-Portfolio-Construction/Outline/
!!! ECON 424/CFRM 462: Computational Finance and Financial Econometrics
C:\Research\R\Tutorials\Zivot
!!! C:\Research\R\Tutorials\Zivot\research
!!! Bloch ebook Quantitative Portfolio Management.pdf
!!! Tourin dynport.zip
!!! Cochrane Advanced Investments
C:\Research\Academic\Cochrane Advanced Investments
!!! Pfaff Package Development in R.pdf
-
suggest to Josh Ulrich fixing HurstIndex() in PerformanceAnalytics
HurstIndex() calculates range of returns, instead of range of prices (cumulative returns)
HurstIndex() doesn't use OHLC data
suggest adding to OHLC Hurst-like technical indicators to TTR -
Piotr Orlowski high frequency option models
Adapt code from package affineModelR as template for RCPP
https://github.com/piotrek-orlowski/affineModelR -
study Kozhan paper
!!! Kozhan Skew Variance Swap Stock Forecasting.pdf -
create simple quantstrat scripts
http://timelyportfolio.blogspot.com/2011/06/reit-momentum-in-quantstrat.html -
reproduce factors results
http://timelyportfolio.blogspot.com/2014/04/all-factors-more-looks.html
http://timelyportfolio.blogspot.com/2014/04/exploring-factors-with-rcharts-and.html -
create R scripts for loading Reuters data
-
source ideas from lectures https://www.quantopian.com/lectures
-
download OHLC data from Quandl
-
adjust stock prices
https://www.quandl.com/blog/guide-to-stock-price-calculation -
Maintaining a database of price files in R
http://www.thertrader.com/2015/12/13/maintaining-a-database-of-price-files-in-r/ -
convert all synthetic t-series to log-normal process
-
Yollin tutorial rDataAccess
-
recreate CRSPpanel.txt fundamental financial data for 265 S&P 500 stocks
trellis plots
nice barchart, dotplot, bwplot, and data munging
Estimating risk and performance measures: volatility, skew, CVaR, Cornish-Fisher VaR, Modified VaR, risk-return ratios (Sharpe, Sortino, Calmar), package PerformanceAnalytics,
-
Normal (Gaussian) distribution is a bad model for price returns, because extreme returns are frequent and determine the mean
Risk and total return determined by a small number of data points
For example, half of the return of the stock market over the past 50 years was associated with just 10 days with the greatest daily change (Taleb) ?
if cumulative returns are mostly generated by a few large returns, then focusing on predicting small returns is a waste of time
bin the returns according to their magnitude
what are the cumulative returns for each return magnitude bucket?
are the cumulative returns for small return magnitude almost zero most of the time?
study the following statistic: cumulative returns of bucket divided by its magnitude
large magnitude buckets should have very high variability because they have smaller number of elements -
Matthieu Lestel article for PerformanceAnalytics reviews risk-return measures
PerformanceAnalytics PA-Bacon.pdf -
measures of dispersion:
STDEV
Maximum or median absolute deviation
trailing range statistic
Hurst exponent
Show that ratio of STDEV over MAD is related to kurtosis -
stock process is not stationary
calculate rolling/running dispersion and moment estimators using vectorized functions: package caTools,
show that stock volatility is time-dependent -
create study of bias-variance tradeoff using volatility estimation example:
http://scott.fortmann-roe.com/docs/BiasVariance.html
create xts of random prices with changing time-dependent deterministic vol parameter,
estimate volatility use look-back window parameter,
too short look-back window increases variance,
too long look-back window increases bias,
tune filter parameters in-sample: study bias-variance tradeoff,
create rCharts and shiny visualizations -
estimate volatility using OHLC data
demonstrate that estimator standard error is lower using OHLC data
use bootstrap to determine estimator standard error confidence intervals
use simulated data with constant volatility -
range OHLC volatility estimation
http://eranraviv.com/intraday-volatility-measures/
http://eranraviv.com/multivariate-volatility-forecasting-2/
http://eranraviv.com/multivariate-volatility-forecasting-3-exponentially-weighted-model/
http://eranraviv.com/multivariate-volatility-forecasting-5-orthogonal-garch/
Bennett review Range Volatility Estimators.pdf
Brandt OHLC Range Volatility Estimators.pdf
Chou Range OHLC GARCH Volatility Estimators.pdf
Bencik Range OHLC HAR GARCH Volatility Estimators.pdf
Yang OHLC Range Volatility Estimators.pdf -
tail risk measures
value-at-risk and conditional value-at-risk as function of skewness and kurtosis parameters
show that value-at-risk is not subadditive
subadditive risk measures, ETL (ES/ETL/CVaR), Omega, Hurst exponent,
conditional value at risk (CVaR)
VaR for generalized Pareto distribution
!!! Maillard Cvar Cornish Fisher Portfolio.pdf
http://www.capitalspectator.com/tail-risk-analysis-in-r-part-i/
https://gist.github.com/jpicerno1/c3af6285713c76a5d124 -
CVAR has a bigger standard error than VAR, and is therefore useless for very large number of risk factors estimated using short span of data
Jon Danielsson and Chen Zhou have demonstrated that to accurately estimate CVAR at 5% confidence would require decades of price history, something that simply doesn't exist for many assets.
Danielsson CVAR Estimation Standard Error http://www.bloomberg.com/view/articles/2016-05-23/big-banks-risk-does-not-compute -
create Normal mixture model and show that it has fat tails,
show that Normal mixture model is similar to t-distr
create distribution with large skew - Poisson -
demonstrate that a small change in the alpha parameter (less than its mean error) changes the CVaR by large amount (plot the relationship)
https://edge.org/conversation/nassim_nicholas_taleb-the-fourth-quadrant-a-map-of-the-limits-of-statistics -
stochastic volatility models: CEV, GARCH, Engle's ARCH volatility clustering permutation test http://www.burns-stat.com/documents/tutorials/the-statistical-bootstrap-and-other-resampling-methods-2/ demonstrate the persistence of volatility (autoregression) in real return data
fit GARCH model into real return data
define GARCH model using OHLC data, and demonstrate its better performance
simulate stochastic volatility process
show that when volatility is stochastic then STDEV is much higher than MAD
create plot displaying ratio of STDEV over MAD as function of volatility of volatility
http://stackoverflow.com/questions/9969962/simulation-of-garch-in-r?rq=1
https://en.wikipedia.org/wiki/Stochastic_volatility
http://jonathankinlay.com/index.php/2011/03/long-memory-and-regime-shifts-in-asset-volatility/
http://www.jonathankinlay.com/articles/Modeling%20Asset%20Volatility.pdf
http://jonathankinlay.com/index.php/2011/03/regarch-option-pricing-models/ -
simulate Heston model and calibrate it to S&P returns use package NMOF
http://stackoverflow.com/questions/15579655/heston-simulation-monte-carlo-slow-r-code
http://stackoverflow.com/questions/27429725/monte-carlo-simulation-in-r
http://stackoverflow.com/questions/15534270/stock-price-simulation-r-code-slow-monte-carlo -
Pareto distribution as model for stocks
fit returns into Student t-distribution, Cauchy, and Pareto distribution
Show that the Pareto distribution has infinite variance but has finite MAD.
Shalizi: pareto.R
http://edge.org/response-detail/25401
Capital Asset Pricing Model CAPM: market portfolio, regressions of asset returns, alpha, beta, CML, SML, package factorAnalytics,
-
factor model and investing explained !!! Cazalet CAPM Factor Models.pdf
Fernandez CAPM Stock Model Review.pdf
Black CAPM Empirical Tests.pdf
Steiner Alpha Misleading Performance Measure.pdf
Ardia CAPM Portfolio Optimization Stock Forecasting.pdf -
calculate rolling/running beta
calculate beta confidence intervals using bootstrap
http://eranraviv.com/bootstrap-example/
http://statistics.ats.ucla.edu/stat/r/library/bootstrap.htm
Fox Regression Bootstrap.pdf
C:\Research\R\Packages\returnanalytics\pkg\PerformanceAnalytics\R\FamaBeta.R -
beta robust regression shrinkage-estimator-for-beta
http://eranraviv.com/a-shrinkage-estimator-for-beta/ -
lm() Model Variable selection
shrinkage methods
AIC, AIC, BIC
update() -
define Capital Market Line (CML) and Security Market Line (SML)
-
Markowitz’s Critical Line Algorithm (CLA) - function CCLA()
http://rnfc.org/2015/06/05/Markowitz/ !!! Bailey Prado Critical Line Algorithm Portfolio Selection -
Equity premium puzzle
Returns on stocks are much higher than predicted by CAPM model using volatility of equity returns and returns on government bonds
The fact that stocks are riskier than bonds doesn't explain the magnitude of the difference,
https://en.wikipedia.org/wiki/Equity_premium_puzzle -
define beta-adjusted risk-return measures
Treynor ratio, Jensen's alpha
Matthieu Lestel article for PerformanceAnalytics reviews risk-return measures
PerformanceAnalytics PA-Bacon.pdf -
Jensen alpha examples of how alpha can be generated: timing market and ex-post portfolios
calculate Jensen alpha for SPX (for example) and demonstrate that it's close to zero
calculate Jensen alpha for timed SPX: buy SPX at lows and sell at highs
calculate Jensen alpha for ex-post portfolio: optimize portfolio in-sample to obtain highest alpha -
Ormos Entropy Asset Pricing Model.pdf
-
Lewellen: momentum is cross-sectional ranking, meaning winners outperform losers Autocorrelation is longitudinal ranking, meaning past performance will continue Is it possible to have zero autocorrelation, but non-zero momentum?
