arXiv subjects

Tim Gebbie

Publications and source records attributed to Tim Gebbie.

54 records · Page 3Linked to original sources

arXiv2019

Learning the dynamics of technical trading strategies

We use an adversarial expert based online learning algorithm to learn the optimal parameters required to maximise wealth trading zero-cost portfolio strategies. The learning algorithm is used to determine the relative population dynamics of technical trading strategies that can survive historical back-testing as well as form an overall aggregated portfolio trading strategy from the set of underlying trading strategies implemented on daily and intraday Johannesburg Stock Exchange data. The resulting population time-series are investigated using unsupervised learning for dimensionality reduction and visualisation. A key contribution is that the overall aggregated trading strategies are tested for statistical arbitrage using a novel hypothesis test proposed by Jarrow et al. (2012) on both daily sampled and intraday time-scales. The (low frequency) daily sampled strategies fail the arbitrage tests after costs, while the (high frequency) intraday sampled strategies are not falsified as statistical arbitrages after costs. The estimates of trading strategy success, cost of trading and slippage are considered along with an online benchmark portfolio algorithm for performance comparison. In addition, the algorithms generalisation error is analysed by recovering a probability of back-test overfitting estimate using a nonparametric procedure introduced by Bailey et al. (2016). The work aims to explore and better understand the interplay between different technical trading strategies from a data-informed perspective.

q-fin.CP↗

arXiv2019

Systematic Asset Allocation using Flexible Views for South African Markets

We implement a systematic asset allocation model using the Historical Simulation with Flexible Probabilities (HS-FP) framework developed by Meucci. The HS-FP framework is a flexible non-parametric estimation approach that considers future asset class behavior to be conditional on time and market environments, and derives a forward looking distribution that is consistent with this view while remaining close as possible to the prior distribution. The framework derives the forward looking distribution by applying unequal time and state conditioned probabilities to historical observations of asset class returns. This is achieved using relative entropy to find estimates with the least distortion to the prior distribution. Here, we use the HS-FP framework on South African financial market data for asset allocation purposes; by estimating expected returns, correlations and volatilities that are better represented through the measured market cycle. We demonstrated a range of state variables that can be useful towards understanding market environments. Concretely, we compare the out-of-sample performance for a specific configuration of the HS-FP model relative to classic Mean Variance Optimization(MVO) and Equally Weighted (EW) benchmark models. The framework displays low probability of backtest overfitting and the out-of-sample net returns and Sharpe ratio point estimates of the HS-FP model outperforms the benchmark models. However, the results are inconsistent when training windows are varied, the Sharpe ratio is seen to be inflated, and the method does not demonstrate statistically significant out-performance on a gross and net basis.

q-fin.PM↗

arXiv2019

Fast Super-Paramagnetic Clustering

We map stock market interactions to spin models to recover their hierarchical structure using a simulated annealing based Super-Paramagnetic Clustering (SPC) algorithm. This is directly compared to a modified implementation of a maximum likelihood approach we call Fast Super-Paramagnetic Clustering (f-SPC). The methods are first applied standard toy test-case problems, and then to a data-set of 447 stocks traded on the New York Stock Exchange (NYSE) over 1249 days. The signal to noise ratio of stock market correlation matrices is briefly considered. Our result recover approximately clusters representative of standard economic sectors and mixed ones whose dynamics shine light on the adaptive nature of financial markets and raise concerns relating to the effectiveness of industry based static financial market classification in the world of real-time data analytics. A key result is that we show that f-SPC maximum likelihood solutions converge to ones found within the Super-Paramagnetic Phase where the entropy is maximum, and those solutions are qualitatively better for high dimensionality data-sets.

q-fin.CP↗

arXiv2017

Can Agent-Based Models Probe Market Microstructure?

We extend prior evidence that naively using intraday agent-based models that involve realistic order-matching processes for modeling continuous-time double auction markets seems to fail to be able to provide a robust link between data and many model parameters, even when these models are able to reproduce a number of well-known stylized facts of return time series. We demonstrate that while the parameters of intraday agent-based models rooted in market microstructure can be meaningfully calibrated, those exclusively related to agent behaviors and incentives remain problematic. This could simply be a failure of the calibration techniques used but we argue that the observed parameter degeneracies are most likely a consequence of the realistic matching processes employed in these models. This suggests that alternative approaches to linking data, phenomenology and market structure may be necessary and that it is conceivable that one could construct a useful model that does not directly depend on the nuances of agent behaviors, even when it is known that the real agents engage in complex behaviors.

q-fin.CP↗

arXiv2017

The Problem of Calibrating an Agent-Based Model of High-Frequency Trading

Agent-based models, particularly those applied to financial markets, demonstrate the ability to produce realistic, simulated system dynamics, comparable to those observed in empirical investigations. Despite this, they remain fairly difficult to calibrate due to their tendency to be computationally expensive, even with recent advances in technology. For this reason, financial agent-based models are frequently validated by demonstrating an ability to reproduce well-known log return time series and central limit order book stylized facts, as opposed to being rigorously calibrated to transaction data. We thus apply an established financial agent-based model calibration framework to a simple model of high- and low-frequency trader interaction and demonstrate possible inadequacies of a stylized fact-centric approach to model validation. We further argue for the centrality of calibration to the validation of financial agent-based models and possible pitfalls of current approaches to financial agent-based modeling.

q-fin.CP↗

arXiv2017

Detecting intraday financial market states using temporal clustering

We propose the application of a high-speed maximum likelihood clustering algorithm to detect temporal financial market states, using correlation matrices estimated from intraday market microstructure features. We first determine the ex-ante intraday temporal cluster configurations to identify market states, and then study the identified temporal state features to extract state signature vectors which enable online state detection. The state signature vectors serve as low-dimensional state descriptors which can be used in learning algorithms for optimal planning in the high-frequency trading domain. We present a feasible scheme for real-time intraday state detection from streaming market data feeds. This study identifies an interesting hierarchy of system behaviour which motivates the need for time-scale-specific state space reduction for participating agents.

q-fin.TR↗

arXiv2016

Deviations in expected price impact for small transaction volumes under fee restructuring

We report on the occurrence of an anomaly in the price impacts of small transaction volumes following a change in the fee structure of an electronic market. We first review evidence for the existence of a master curve for price impact on the Johannesburg Stock Exchange (JSE). On attempting to re-estimate a master curve after fee reductions, it is found that the price impact corresponding to smaller volume trades is greater than expected relative to prior estimates for a range of listed stocks. We show that a master curve for price impact can be found following rescaling by an appropriate liquidity proxy, providing a means for practitioners to approximate price impact curves without onerous processing of tick data.

q-fin.TR↗

arXiv2016

Learning zero-cost portfolio selection with pattern matching

We consider and extend the adversarial agent-based learning approach of Gy{ö}rfi {\it et al} to the situation of zero-cost portfolio selection implemented with a quadratic approximation derived from the mutual fund separation theorems. The algorithm is applied to daily sampled sequential Open-High-Low-Close data and sequential intraday 5-minute bar-data from the Johannesburg Stock Exchange (JSE). Statistical tests of the algorithms are considered. The algorithms are directly compared to standard NYSE test cases from prior literature. The learning algorithm is used to select parameters for agents (or experts) generated by pattern matching past dynamics using a simple nearest-neighbour search algorithm. It is shown that there is a speed advantage associated with using an analytic solution of the mutual fund separation theorems. It is argued that the expected loss in performance does not undermine the potential application to intraday quantitative trading and that when transactions costs and slippage are considered the strategies can still remain profitable when unleveraged. The paper demonstrates that patterns in financial time-series on the JSE can be systematically exploited in collective but that this does not imply predictability of the individual asset time-series themselves.

q-fin.CP↗

arXiv2015

High-speed detection of emergent market clustering via an unsupervised parallel genetic algorithm

We implement a master-slave parallel genetic algorithm (PGA) with a bespoke log-likelihood fitness function to identify emergent clusters within price evolutions. We use graphics processing units (GPUs) to implement a PGA and visualise the results using disjoint minimal spanning trees (MSTs). We demonstrate that our GPU PGA, implemented on a commercially available general purpose GPU, is able to recover stock clusters in sub-second speed, based on a subset of stocks in the South African market. This represents a pragmatic choice for low-cost, scalable parallel computing and is significantly faster than a prototype serial implementation in an optimised C-based fourth-generation programming language, although the results are not directly comparable due to compiler differences. Combined with fast online intraday correlation matrix estimation from high frequency data for cluster identification, the proposed implementation offers cost-effective, near-real-time risk assessment for financial practitioners.