-
Distinguish between cross-sectional regressions (in-sample), and predictive regressions (out-of-sample),
-
Statistical factor models examine returns over many time periods, and from them identify relationships between and among the different assets, unlike fundamental factor models, which from the outset group assets that are likely to experience similar returns.
-
Fama and French three-factor model tutorial
http://www.bogleheads.org/wiki/Fama_and_French_three-factor_model
https://www.bogleheads.org/wiki/Fama-French_three-factor_model_analysis
https://www.bogleheads.org/wiki/CAPM_-_Capital_Asset_Pricing_Model
https://www.bogleheads.org/wiki/Category:Financial_theory
http://www.capitalspectator.com/portfolio-analysis-in-r-part-v-risk-analysis-via-factors/
http://jonathankinlay.com/index.php/2015/03/combining-momentum-mean-reversion-strategies/ -
Zivot Factor Models
C:\Research\R\Tutorials\Zivot\research\factorModels.r
Zivot Factor Models.pdf -
qmj factor package David Kane Hutchin Hill
portfolio.pdf
portfolio Vignette.pdf
https://github.com/anttsou/qmj
https://github.com/anttsou/qmjdata -
AQR data library
https://www.aqr.com/library
https://www.aqr.com/library/data-sets/quality-minus-junk-factors-monthly
http://stackoverflow.com/questions/28031008/reading-an-online-xlsx-file-into-r -
Cochrane asset pricing
C:\Research\Academic\Cochrane Advanced Investments
Cochrane_asset_pricing_CH12_229-250.pdf
https://github.com/shabbychef/coursera_ap2013 -
Fama-MacBeth two-pass regressions to explain cross-sectional returns/values by factors
Campbell Market Factors Stock Forecasting.pdf
http://quant.stackexchange.com/questions/16855/how-to-test-the-5-factor-capm-of-fama-french-2014
http://quant.stackexchange.com/questions/17125/please-give-a-step-by-step-explanation-on-how-to-build-a-factor-model http://quant.stackexchange.com/questions/8697/r-fast-and-efficient-way-of-running-a-multivariate-regression-across-a-really factorAnalytics fitTsfm_vignette.pdf -
Equity Factors with Principal Component Analysis
http://www.calculatinginvestor.com/2013/03/01/principal-component-analysis/
http://www.calculatinginvestor.com/2013/03/18/pca-factors-vs-fama-french-factors/
http://www.calculatinginvestor.com/octave-code/calculating-fama-french-loading/ -
Fama-French three factor model viewer
http://systematicinvestor.wordpress.com/2012/06/20/factor-attribution/
https://systematicinvestor.wordpress.com/category/factor-model/ -
Factor Attribution
https://systematicinvestor.wordpress.com/2012/06/ -
Barra and Northfield factor models
https://systematicinvestor.wordpress.com/2012/02/21/multiple-factor-model-building-risk-model/
https://systematicinvestor.wordpress.com/2012/01/29/multiple-factor-model-fundamental-data/ -
timely factors
http://timelyportfolio.blogspot.com/2014/04/exploring-factors-with-rcharts-and.html
http://timelyportfolio.github.io/rCharts_factor_analytics/factors_with_new_R.html -
PCA factors
CFA Stock Premium Factors.pdf
CFM Principal Component Market Factors.pdf
scatterplot of two stocks
rotate axes to get PComps
PCA: don't scale factors
PCA: stock regressed against PCA factors
bootstrap PCA analysis to obtain distribution of loadings
https://tgmstat.wordpress.com/2013/11/28/computing-and-visualizing-pca-in-r/
apply ADF test to higher order PC's to demonstrate that higher order PC's are more stationary
http://fabian-kostadinov.github.io/2015/01/27/comparing-adf-test-functions-in-r/
http://www.statmethods.net/advstats/factor.html
http://davetang.org/wiki/tiki-index.php?page=Principal+component+analysis
CFM Principal Component Market Factors.pdf
Alexander Principal Component Multivariate GARCH Model.pdf -
PCA as eigenvectors by hand
http://eranraviv.com/multivariate-volatility-forecasting-4-factor-models/
https://en.wikipedia.org/wiki/Eigendecomposition_of_a_matrix -
PCA Principal Components and Clustering
http://systematicinvestor.wordpress.com/2012/12/22/visualizing-principal-components
http://systematicinvestor.wordpress.com/2012/12/29/clustering-with-selected-principal-components
http://systematicinvestor.wordpress.com/2013/01/17/optimal-number-of-clusters -
Survey of multi-factor models
https://factorinvestingtutorial.wordpress.com/
https://github.com/hkuhn/multifactor-models
https://github.com/hkuhn/multifactor-models/blob/master/src/6-pca-factor-model/pca-factor-model.R -
data mining for factors
create synthetic time series with correlations using factors, and extract the factors
how much data is needed to obtain accurate estimates?
create synthetic time series without correlations
show that bogus factors are extracted with high t-values
perform bootstrap to reduce data mining for factors
Harvey Factor Model Data Mining Bonferroni Adjustment.pdf
Harvey Bootstrap Factor Models.pdf -
significant PCA factors
http://gestaltu.com/2015/11/tactical-alpha-in-theory-and-practice-part-ii-principal-component-analysis.html
Guttman (1954) and Kaiser (1960, 1970), asserted that in order to be significant, "a factor must account for at least as much variance as an individual variable" (Nunnally and Bernstein, 1994).
Horn proposed that factors should only be considered significant if they explain a greater proportion of variance than what might be expected from random chance. -
factorAnalytics package vignettes
https://r-forge.r-project.org/scm/viewvc.php/pkg/FactorAnalytics/vignettes/?root=returnanalytics
https://r-forge.r-project.org/projects/factoranalytics/
http://rpackages.ianhowson.com/rforge/factorAnalytics/
http://timelyportfolio.blogspot.com/2014/08/famafrench-factors-in-1-line-of-code.html
https://github.com/R-Finance/FactorAnalytics/blob/master/vignettes/fundamentalFM.Rnw -
factorAnalytics fundamental factor model
fitFundamentalFactorModel is is now called fitFfm
?fitFfm
factorAnalytics YiAnChen.pdf
https://r-forge.r-project.org/scm/viewvc.php/pkg/FactorAnalytics/?root=returnanalytics
http://r.789695.n4.nabble.com/factoranalytics-vs-factoranalyticsuw-td4709951.html -
factorAnalytics fundamental data and returns for 447 NYSE stocks
stock (Stock.df) Fundamental and return data for 447 NYSE stocks
stock is used by function fitFfm() ?Stock.df
ls("package:factorAnalytics")
data(package="factorAnalytics")
factors.M (CommonFactors) Factor set of several commonly used factors
factors.Q (CommonFactors) Factor set of several commonly used factors
managers Hypothetical Alternative Asset Manager and Benchmark Data
r.M (StockReturns) Stock Return Data
r.W (StockReturns) Stock Return Data
tr.yields (TreasuryYields) Treasury yields at different maturities -
use following data:
data(package="factorAnalytics")
C:\Research\R\Packages\factorAnalytics\extdata
data(package="GARPFRM")
data(package="mpo") -
factorAnalytics and rCharts
http://timelyportfolio.blogspot.com/2014/04/exploring-factors-with-rcharts-and.html -
BARRA factor model with factorAnalytics
https://github.com/BradGalton/R-Factor-Models/blob/master/Barra%20Industry -
factorAnalytics fundamentalFM.Rnw - BARRA fundamental factor model - can't find pdf
https://github.com/R-Finance/FactorAnalytics/blob/master/vignettes/fundamentalFM.Rnw -
factor data mining: 24 return factors forecast monthly stock returns
Green Factor Models Stock Forecasting.pdf
quote:
Fama and French (1992, FF92) measured the dimensionality of the cross-section of expected monthly U.S. stock returns by regressing the potential factors beta, firm size, book-to-market, earnings-to-price and leverage.
They found that beta was not explanatory of expected returns, but size and book-to-market were, and that they absorbed the explanatory power of
earnings-to-price and leverage.
FF92 concluded that the cross-section of expected monthly U.S. stock returns was two-factor, although neither factor was consistent with the CAPM.
A third factor in the form of 12-month return momentum (Jegadeesh, 1990;
Jegadeesh and Titman, 1993) was incorporated by Fama and French (1996) and
Carhart (1997) to create the three-factor model of risks that explain equity returns. -
Green: most predictive factors of monthly stock returns from Green Factor Models
use standardized unexpected earnings
bm book-to-market
mom12m 12-month momentum
mom36m 36-month momentum
chfeps Change in forecasted annual EPS
ear three-day return centered on the most recent earnings announcement
sfe the ratio of forecasted annual earnings-to-price
rsup quarterly sales growth
indmom 12-month industry return momentum
turn Share turnover
dolvol trading volume in month t-2
rsup Revenue surprise
mve log of market cap at month-end immediately prior to signal date -
factor data mining and stock forecasting additional sources for above Green Factor Models
Lewellen Factor Models Stock Forecasting.pdf
Harvey Factor Model Data Mining Bonferroni Adjustment.pdf
Pukthuanthong Ranking Factor Models Stock Forecasting.pdf -
demonstrate that the time variation of factors may lead to spurious evidence of additional risk factors
-
multiple regression
variance inflation factor for multicollinearity in explanatory variables regression analysis
https://en.wikipedia.org/wiki/Variance_inflation_factor
LASSO reduces artefacts from multicollinearity of explanatory variables -
factor model regularization (shrinkage)
Regularization (shrinkage) is an example of Occam’s Razor, which was postulated by the fourteenth-century philosopher Sir William of Ockham.