q-fin.CP↗

arXiv2014

Hierarchical causality in financial economics

Hierarchical analysis is considered and a multilevel model is presented in order to explore causality, chance and complexity in financial economics. A coupled system of models is used to describe multilevel interactions, consistent with market data: the lowest level is occupied by agents generating the prices of individual traded assets; the next level entails aggregation of stocks into markets; the third level combines shared risk factors with information variables and bottom-up, agent-generated structure, consistent with conditions for no-arbitrage pricing theory; the fourth level describes market factors which originate in the greater economy and the highest levels are described by regulated market structure and the customs and ethics which define the nature of acceptable transactions. A mechanism for emergence or innovation is considered and causal sources are discussed in terms of five causation classes.

q-fin.GN↗

arXiv2013

Factorising equity returns in an emerging market through exogenous shocks and capital flows

A technique from stochastic portfolio theory [Fernholz, 1998] is applied to analyse equity returns of Small, Mid and Large cap portfolios in an emerging market through periods of growth and regional crises, up to the onset of the global financial crisis. In particular, we factorize portfolios in the South African market in terms of distribution of capital, change of stock ranks in portfolios, and the effect due to dividends for the period Nov 1994 to May 2007. We discuss the results in the context of broader economic thinking to consider capital flows as risk factors, turning around more established approaches which use macroeconomic and socio-economic conditions to explain Foreign Direct Investment (into the economy) and Net Portfolio Investment (into equity and bond markets).

q-fin.GN↗

arXiv2008

How many independent bets are there?

The benefits of portfolio diversification is a central tenet implicit to modern financial theory and practice. Linked to diversification is the notion of breadth. Breadth is correctly thought of as the number of in- dependent bets available to an investor. Conventionally applications us- ing breadth frequently assume only the number of separate bets. There may be a large discrepancy between these two interpretations. We uti- lize a simple singular-value decomposition (SVD) and the Keiser-Gutman stopping criterion to select the integer-valued effective dimensionality of the correlation matrix of returns. In an emerging market such as South African we document an estimated breadth that is considerably lower than anticipated. This lack of diversification may be because of market concentration, exposure to the global commodity cycle and local currency volatility. We discuss some practical extensions to a more statistically correct interpretation of market breadth, and its theoretical implications for both global and domestic investors.

physics.soc-ph↗

arXiv2007

Serial Correlation, Periodicity and Scaling of Eigenmodes in an Emerging Market

We investigate serial correlation, periodic, aperiodic and scaling behaviour of eigenmodes, i.e. daily price fluctuation time-series derived from eigenvectors, of correlation matrices of shares listed on the Johannesburg Stock Exchange (JSE) from January 1993 to December 2002. Periodic, or calendar, components are detected by spectral analysis. We find that calendar effects are limited to eigenmodes which correspond to eigenvalues outside the Wishart range. Using a variance ratio test, we uncover serial correlation in the first eigenmodes and find slight negative serial correlation for eigenmodes within the Wishart range. Our spectral analysis and variance ratio investigations suggest that interpolating missing data or illiquid trading days with zero-order hold introduces high frequency noise and spurious serial correlation. Aperiodic and scaling behaviour of the eigenmodes are investigated by using rescaled-range (R/S) methods and detrended fluctuation analysis (DFA). We find that DFA and classic and modified R/S exponents suggest the presence of long-term memory effects in the first five eigenmodes.

cond-mat.stat-mech↗

arXiv2006

An analysis of Cross-correlations in South African Market data

We apply random matrix theory to compare correlation matrix estimators C obtained from emerging market data. The correlation matrices are constructed from 10 years of daily data for stocks listed on the Johannesburg Stock Exchange (JSE) from January 1993 to December 2002. We test the spectral properties of C against random matrix predictions and find some agreement between the distributions of eigenvalues, nearest neighbour spacings, distributions of eigenvector components and the inverse participation ratios for eigenvectors. We show that interpolating both missing data and illiquid trading days with a zero-order hold increases agreement with RMT predictions. For the more realistic estimation of correlations in an emerging market, we suggest a pairwise measured-data correlation matrix. For the data set used, this approach suggests greater temporal stability for the leading eigenvectors. An interpretation of eigenvectors in terms of trading strategies is given in lieu of classification by economic sectors.

cond-mat.stat-mech↗

arXiv1999

1+3 Covariant Cosmic Microwave Background anisotropies II: The almost - Friedmann Lemaitre model

This is the second of a series of papers extending the 1+3 covariant and gauge invariant treatment of kinetic theory to an examination of Cosmic Microwave Background temperature anisotropies arising from inhomogeneities in the early universe. The first paper dealt with algebraic issues. Here we derive the mode form of the integrated Boltzmann equations, first, giving a covariant version of the standard derivation using the mode recursion relations, second, demonstrating the link to the multipole divergence equations and finally various analytic ways of solving the resulting equations are discussed. A general integral form of solution is obtained for the equations with Thomson scattering. The covariant Friedmann-Lemaitre multipole form of the transport equations are found using the covariant and gauge-invariant generalization of the Peebles and Yu expansion in Thompson scattering time. The dispersion relations and damping scale are then obtained from the covariant approach. The equations are integrated to give the covariant and gauge-invariant equivalent of the canonical scalar sourced anisotropies. We carry out a simple treatment of the matter dominated free-streaming projection, slow decoupling, and tight-coupling cases, with the aim both giving a unified transparent derivation of this range of results and clarifying the connection between the more usual approaches (for example that of Hu and Sugiyama) and the treatment for scalar perturbations (for example the treatment of Challinor and Lasenby).

astro-ph↗

arXiv1999

1+3 Covariant Cosmic Microwave Background anisotropies I: Algebraic relations for mode and multipole representations

This is the first of a series of papers extending a covariant and gauge invariant (CGI) treatment of kinetic theory in curved space-times to a treatment of Cosmic Background Radiation (CBR) temperature anisotropies arising from inhomogeneities in the early universe. This paper deals with algebraic issues, both generically and in the context of models linearised about RW geometries. The approach represents radiation anisotropies by PSTF tensors. The Angular correlation functions for the mode coefficients are found in terms of these quantities, following the Wilson-Silk approach, but derived and dealt with in CGI form. The covariant multipole and mode-expanded angular correlation functions are related to the usual treatments in the literature. The CGI mode expansion is related to the coordinate approach by linking the Legendre functions to the PSTF representation, using a covariant addition theorem for the tensors to generate the Legendre Polynomial recursion relation. This paper lays the foundation for further papers in the series, which use this formalism in a CGI approach to develop solutions of the Boltzmann and Liouville equations for the CBR before and after decoupling, thus providing a unified CGI derivation of the variety of approaches to CBR anisotropies in the current literature.

astro-ph↗

arXiv1999

The Limits on Cosmological Anisotropies and Inhomogeneities from COBE Data

Assuming that the cosmological principle holds, Maartens, Ellis and Stoeger (MES) recently constructed a detailed scheme linking anisotropies in the cosmic background radiation (CMB) with anisotropies and inhomogeneities in the large scale structure of the universe and showed how to place limits on those anisotropies and inhomogeneities simply by using CMB quadrupole and octupole limits. First we indicate and discuss the connection between the covariant multipole moments of the temperature anisotropy used in the MES scheme and the quadrupole and octupole results from COBE. Then we introduce those results into the MES limit equations to obtain definite quantitative limits on the complete set of cosmological measures of anisotropy and inhomogeneity. We find that all the anisotropy measures are less than 10^{-4} in the case of those not affected by the expansion rate H, and less than 10^{-6} Mpc^{-1} in the case of those which are. These results quantitatively demonstrate that the observable universe is indeed close to Friedmann-Lemaitre-Robertson-Walker (FLRW) on the largest scales, and can be adequately modelled by an almost-FLRW model -- that is, the anisotropies and inhomogeneities characterizing the observable universe on the largest scales are not too large to be considered perturbations to FLRW.

astro-ph↗

arXiv1999

Cosmic microwave background anisotropies: Nonlinear dynamics

We develop a new approach to local nonlinear effects in cosmic microwave background anisotropies, and discuss the qualitative features of these effects. New couplings of the baryonic velocity to radiation multipoles are found, arising from nonlinear Thomson scattering effects. We also find a new nonlinear shear effect on small angular scales. The full set of evolution and constraint equations is derived, including the nonlinear generalizations of the radiation multipole hierarchy, and of the dynamics of multi-fluids. These equations govern radiation anisotropies in any inhomogeneous spacetime, but their main application is to second-order effects in a universe that is close to the Friedmann models. Qualitative analysis is given here, and quantitative calculations are taken up in further papers.

astro-ph↗