Occam’s Razor states that the most likely solution to be correct is the simplest solution, and that any solution should not be more complicated than necessary ("the law of parsimony").
-
show that it's easier to forecast returns over longer horizons in the future
compare forecasts of daily returns, weekly, monthly, annual, etc. using past returns over different horizons - weekly, monthly, annual, etc.
perform apply loops over different horizons
It's easier to forecast long-term returns (over next decade) than short-term returns (over next year), because long-term returns are determined mostly by economic fundamentals, while short-term returns are determined mostly by speculative returns
Similarly, it's easier to forecast the weather over a longer term (next Summer) than over a shorter term (next week), but it's also easier to forecast the weather over a very short term (next day) -
forecast a simulated time series
simulate a time series of returns using the Vasicek or Heston models
create a linear forecasting model of returns
evaluate forecasting performance using various measures: MSE, sign of forecast versus realized returns, etc.
use bootstrap to obtain a distribution of forecasting performance ?
define objective function based on distribution of measures ?
calibrate the forecasting model parameters to maximize the objective function
demonstrate that Hurst measures the level of forecastability
plot Hurst as function of model parameters -
Using the LASSO to Forecast Returns
http://www.alexchinco.com/using-the-lasso-to-forecast-returns/ -
Bias in Time-Series Regressions
http://www.alexchinco.com/bias-in-time-series-regressions/ -
simulate GARCH model and forecast volatility
demonstrate forecasting ability as function of GARCH parameters
Goyal GARCH Volatility Forecasting.pdf -
stochastic volatility and rebalancing - solve Hamilton-Jacobi-Bellman equation
Goyal Cross Sectional Factors Stock Forecasting.pdf
Rapach Equity Stock Forecasting.pdf
DeMiguel VAR Model Stock Selection Forecasting.pdf
correlation forecasting - is it possible? -
is Hurst exponent forecastable?
calculate running Hurst over sliding interval - is Hurst persistent?
calculate Hurst for different assets, and sort them
are stock indexes more forecastable than individual stocks? -
Forecasting returns using momentum factor
Vogel Absolute Momentum Stock Forecasting.pdf
Gulen Absolute Momentum Stock Forecasting.pdf -
Forecasting returns conditional on volatility
Interaction between returns and volatility
Sort stocks by volatility and test which deciles have highest momentum
Vogel Volatility Momentum Stock Forecasting.pdf -
forecast returns using volatility-adjusted momentum - just like Sharpe ranking
Calculate volatility-adjusted momentum rankings by dividing the prior twelve month total return by the realized volatility over the same period and then ranking in the standard fashion.
Clare Volatility Momentum Trend Following Asset Allocation.pdf
Baltas Volatility Momentum Trend Following Asset Allocation.pdf
Zakamulin Momentum Indicators Stock Forecasting.pdf -
demonstrate negative correlation between the monthly return of S&P index versus monthly volatility of returns on the index
unexpected volatility is the difference between the realized volatility minus the GARCH forecast
unexpected volatility predicts future excess return and volatility
two strategies that dynamically reallocate between stocks and the risk-free asset, depending on the value of unexpected volatility.
Zakamulin Volatility Forecasting Asset Allocation.pdf
Vogel Absolute Momentum Stock Forecasting.pdf -
Testing forecasting accuracy using Diebold Mariano Test
package forecast dm.test() for Diebold Mariano Test ROC curve
Diebold Mariano Forecast Accuracy Test.pdf
C:\Research\R\Tutorials\Zivot\Econ 584\dieboldMariano.pdf
http://stats.stackexchange.com/questions/139462/diebold-mariano-test-for-predictive-accuracy
http://stats.stackexchange.com/questions/143079/what-is-prediction-accuracy-auc-and-how-is-it-the-number-conducted-in-machi?rq=1 -
measuring predictive ability using bootstrapping
Hansen Forecasting Cross-Validation Bootstrap.pdf
Hansen test improves on White's Reality Check for Data Snooping
Hambuckers Forecasting Cross-Validation Bootstrap.pdf
http://thestatsgeek.com/2014/10/04/adjusting-for-optimismoverfitting-in-measures-of-predictive-ability-using-bootstrapping/ -
Ian Kaplan (UofWash) Value Factors Do Not Forecast Returns for S&P 500 Stocks
http://www.bearcave.com/finance/thesis_project/
http://www.bearcave.com/finance/etf2/index.html
Kaplan Constructing ETF Portfolio.pdf
Kaplan Value Factor Model Forecast Returns.pdf -
demonstrate that beta and correlations are difficult to forecast
calculate rolling/running beta
forecast beta out-of-sample, and show it doesn't work
The only thing we can do is to short correlation at +1 and buy it at -1.
beta and correlations are as difficult to forecast as returns -
steady momentum frog-in-the-pan indicator: number of winning periods minus number of losing periods
steady momentum indicator should be related to skew: large gains in a short period should produce positive skew -
Stroud C programs for forecasting returns, variance, skew, kurtosis,
!!! Stroud High Frequency Forecasting Volatility VIX VXX Strategy.pdf
http://www.jonathanrstroud.com/code.html -
forecasting intraday returns after price jumps
Zawadowski Intraday Reversal Stock Forecasting.pdf
Grant Intraday Reversal Stock Forecasting.pdf
Duyvesteyn Intraday Reversal Bond Forecasting.pdf
Schneider Skew Fear Volatility Risk Premium Forecasting.pdf -
implied variance and skew forecast realized variance and skew
!!! Kozhan Skew Variance Swap Stock Forecasting.pdf
Kozhan Skew Variance Swap Stock Forecasting SSRN.pdf
Mijatovic VIX Market Factors Stock Forecasting SSRN.pdf -
VIX squared minus the five-minute realized variance forecast stocks
volatility risk premium forecasts stocks
does it demonstrate negative relationship between volatility and future return ?
Bollerslev Volatility Stock Forecasting.pdf
Bollerslev Implied Realized Volatility Stock Forecasting.pdf -
Hull kitchen sink forecasting
Hull Indicators Stock Forecasting.pdf
Hull Tactical US ETF (HTUS)
http://www.thestreet.com/story/13349919/1/will-this-quant-based-eft-be-able-to-time-the-market.html -
Neely combine fundamental and technical indicators
Neely Indicators Stock Forecasting.pdf
Rapach Short Interest Indicator Stock Forecasting.pdf
https://sites.google.com/site/xiaoqiao10/
http://blog.alphaarchitect.com/2015/02/23/can-you-predict-stock-market-returns-with-short-interest/#.VPEI3EtN3wJ
http://www.superforecasting.com/asset-return-forecasting/ -
Do valuation ratios forecast stock returns ?
The Campbell-Shiller identity connects current dividend yield to future returns, dividend growth, and dividend yield
Campbell Stock Forecasting.pdf
add constraints on coefficients to improve forecasting out-of-sample R2 is positive
If R2 is large relative to S2, then an investor can use the information in the predictive regression to obtain a large proportional increase in portfolio return -
changes in the analyst rankings of P/E ratios forecasts stocks
Gray Price Earnings Ratio Stock Forecasting.pdf -
Kakushadze Alpha Forecasting
Kakushadze Factor Model Stock Alpha Forecasting.pdf
Kakushadze Factor Models Alpha Streams.pdf -
yield curve forecasting example
http://eranraviv.com/yield-curve-forecasting/ -
Kalman filter
http://bilgin.esme.org/BitsBytes/KalmanFilterforDummies.aspx
http://intelligenttradingtech.blogspot.com/2010/05/kalman-filter-for-financial-time-series.html
http://stats.stackexchange.com/questions/8055/how-to-use-dlm-with-kalman-filtering-for-forecasting
http://www.magesblog.com/2015/01/extended-kalman-filter-example-in-r.html
Arnold Kalman Filter Expectation Maximization.pdf
Sorensen Kalman Filter.pdf -
Kalman filter
Prado Kinetic Component Analysis Forecasting.pdf
Backtesting: cross-validation, parameter regularization, model overfitting, data snooping, data mining,
-
verbiage
cross-validation in machine learning is also called backtesting or walk forward analysis
https://en.wikipedia.org/wiki/Backtesting
terminology: formation period or lookback window (in-sample), evaluation period (out-of-sample) -
backtesting rule of thumb is to accept a strategy only if out-of-sample (OOS)
dictates that a system passes cross-validation if OOS performance is greater than 50% of in-sample (IS) performance. -
two resampling methods: cross-validation and bootstrap
C:\Research\R\Tutorials\Stanford Statistical Learning\cv_boot.pdf
C:\Research\R\Tutorials\Zivot\Econ 424\bootStrapPowerPoint.pdf
C:\Research\R\Tutorials\Shalizi Advanced Data Analysis\which-bootstrap-when.pdf -
beta rolling regression: illustrate variance-bias tradeoff
http://scott.fortmann-roe.com/docs/BiasVariance.html -
create backtesting system for strategies - walk forward analysis
!!! Peterson Developing Backtesting Systematic Strategies.pdf
system should recalibrate forecasting model at lower frequency than the base data frequency
system should apply forecasting model at base data frequency -
quantstrat examples
https://timtrice.github.io/backtesting-strategies/
https://quantstrattrader.wordpress.com/2014/09/09/nuts-and-bolts-of-quantstrat-part-i/
https://quantstrattrader.wordpress.com/2014/09/16/nuts-and-bolts-of-quantstrat-part-ii/
https://quantstrattrader.wordpress.com/2014/09/20/nuts-and-bolts-of-quantstrat-part-iii/
https://quantstrattrader.wordpress.com/2014/09/24/nuts-and-bolts-of-quantstrat-part-iv/
https://quantstrattrader.wordpress.com/2015/09/03/introduction-to-hypothesis-driven-development-overview-of-a-simple-strategy-and-indicator-hypotheses/ https://quantstrattrader.wordpress.com/2015/09/09/hypothesis-driven-development-part-ii/
https://github.com/milktrader/A-Mustering-of-Storks/tree/master/demo
calculate turnover per year,
measure timing performance, -
create synthetic time series with autocorrelation and drift, and perform trend following strategy
Nilsson Momentum Trend Following.pdf
tune model parameters in-sample: study bias-variance tradeoff,
study profitability as function of model parameters -
ARIMA GARCH strategy
!!! Halls-Moore ARIMA GARCH Strategy.pdf -
portfolio backtest packages Jeff Enos and David Kane
https://github.com/dgerlanc -
strategery package strategy workflow, backtesting, optimization
https://github.com/danielkrizian/strategery/ -
packages ttrTests and fTrading for backtesting Technical Trading Rules
White Strategy Backtesting Overfitting Data Mining Cross-validation Bootstrap.pdf
http://www.inside-r.org/packages/cran/ttrTests/docs/dataSnoop
https://www.linkedin.com/in/david-st-john-96798745 -
backtest with properly accounting for adjusted prices
http://systematicinvestor.github.io/Backtest-Reality-Check/ -
create rebalancing strategy for portfolio with two assets: stock index plus bond index
http://www.capitalspectator.com/portfolio-analysis-in-r-a-6040-us-stockbond-portfolio/
create portfolio return scatterplot and show that it's negatively skewed, because rebalancing strategy is equivalent to selling put options
demonstrate that a strategy using asset rebalancing to maintain constant market value adds risk
Granger Portfolio Rebalancing Momentum Trend Following.pdf
Qian Asset Allocation Portfolio Rebalancing Alpha.pdf
Qian Asset Allocation Portfolio Rebalancing.pdf -
GARCH volatility forecast trading
http://systematicinvestor.wordpress.com/2012/01/06/trading-using-garch-volatility-forecast/ -
rugarch ARFIMA ugarchroll rolling estimation forecasting
https://theaverageinvestor.wordpress.com/2011/12/15/more-orthodox-armagarch-trading/ -
ARMA Models for Trading - ARIMA backtesting
http://www.quintuitive.com/2012/08/22/arma-models-for-trading
http://www.quintuitive.com/2012/12/27/armagarch-experiences/
http://www.quintuitive.com/category/research/armagarch/ http://www.quintuitive.com/2013/03/24/automatic-armagarch-selection-in-parallel/ -
SVM Models for Trading
http://www.quintuitive.com/category/research/svm/ -
distinguish single-asset strategies from cross-sectional strategies
ranking and sorting stocks -
Probit Model
http://www.capitalspectator.com/a-partial-solution-for-narrative-risk-probit-modeling/
https://gist.github.com/jpicerno1/f8307d0c14cc39d25c53 -
XIV VXX ZIV strategies by Harry Long
http://seekingalpha.com/article/2616495-a-weird-all-long-strategy-that-beats-the-s-and-p-500-every-year-ii
http://seekingalpha.com/article/2627145-a-refined-all-long-strategy-iii -
Ilya Kipnis backtesting Harry Long strategies
http://quantstrattrader.wordpress.com/2014/10/08/structural-arbitrage-a-working-long-history-backtest/
http://quantstrattrader.wordpress.com/2014/11/02/its-amazing-how-well-dumb-things-get-marketed/
http://quantstrattrader.wordpress.com/2014/11/03/seeking-volatility-and-leverage/ -
parameter heatmaps
http://quantstrattrader.wordpress.com/2014/11/19/trading-the-odds-volatility-risk-premium-addressing-data-mining-and-curve-fitting/ -
if VIX contango (VXV/VXMT<1), then long XIV (short VIX) - if VIX backwardation (VXV/VXMT>1), then long VXX (long VIX)
http://quantstrattrader.wordpress.com/2014/12/04/a-new-volatility-strategy-and-a-heuristic-for-analyzing-robustness/ -
same as above but delayed execution
http://quantstrattrader.wordpress.com/2014/12/10/an-update-to-the-robustness-heuristic-and-a-variation-of-a-volatility-strategy/ -
VIX strategies
Donninger VIX Futures Skew Strategy.pdf
Mixon VIX Futures Markets Review.pdf -
VIX strategies
http://systematicinvestor.github.io/TradingTheOdds/
http://quantstrattrader.wordpress.com/2014/12/12/the-zomma-warthog-index/ -
volatility risk premium heatmap
http://quantstrattrader.wordpress.com/2014/11/19/trading-the-odds-volatility-risk-premium-addressing-data-mining-and-curve-fitting/
http://volatilitymadesimple.com/chasing-the-volatility-risk-premium/ -
volatility VXX XIV shiny app
https://alphaminer.shinyapps.io/VolaStrat/ -
create option selling strategy with position limits:
double down as option premiums increase (this will increase position size)
reduce position size when market turns
this is a path-dependent strategy -
create strategy using as inputs Treasury curve, VIX curve, stock momentum, etc.
Gayed Treasury Curve Anomaly Asset Allocation.pdf -
create strategy that forecasts returns over longer horizons
should it rebalance over longer horizons too ?
what if monthly returns are forecast but rebalancing is performed daily ?
is it better to forecast monthly returns and rebalance daily,
instead of forecasting daily returns and rebalancing daily ?
compare best strategies over different horizons - change horizon parameter
you can use three parameters: lookback window, forecast horizon, and rebalancing frequency -
trading strategy with stochastic volatility
Johannes Stock Forecasting Intertemporal Universal Portfolios Choice.pdf -
create examples of strategies with great performance despite low Sharpe ratios (trend following?) and vice versa (mean reversion?)
better Sharpe ratio:
http://bettersystemtrader.com/sharpe-ratio-right-answer-wrong-question/ -
distinction between long volatility strategies and short volatility strategies
mean reverting strategies have natural profit caps (exit when price has reverted to mean) but no natural stop losses (we should buy more of something if it gets cheaper), so it is very much subject to left tail risk, but cannot take advantage of the unexpected good fortune of the right tail.
On the contrary, momentum strategies have natural stop losses (exit when momentum reverses) and no natural profit caps (keep same position as long as momentum persists). -
create league table of indicators from package TTR
Green Indicators Stock Forecasting.pdf
Zakamulin Indicators Stock Forecasting.pdf -
check for data snooping (leaking or look ahead bias) in backtest
propagate price spike in backtest
feed random data into backtest -
data mining (synonyms significance inflation, multiple testing), and false discovery rate
https://en.wikipedia.org/wiki/Look-elsewhere_effect
create example of data mining: create tech indicator with several parameters
http://datagrid.lbl.gov/backtest/
http://www.financial-math.org/software/ -
controlling the false-discovery rate using Bonferroni method Sidak correction
http://www.alexchinco.com/screening-using-false-discovery-rates/
http://eranraviv.com/sample-data-snooping/
http://eranraviv.com/modern-statistical-discoveries/
Bailey Prado Deflated Sharpe Ratio Overfitting.pdf
Bailey Prado Strategy Backtesting Overfitting Cross-validation.pdf
Bailey Prado Strategy Backtesting Overfitting.pdf
Harvey Backtesting Data Mining Bonferroni Adjustment.pdf
Harvey Evaluating Trading Strategies.pdf
White Strategy Backtesting Overfitting Data Mining Cross-validation Bootstrap.pdf -
add transaction costs
Donninger Overnight Momentum Seasonal Anomaly.pdf
$12.5 broker fee per trade ($25 round-trip)
for Nasdaq the money-value of 1 tick is $5 - 1 tick is also the typical bid-ask spread for NQ
for ES E-mini futures the money-value of 1 tick is 12.5$ -
metaparameter data mining increases false-discovery rate
momentum indicators cross-validation for determining optimal filter parameters
!!! Bruder Momentum Indicators Kalman Filter SVM.pdf
Wojtow Momentum Trend Following.pdf
create heatmap of model parameters using expand.grid -
perform grid search of model parameters on heatmap
find max and min and indices with arr.ind=FALSE
http://sebastianraschka.com/Articles/heatmaps_in_r.html
http://stackoverflow.com/questions/8421536/a-true-heat-map-in-r
http://stat.ethz.ch/R-manual/R-patched/library/stats/html/heatmap.html
http://hosho.ees.hokudai.ac.jp/~kubo/Rdoc/library/gplots/html/heatmap.2.html
http://digitheadslabnotebook.blogspot.com/2011/06/drawing-heatmaps-in-r.html -
plot grid heatmaps 3d charts
use random data to show that optimal in-sample parameters are not best out-of-sample, because of noise in-sample
show that the best out-of-sample parameters depend on level of noise
with lower noise level, the best out-of-sample parameters is closer to the optimal in-sample parameters -
System Parameter Permutation (SPP):
we don't know the appropriate model parameters for the best future performance, so run the model using all possible parameter permutations, to obtain a distribution of possible future performance.
SPP provides distribution of model performance
worst-case contingencies must be tolerated in order to achieve the long-run expectation?
Walton Backtesting Overfitting Parameter Permutation.pdf -
calculate histogram of performance on model parameters
the median serves as the best estimate of future strategy performance out-of-sample
The median performance can be used as the best estimate of future strategy performance -
data resampling: divide the data into non-overlapping intervals
for each interval find the optimal in-sample strategy parameters
create distribution of optimal in-sample strategy parameters from all the intervals, and calculate parameter standard error
calculate distribution of strategy performance using the distribution of parameters
the distribution of performance provides an estimate of possible future strategy performance
summarize the distribution of strategy performance (mean and SD ?)
this summary provides an estimate of the strategy's usefulness potential
create benchmark random data with the same moments as the real data - shuffle real data ?
compare the distribution of strategy performance for real data with that for random data
random data should produce high mean performance but also very high SD of parameters
random data should have slightly higher mean (?) but much higher SD
create single performance measure defined as excess performance over random benchmark:
decrease in SD minus wieghted decrease in mean (decrease in SD wieghs more than decrease in mean) -
quantile optimization with data resampling: repeat the above, but instead of optimal in-sample strategy parameters, select median (quantile) parameters
for each interval find the strategy parameters corresponding to the median (quantile) performance
apply those parameters to calculate the strategy performance for all the data out-of-sample
calculate the distribution of out-of-sample strategy performance
which quantile corresponds to the best out-of-sample strategy performance? -
in each interval select the parameters corresponding to a certain performance quantile
apply the quantile parameters in other intervals, and collect performance distribution
this way create performance distributions as function of performance quantile
find the performance quantile with the best performance distribution - how does it change with level of noise ?
this should demonstrate that the best in-sample parameters are far from the best out-of-sample parameters
modify the above by using all the parameters in the full quantile (tail), instead of just the ones at the cutoff -
perform model ensemble backtesting exercise
perform backtesting, but at each step don't select the optimal parameters, because the optimal model likely outperforms because it is fit to noise
instead select several models with parameters corresponding to a certain performance quantile (ensemble)
the choice of performance quantile depends on the level of noise -
the performance quantile with the best performance distribution corresponds to a distribution of parameters - which is an ensemble of models
does this ensemble of models also have the best backtesting results ?
this approach using ensembles of models defines a new backtesting paradigm -
Perform futility analysis to determine if model is close to random
Futility Analysis versus random benchmark (in clinical trials)
When is enough data collected to conclude that a model is close to random ?
Answer: when the likelihood that it's not random is very small
http://onbiostatistics.blogspot.com/2012/03/futility-analysis-in-clinical-trials.html
random prices as benchmark
https://en.wikipedia.org/wiki/Sequential_analysis !!! Kuhn Parallel Adaptive Resampling.pdf
!!! Kuhn Futility Analysis Cross-Validation Machine Learning Models.pdf -
caret package by Max Kuhn for creating predictive models
caret Predictive Modeling.pdf
caret Predictive Modeling.R
http://caret.r-forge.r-project.org/
http://appliedpredictivemodeling.com/data/
http://stackoverflow.com/users/1078601/topepo -
Determine stop-loss policy parameters using sequential hypothesis testing
Sequential hypothesis testing -
mean reverting strategies
Blitz Short-Term Residual Reversal.pdf -
overfitting genetic programming models
http://fabian-kostadinov.github.io/2015/01/14/evolving-trading-strategies-with-genetic-programming-punishing-complexity/ -
backtesting fallacies
http://www.philosophicaleconomics.com/2015/12/backtesting/ -
get R code from Krauss
Krauss Piotroski Score Value Stock Strategy.pdf -
many quant indexes have underperformed compared to backtests Beware of optimistic results of backtesting hypothetical investment strategies! Past backtesting results are no guarantee of future performance, as they say. http://www.bloomberg.com/news/articles/2016-01-21/how-wall-street-finds-new-ways-to-sell-old-opaque-products-to-retail-investors
-
equity risk premium anomalies Antti Ilmanen: The equity risk premium (ERP) refers to the expected return of a broad equity index in excess of some fixed-income alternative.
Arnott (Research Affiliates):
The ERP Puzzle: Stocks beat bonds by more than they should.
Historical excess returns exhibit large negative correlation.
The correlation between consecutive 10-year stock market excess returns over 10-year government bonds has been a whopping –38 percent.
When stocks beat bonds by a wide margin in one decade, they reversed with
reasonable reliability over the next decade.
This correlation is both statistically significant and economically meaningful. -
pricing anomalies papers !!! Bouchaud Momentum Volatility Market Anomalies.pdf Fama French Dissecting Anomalies.pdf
C:\Research\Academic\Cochrane Advanced Investments\new_anomalies.pdf
Vogel Factor Model Momentum Anomaly.pdf
CFM Momentum Trend Following Strategy Anomaly.pdf
Han Trend Factor Cross-Section Momentum Stock Returns.pdf
Israel Size Value Momentum Anomalies.pdf
DeBondt Stock Premium January Anomaly.pdf -
Asness anomalies
Asness Fama French Small-Cap Anomalies.pdf
Asness data files in: Asness*.xlsx -
Anomalies aren't persistent
Edwards Market Anomaly Smart Beta Persistent Spurious.pdf -
Low volatility anomaly
Boudt Low Volatility Anomaly High Frequency Data.pdf Gray Low Volatility Anomaly.pdf
Baker Low Volatility Anomaly.pdf
Li Low Volatility Anomaly FAJ.pdf
Han Volatility Decile Cross-Sectional Momentum Anomaly.pdf -
Low beta anomaly caused by demand for positive skewness (lottery) which reduces future returns
Bali Betting Against Beta Lottery Demand.pdf -
Jacobs: momentum anomaly enhanced by skewness
skewness enhanced momentum is about twice as large as traditional momentum
skewness is among the most important cross-sectional determinants of momentum
Jacobs Skewness Cross-Sectional Momentum Anomaly.pdf
Amaya Skewness Momentum Equity Returns http://www.etf.com/sections/index-investor-corner/swedroe-keep-skewness-perspective -
Show that the returns of momentum strategies have negative skewness: momentum strategies have positive returns but also experience infrequent but significant negative returns http://blog.alphaarchitect.com/2015/05/11/momentum-investing-skewness-enhanced-momentum-yields-double-alpha/#gs.YexM_xM
-
Schneider: CAPM betas overestimate true market risk
demonstrate that if asset value follows a lognormal process, then the equity price in Merton model has a skewed distribution of returns
demonstrate that equity returns in Merton model have positive skewness, since they are a call option
high credit risk produces time-varying skewness in Merton model
demonstrate that high credit risk produces time-varying skewness in Merton model
Schneider Volatility Anomaly Skew Risk Premium.pdf
Schneider Skew Anomaly Merton Credit Risk Forecasting.pdf -
Ang Idiosyncratic Volatility Anomaly.pdf
R code to replicate main results in Ang, Hodrick, Xing, and Zhang (2006)
https://gist.github.com/alexchinco/d58ebd7750904db1b94c
https://gist.github.com/alexchinco
https://github.com/alexchinco -
Treasury Curve Anomaly
Gayed Treasury Curve Anomaly Asset Allocation.pdf -
Value strategies can be implemented in many different ways, leading to widely different performance
http://investorfieldguide.com/three-value-investors-meet-in-a-bar/
Stock value can be measured in several different ways including book value, earnings, and sales.
The Russell 1000 Value has underperformed the Russell 1000 by -22% and the Russell 1000 Growth by -43% over the past decade (10 years ending 11/30/15).
idea: apply value investing to different value indices: buy more of the cheap ones
-
demonstrate sell in May anomaly by subsetting S&P returns
Bouman Sell in May Halloween Seasonal Anomaly.pdf
Afik Sell in May Halloween Seasonal Anomaly
Matilde Sell in May Halloween Seasonal Anomaly.pdf
Dzhabarov Seasonal Anomalies.pdf -
Cieslak Market Timing FOMC Calendar Seasonal Anomaly.pdf
-
daily overnight seasonal anomaly
buy ES in last 30min, and sell ES next morning in first 30min - trade only on reversals
Gray Lou Overnight Momentum Seasonal Anomaly.pdf
Lou Overnight Momentum Seasonal Anomaly CAPM Factor Models.pdf
Cliff Overnight Momentum Seasonal Anomaly.pdf
Donninger Overnight Momentum Seasonal Anomaly.pdf
http://jonathankinlay.com/index.php/2015/11/overnight-trading-in-the-e-mini-sp-500-futures/
http://www.priceactionlab.com/Blog/2015/11/overnight-trading-anomaly-backtesting-r/
http://blog.fosstrading.com/2015/11/overnight-spy-anomaly.html
http://systemtradersuccess.com/overnight-edge/
http://systemtradersuccess.com/market-seasonality-study/
http://systemtradersuccess.com/seasonality-sp-market-session/ -
daily overnight overreaction gap reversal anomaly
create morning strategy based on open-close (daytime) and close-open (overnight) returns
is there price gap in morning ? what is best rule based on combination of all three returns?
Donninger Intraday Reversal Stock Forecasting.pdf
Kudryavtsev Intraday Reversal Stock Forecasting.pdf
Kudryavtsev abstract Intraday Reversal Stock Forecasting.pdf
-
derive CAPM from utility
Show that logarithmic utility implies max Sharpe -
skew demand causes underperformance, and creates stock premium factor
Ilmanen Buying Selling Insurance Lottery Tickets.pdf
Nekrasov Kelly Criterion Multivariate Portfolios.pdf
Estimation of covariance and correlation matrices, Akaike and Bayesian information criteria, coefficient shrinkage,
-
estimating covariance and correlation matrices
https://en.wikipedia.org/wiki/Estimation_of_covariance_matrices
http://quant.stackexchange.com/questions/44/what-methods-do-you-use-to-improve-expected-return-estimates-when-constructing-a
http://quant.stackexchange.com/questions/10101/portfolio-optimization-shrinkage-of-covariance-matrix-when-data-is-available -
covmat package for asset return correlation matrix estimation
https://github.com/rstats-gsoc/gsoc2015/wiki/Covariance-Matrix-Estimators
https://github.com/arorar/covmat
data(package="covmat") -
correlation covariance estimation and shrinkage
!!! https://bwlewis.github.io/covariance-shrinkage/
http://bwlewis.github.io/covar/missing.html -
correlation parameter uncertainty
demonstrate how correlation parameter uncertainty increases with smaller number of observations or larger number of assets. -
indeterminate correlation matrix
Cholesky decomposition fails when correlation matrix is mis-specified (not orthogonal and not positive definite)
create example of mis-specified matrix and demonstrate how to fix it
correlation matrix estimation error bands
Kwan Correlation Estimation Error.pdf -
SVD and covariance matrix inverse:
inverse of covariance matrix using factors
Karhunen-Loeve Decomposition -
tawny package for regularizinging correlation matrices using random matrix theory and shrinkage estimation
Rowe Random Matrix Shrinkage Covariance Estimation.pdf
Gatheral Random Matrix Shrinkage Covariance Estimation.pdf
Plerou Random Matrix Correlation Estimation.pdf -
package irlba
Lewis RFinance 2012 Cointegration SVD.pdf
C:\Research\R\R-Finance 2015\BryanLewis.html -
Factor Augmented Regression for shrinking correlation matrix
Fit asset returns into multifactor model (start with CAPM),
Fitted asset returns should equal weighted sum of factors plus random uncorrelated residual,
Calculate correlation matrix of the fitted asset returns,
The correlation matrix should depend only on the factor correlations and asset betas,
Chiara Factor Model Forecasting.pdf
Kakushadze Correlation Shrinkage Factor Models.pdf -
demonstrate that the term structure of correlation decreases with tenor
show that correlation depends on time scale, and decreases with shorter time scale
on short time scales correlation is very small
on intermediate time scales correlation is greater
on long time scales correlation is lower
study Lo and MacKinlay variance ratio test in: Kinlaw Variance Ratio Correlation Term Structure.pdf
package vrtest -
package corpcor for estimation of correlation for biostatistics
http://strimmerlab.org/software/corpcor/ -
Multivariate volatility and correlation forecasting DCC GARCH model
http://eranraviv.com/multivariate-volatility-forecasting-1/ -
estimate correlation using OHLC data
Bannouh Range High Frequency Covariance Estimators.pdf Rogers OHLC Range Covariance Estimators.pdf -
introduce the Gerber Statistic
Gerber Statistic Portfolio Optimization.pdf http://nextlevelanalytics.github.io/2016/05/26/Gerber/
-
Efficient Frontier Portfolios
http://zoonek.free.fr/blosxom/R/2012-06-01_Optimization.html
https://gist.github.com/jpicerno1/565be39ca4226ecd004c
http://www.capitalspectator.com/efficient-frontier-portfolios-impractical-but-still-useful/
http://moderndata.plot.ly/portfolio-optimization-using-r-and-plotly/ -
Show that any convex combination of efficient frontier portfolios is also an efficient frontier portfolio.
The efficient frontier consists of convex combinations of any two efficient frontier portfolios. -
Optimizing portfolios under different correlation assumptions
create risk/return scatterplot for portfolios with two assets: stocks plus bonds
create vector of weights and plot line from stocks to bonds
simulate stock and bond returns using different correlations, and study effect on the line
solve for the most efficient portfolios (highest Sharpe) and create plot of bond percentage as function of correlation
create xts plot with slider for bond weight, display how Sharpe ratio changes -
proporties of the Market Portfolio under the CAPM model
the Market Portfolio is assumed to be the optimal portfolio with the highest utility under the CAPM model,
the Market Price of Risk equals the highest Sharpe ratio of the optimal portfolio
How can Market Portfolio be obtained ?
What is Market Portfolio isn't it the same as highest Sharpe portfolio ?
Market Portfolio can be obtained by optimizing Sharpe ratio ?
The Market Portfolio isn't necessarily equal to the cap-weighted portfolio of all assets -
CAPM holds by construction when market portfolio is the efficient frontier portfolio.
When individual stock returns are regressed on the efficient portfolio returns, then their residuals are uncorrelated, because if they weren't then a more efficient portfolio could be constructed.
If the residuals of returns in SML were correlated with each other, then a different Market Portfolio would exist portfolios on the CML satisfy the SML equation
provide reasons why CAPM may not hold?
Can Security Market Line (SML) be derived from Capital Market Line (CML)?
Yes, because if we choose the most efficient portfolio as the reference portfolio for CAPM, then an asset's idiosyncratic returns will on average have mean equal to zero -
"Zivot portfolio.r" from econ424
http://faculty.washington.edu/ezivot/econ424/portfolio.r
Zivot Efficient Portfolios in R
C:\Research\R\Tutorials\Zivot\Econ 424\bootstrapPortfoliosPowerpoint.pdf
C:\Research\R\Tutorials\Zivot\Econ 424\bootstrapPortfolio.R -
portfolio optimization using optim
can mean variance portfolio optimization be converted to min variance optimization ?
library(quadprog)
solve.QP
different objective functions,
constraints -
DEoptim
Ardia DEoptim Portfolio Optimization.pdf
Boudt DEoptim Portfolio Optimization.pdf
Boudt Asset Allocation Conditional Value-at-Risk Budgets.pdf -
use MC and bootstrap to create scatterplot of optimal portfolios or weights, due to parameter uncertainty
place the zero correlation optimal portfolio on that scatterplot to show that it's as good as the optimal portfolio -
portfolio optimization with different objective functions - VaR, CVaR,
Ian Kaplan (UofWash) minimum variance and tangency portfolios, CVaR portfolio optimization, ETF portfolios, Wharton Research Data Service (WRDS) data set and Factor Model Factors
http://www.bearcave.com/finance/
Shaw Portfolio Optimization CVaR Omega Utility.pdf -
Guy Yollin’s "effFrontier" and "maxSharpe" functions use the core function of "portfolio.optim" in the "tseries" R package
http://blog.streeteye.com/blog/2012/01/portfolio-optimization-and-efficient-frontiers-in-r/
C:\Research\R\Tutorials\Guy Yollin Presentations
Levy Alpha Sharpe Portfolio Optimization.pdf -
package NMOF PMwR
Schumann Take the Best Portfolio Selection Heuristic.pdf
portfolio optimization adds no incremental value because correlation forecast error is so large that best to rely on marginal risk for portfolio choice.
pick assets that are good on their own, not for diversification,
simple sorting rules or cutoff rules are likely "more optimal" than is sometimes thought.
C:\Research\R\Packages\NMOF\doc
C:\Research\R\Packages\NMOF\book
DEopt
PSopt
NMOF Portfolio Optimization Threshold Accepting.pdf
quadprog package
solve.QP
TAopt -
nice formulas Global Minimum Variance Weights
http://www.bearcave.com/finance/portfolio_equations/ -
package fPortfolio in Rmetrics for portfolio optimization
-
constrained portfolio optimization shrinkage
http://www.finance-r.com/s/efficient_frontier_fPortfolio/complete/
http://www.finance-r.com/s/simple_portfolio_optimization_tseries/complete/
http://www.portfolioprobe.com/2011/04/28/a-test-of-ledoit-wolf-versus-a-factor-model
http://quant.stackexchange.com/questions/10101/portfolio-optimization-shrinkage-of-covariance-matrix-when-data-is-available
https://systematicinvestor.wordpress.com/2011/11/11/resampling-and-shrinkage-solutions-to-instability-of-mean-variance-efficient-portfolios/
https://systematicinvestor.wordpress.com/2013/10/29/updates-for-proportional-minimum-variance-and-adaptive-shrinkage-methods/
http://quant.stackexchange.com/questions/1504/robust-portfolio-optimization-re-balancing-with-transaction-costs
Golts Constrained Shrinkage Portfolio Optimization.pdf
Demiguel Shrinkage Estimators Portfolio Optimization.pdf
Ledoit Wolf Covariance Shrinkage Estimators Portfolio Optimization.pdf -
SharpeR and MarkowitzR packages by Steven Pav
Sharpe ratio as Hotelling's t-squared distribution
https://github.com/shabbychef
Pav Sharpe Ratio Notes Hotelling Statistic
Pav Strategy Overfit
Pav code for Cochrane Asset Pricing
https://github.com/shabbychef/coursera_ap2013
SharpeR Vignette.pdf
MarkowitzR Vignette.pdf
MarkowitzR AsymptoticMarkowitz.pdf
finding optimal portfolio in-sample is the same as finding optimal strategy in-sample - both are over-fit and require shrinkage
Britten-Jones Sampling Error Mean-Variance Efficient Portfolio Weights
http://blog.fosstrading.com/2014/03/intro-to-portfolioanalytics.html
-
demo_efficient_frontier.R
portfolio object specifies the constraints and objectives for the optimization -
PortfolioAnalytics package for portfolio optimization
PortfolioAnalytics Bennett Random Portfolios Swarm Optimization.pdf
C:\Research\R\Packages\PortfolioAnalytics Bennett
https://github.com/rossb34/PortfolioAnalyticsPresentation2015 -
visualize portfolio optimization
chart.VaRSensitivity
chart.RiskReward(risk.col="StdDev") or (risk.col="ES")
combine portfolios into list and chart: chart.EfficientFrontierOverlay -
create portfolios satisfying combinations of constraints and objectives:
objectives: maxSR, maxSRES, minVAR, minVARES,
constraints: long-only, long-short, neutral, box, leverage (=sum of absolute values of weights),
mean-ES (Expected Shortfall) portfolio
min-ES (Expected Shortfall) portfolio
mean-variance portfolio
mean-variance long-only portfolio
min-variance long-only portfolio
Maximize portfolio mean return per unit of ES/ETL/CVaR=STARR Ratio
ES=Expected Shortfall=Conditional VaR=CVaR=STARR (Stable Tail Adjusted Return Ratio)
method="historical", method="gaussian" or method="modified" -
optimize.portfolio: study and explain effect of choosing different
optimize_method="DEoptim", "random", "ROI",
ROI package=R Optimization Infrastructure -
optimize simultaneously several portfolios with different constraints and objectives
optimize.portfolio.rebalancing -
introduce Hierarchical Risk Parity (HRP), the Gerber Statistic, and the Critical Line Algorithm (CLA)
http://gallery.rcpp.org/articles/HRP/ https://github.com/RcppCore/rcpp-gallery/blob/gh-pages/src/2016-05-27-HRP.Rmd http://nextlevelanalytics.github.io/2016/05/30/Gerber_Statistic_and_Hierarchical_Risk_Parity/ -
Michaud Resampled Efficiency Portfolio Optimization (patented)
https://newfrontieradvisors.com/Research/Articles/MichaudResampledEfficiency.html
https://systematicinvestor.wordpress.com/2011/11/11/resampling-and-shrinkage-solutions-to-instability-of-mean-variance-efficient-portfolios/ -
Random Subspace Optimization is a generalization of the random forest algorithm
https://systematicedge.wordpress.com/2013/10/14/random-subspace-optimization-max-sharpe/
Active portfolio management strategies: out-of-sample performance of optimized portfolios, tactical asset allocation, risk parity, minimum correlation, minimum variance, maximum Sharpe, maximum CVaR, universal portfolios,
-
simulate terminal distribution of stock prices
simulate 500 correlated stocks time series random lognormal with positive drift,
use them for random portfolios
create a value-weighted index
show that cap-weighted index investors are inherently trend-following because index keeps buying more of the outperforming stocks
compare to equally weighted index
which investors perform better?
expand on: cap-weighted indices have large concentrations and undesirable factor exposures to momentum -
demonstrate that active managers are likely to underperform index, unless they have extraordinary skill
http://www.bellmanoptimality.com/programming/ http://www.bellmanoptimality.com/ Heaton Stock Index Selection Active Portfolio Management.pdf -
Grinold fundamental law of active management
Grinold Synopsis Active Portfolio Management.pdf -
Bogle's message is: it's better to invest in indices, unless you're a genius stock picker or a genius speculator.
http://blogs.wsj.com/moneybeat/2015/12/24/this-simple-way-is-the-best-way-to-predict-the-market/
Bogle Investing Factor Models.pdf
Bogle postulates that long-term returns on investments consist of an "investment return" (initial yield plus earnings growth) plus the "speculative return" (discount factor determined by investor psychology and risk appetite).
The cumulative investment return is positive, while the cumulative speculative return is close to zero. -
Merton model: simulate dynamic investment and consumption strategies
Merton Dynamic Consumption and Portfolio Choice
simulate Merton consumption wealth model
Guasoni Merton Optimal Consumption Utility Shortfall Aversion.pdf
An Merton Utility Asset Allocation.pdf
https://en.wikipedia.org/wiki/Intertemporal_portfolio_choice
https://en.wikipedia.org/wiki/Merton%27s_portfolio_problem -
Intertemporal portfolio choice
calculate out-of-sample performance of optimized portfolios,
perform rolling portfolio optimization and study stability of weights over time -
optimize portfolio assuming zero or constant asset correlations
demonstrate that this portfolio outperforms out-of-sample
Sivaramakrishnan Intertemporal Portfolio Choice.pdf
Garleanu Intertemporal Portfolio Choice.pdf -
simulate static asset allocation strategies
all weather portfolios
Faber Arnott Portfolio Asset Allocation.pdf -
Risk Parity Portfolios
!!! Roncalli Risk Parity Factor Models.pdf
Steiner Risk Parity Portfolios.pdf
Griveau-Billion Risk Parity Portfolio Cyclical Coordinate Descent Algorithm.pdf -
simulate CPPI strategy: CPPI strategy is similar to Kelly betting strategy
http://epchan.blogspot.com/search/label/Book%20reviews
the only way to ensure that our maximum drawdown will not exceed a certain limit is through Constant Proportion Portfolio Insurance (CPPI): trading risky assets with Kelly-leverage in a limited liability company, putting money that you never want to lose in a FDIC-insured bank, with regular withdrawals from the LLC to the bank (but not the other way around).
Jacquier Merton Kelly Bayesian Utility Asset Allocation.pdf -
factor investing
Blitz Investing Asset Allocation Factor Models.pdf
Bender Smart Beta Asset Allocation Investing Factor Models.pdf -
Andrew Ang at Columbia book Asset Management: A Systematic Approach to Factor Investing
!!! Ang Factor Models Investing.pdf
http://factorinvestingbook.com/
http://factorinvestingbook.com/book.html
!!! Richard Smart Beta Minimum Variance Factor Models.pdf
Maillard Risk Parity Minimum Variance Portfolios.pdf
Goldberg Value Minimum Variance Portfolio Factor Models
Hsu Minimum Variance Portfolio Factor Models.pdf
Clarke Risk Parity Minimum Variance Portfolios.pdf
Chow Minimum Variance Stock Strategy.pdf
-
Smart Beta doesn't outperform
Jacobs:
Glushkov found the Sharpe ratios of smart beta funds and their benchmarks to be nearly identical, at 0.46 versus 0.48, respectively, while the average information ratio was 0.08, inconsistent with the idea that smart beta ETFs offer a distinct advantage over traditional cap-weighted indexes. Furthermore, according to an analysis performed for Reuters by ETF.com, and reported in Barlyn [2015], recent smart beta performance results have been disappointing.
Gestalt Adaptive Asset Allocation.pdf
Gestalt Smart Beta Factor Models Active Portfolio Management
Malkiel Smart Beta Dumb Factor Models.pdf
Glushkov Smart Beta Factor Models.pdf
Richard Smart Beta Minimum Variance Factor Models.pdf
Amenc Smart Beta Investing Factor Models.pdf
Amenc Smart Beta Factor Models.pdf
Amenc Smart Beta Factor Models JPM.pdf -
Russell factor model
Barber Russell Smart Beta Factor Models.pdf
Blitz Factor Models Investing.pdf -
Beta Rotation
Bilello Utilities Sector Indicator Beta Rotation Strategy.pdf
Bilello Lumber Gold Indicator Beta Rotation Strategy.pdf -
create strategy for betting against beta
Asness Betting Against Beta.pdf
Frazzini Betting Against Beta.pdf
https://gist.github.com/timelyportfolio/11148198
https://gist.github.com/timelyportfolio/11232439
http://blog.alphaarchitect.com/2014/06/09/betting-beta-demand-lottery/ -
Minimum Variance Strategy
even better: Minimum Variance minus High Variance Strategy ?
what about time-dependent beta? -
Tactical Asset Allocation simple script
http://blog.fosstrading.com/2009/11/tactical-asset-allocation-using-blotter.html
http://petewerner.blogspot.com/2012/04/mebane-faber-tactical-asset-allocation.html
https://github.com/petewerner/misc/blob/master/gtaa-script.R -
Tactical Asset Allocation Faber
http://unstarched.net/2013/06/18/the-fallacy-of-1n-and-static-weight-allocation/ -
Elastic Asset Allocation using package !!! IKTrading Keller (2014) Elastic Asset Allocation
https://quantstrattrader.wordpress.com/2015/01/30/comparing-flexible-and-elastic-asset-allocation/ -
Flexible Asset Allocation returns algorithm
Keller Elastic Asset Allocation.pdf -
momentum e-book
http://www.investfy.co/little-book-of-momentum/ -
the 12M-1M momentum is the 11-month return up to one month ago
12-month-1-month momentum strategy
Practically, it can be viewed as an 11-month momentum strategy executed with a one-month delay.
A third factor in the form of 12-month return momentum (Jegadeesh, 1990;
Jegadeesh and Titman, 1993) was incorporated by Fama and French (1996) and -
Ross Bennett: Momentum with R
http://rbresearch.wordpress.com/2012/08/23/momentum-with-r-part-1
http://rbresearch.wordpress.com/2012/10/20/momentum-in-r-part-2
http://rbresearch.wordpress.com/2012/11/18/momentum-in-r-part-3
http://rbresearch.wordpress.com/2013/02/19/momentum-in-r-part-4-with-quantstrat/ -
AQR "Dispelling Myths of Momentum": replicate paper with R and rCharts:
http://timelyportfolio.github.io/rCharts_factor_analytics/aqr_fact_fiction_momentum.html
https://github.com/timelyportfolio/rCharts_factor_analytics/
http://timelyportfolio.blogspot.com/2014/06/dispelling-myths-of-momentum-aqr.html -
is momentum waning?
http://www.philosophicaleconomics.com/2015/12/momentum/ -
Momentum crashes
Barroso Momentum Volatility Crash Forecasting.pdf
https://quantstrattrader.wordpress.com/2015/09/16/hypothesis-driven-development-part-iv-testing-the-barrososanta-clara-rule/ -
invest in portfolio with highest momentum or one year trailing Sharpe Ratio
Gogerty Portfolio Optimization Momentum Asset Allocation.pdf
!!! Keller Momentum Markowitz Asset Allocation.pdf
https://quantstrattrader.wordpress.com/2015/06/05/momentum-markowitz-and-solving-rank-deficient-covariance-matrices-the-constrained-critical-line-algorithm/
https://systematicinvestor.wordpress.com/2013/03/22/maximum-sharpe-portfolio/ https://github.com/drquant/R_Finance/blob/master/Momentum_and_Markowitz/kellerCLAfun.R
http://systematicinvestor.github.io/Review-Momentum-Markowitz/ -
similar to above, but applies shrinkage
!!! Keller Momentum Markowitz Shrinkage Asset Allocation.pdf -
use ETFs from:
Antonacci Optimal Momentum.pdf -
PerformanceAnalytics portfolio rebalancing
PerformanceAnalytics Return.portfolio.pdf
http://tradeblotter.wordpress.com/2014/09/25/aggregate-portfolio-contributions-through-time/
Omega ratio
Adjusted Sharpe ratio
add various table.*
table.SpecificRisk()
table.Distributions()
table.DrawdownsRatio()
table.DownsideRiskRatio() -
examples:
http://seekingalpha.com/article/3222126-the-world-country-top-4-etf-strategy-a-way-to-fight-rising-rates-and-a-stalling-u-s-stock-market
http://seekingalpha.com/article/3536476-lower-risk-versions-of-a-dual-momentum-fixed-income-strategy
http://seekingalpha.com/article/3578136-a-paradigm-shift-for-tactical-strategies-trading-mutual-funds-on-a-monthly-basis
-
luck versus skill by Michael Mauboussin, head of financial strategies at Credit Suisse and adjunct professor at Columbia Business School
The null hypothesis should be that outperformance is due to luck only.
https://www.dimensional.com/famafrench/essays/luck-versus-skill-in-mutual-fund-performance.aspx
https://hbr.org/2011/02/untangling-skill-and-luck -
paradox of skill by Michael Mauboussin
Mauboussin Paradox of Skill.pdf
The standard deviation of skill goes down over time.
As skill improves in the population, luck becomes more important in producing outperformance.
The paradox of skill states that in fields where there is no offsetting interaction (for example, pitcher versus hitter) and no luck, we should see absolute results improve and relative results cluster.
The variance of quality in consumer goods has narrowed over time, another finding that’s consistent with the paradox of skill. -
use: data(edhec) from library(PerformanceAnalytics)
chart.CumReturns(edhec) -
demonstrate that: terminal lognormal asset price distribution is very skewed
most paths are below expected value
if we start with a portfolio of 500 stocks, most will underperform
therefore most randomly selected portfolios will underperform the index
therefore most PMs who randomly select portfolios will underperform the index -
Benchmark portfolio management skill using random portfolios http://gestaltu.com/2015/10/apples-and-oranges-a-random-portfolio-case-study.html
Novomestky package: rportfolios
Stein Random Portfolios Fund Analysis.pdf
Resampling Methods Bootstrap Cross Validation Random Portfolios
http://www.capitalspectator.com/using-random-portfolios-to-test-asset-allocation-strategies/
https://quantstrattrader.wordpress.com/2015/09/10/monte-carlo-in-asset-allocation-tests/
https://gist.github.com/jpicerno1/fbc2e589023be56dde42
http://www.capitalspectator.com/skewed-by-randomness-testing-arbitrary-rebalancing-dates/
https://gist.github.com/jpicerno1/af88861bcbbb80687cfb
http://www.burns-stat.com/documents/tutorials/the-statistical-bootstrap-and-other-resampling-methods-2/ -
random portfolios can outperform market if they're equally weighted - because they're overweight value and small-cap stocks
Research Affiliates indexes (known as RAFIs) rank stocks based on book value as well as trailing five-year average cash flow, sales and dividends. -
not a single of the 1000 random portfolios of size 50 delivers a annualized return below the S&P 500 index
https://predictivealpha.wordpress.com/2015/12/24/towards-a-better-equity-benchmark-random-portfolios/
http://robotwealth.com/benchmarking-backtest-results-against-random-strategies/ -
random portfolios indicate additional factors not included in FF4
Arnott Random Portfolios Factor Models.pdf
Amenc Random Portfolios Factor Models.pdf -
Benchmarking portfolio management skill using bootstrapping random investment choices
for a given model, simulate random investment choices who make random decisions to buy or sell at the same time as the model does
calculate the distribution of random manager performance, and calculate the measure of model outperformance compared to the random managers
Harvey Bootstrap Factor Models.pdf -
benchmarking investor timing skill Merton
Roy D. Henriksson and Robert C. Merton. -
benchmark investor skill using portfolio convexity skewness
create scatterplot of returns of managed strategy versus benchmark strategy
should illustrate convexity profile -
Performance attribution
PerformanceAnalytics PA-Bacon.pdf
!!! pa package Kane Performance Attribution.pdf
pa package Lu Performance Attribution.pdf
performance attribution: asset allocation, asset picking, and timing
Stubbs Portfolio Performance Attribution Factor Models.pdf
Guasoni Alpha Actively Managed Funds.pdf
Ferson Portfolio Performance Attribution Bootstrap Factor Models.pdf
-
consider a seasonal process that is the sum of two AR processes for example in the AM a process with Hurst=0.4, and in the PM a process with Hurst=0.6 what is the Hurst for such a process?
-
consider a process for which the Hurst depends on the level of volatility for example the Hurst=0.6 for high volatility, and the Hurst=0.4 for low volatility what is the Hurst for such a process?
-
simulate Ornstein-Uhlenbeck process AR(1) model and trade it
forecast returns and demonstrate that forecasting is easier with stronger mean-reversion
http://robotwealth.com/exploring-mean-reversion-and-cointegration-with-zorro-and-r-part-1/
http://robotwealth.com/exploring-mean-reversion-and-cointegration-part-2/ -
perform Engle-Granger Cointegration test
find cointegrated pairs and demonstrate that cointegration fails out-of-sample
package egcm: Engle-Granger Cointegration Models
package PairTrading.pdf
http://denizstij.blogspot.com/2013/11/stationary-tests-of-time-series-within-r.html -
Granger Causality
http://davegiles.blogspot.com/2011/04/testing-for-granger-causality.html -
using historical data calculate average returns after reaching peak
calculate distribution of mean-reversion times
fit to OU decay model -
cointegration and VAR models
C:\Research\R\Tutorials\Zivot\Econ 584\cointegration.pdf
ADF test for cointegration
Phillips-Ouliaris test for cointegration
package urca
cointegrationPowerPoint.pdf
cointegrationslides.pdf
cointegrationslides2.pdf -
cointegration package irlba
Lewis RFinance 2012 Cointegration SVD.pdf
C:\Research\R\R-Finance 2015\BryanLewis.html
apply Doornik’s method using the SVD to solve the cointegration problem -
pairs trading
Krauss Statistical Arbitrage Pairs Trading Review.pdf
Krauss Copula Pairs Trading Cointegration.pdf
Steffen Hurst Cointegration Pairs Trading.pdf
Leung Pairs Trading Stop-loss Rule.pdf
Clegg Cointegration Pairs Trading.pdf
Miao Statistical Arbitrage Cointegration.pdf
Grabovsky Statistical Arbitrage Pairs Trading.pdf
Kakushadze Statistical Arbitrage.pdf
Kakushadze Pairs Trading Factor Models.pdf
Rudy Pairs Trading Stocks ETFs.pdf
QUSMA ETF Daily Mean Reversion.pdf
C:\Research\Stat Arb Peter\Pairs Trading Cointegration\
Gatev, Goetzmann, and Rouwenhorst 2006.pdf
Tourin Pairs Trading HJB Stochastic Control.pdf -
demonstrate that the pairs trading returns have negative skewness
because upside returns are capped by trade exit rule
while there can be large negative returns if the trade moves against you -
high frequency data cointegration
Krauss High Frequency Cointegration -
index arbitrage
Avellaneda Statistical Arbitrage 2008.pdf
C:\Research\Academic\Avellaneda Quantitative Investment Strategies -
combine mean reversion and momentum strategies
!!! Velissaris Statistical Arbitrage Momentum Strategies.pdf -
residual momentum strategy Blitz Short-Term Residual Reversal.pdf
https://factorinvestingtutorial.wordpress.com/9-residual-momentum-david-blitz/ -
yield curve butterfly strategy
futures butterfly strategy