Quantitative Finance

In this chapter we review some recent results on the dynamics of price formation in financial markets and its relations with the efficient market hypothesis. Specifically, we present the limit order book mechanism for markets and we introduce the concepts of market impact and order flow, presenting their recently discovered empirical properties and discussing some possible interpretation in terms of agent's strategies. Our analysis confirms that quantitative analysis of data is crucial to validate qualitative hypothesis on investors' behavior in the regulated environment of order placement and to connect these micro-structural behaviors to the properties of the collective dynamics of the system as a whole, such for instance market efficiency. Finally we discuss the relation between some of the described properties and the theory of reflexivity proposing that in the process of price formation positive and negative feedback loops between the cognitive and manipulative function of agents are present.

We make an extensive empirical study of the market impact of large orders (metaorders) executed in the U.S. equity market between 2007 and 2009. We show that the square root market impact formula, which is widely used in the industry and supported by previous published research, provides a good fit only across about two orders of magnitude in order size. A logarithmic functional form fits the data better, providing a good fit across almost five orders of magnitude. We introduce the concept of an "impact surface" to model the impact as a function of both the duration and the participation rate of the metaorder, finding again a logarithmic dependence. We show that during the execution the price trajectory deviates from the market impact, a clear indication of non-VWAP executions. Surprisingly, we find that sometimes the price starts reverting well before the end of the execution. Finally we show that, although on average the impact relaxes to approximately 2/3 of the peak impact, the precise asymptotic value of the price depends on the participation rate and on the duration of the metaorder. We present evidence that this might be due to a herding phenomenon among metaorders.

Attempts to accurately measure the monetary velocity or related properties of bitcoin used in transactions have often attempted to either directly apply definitions from traditional macroeconomic theory or to use specialized metrics relative to the properties of the Blockchain like bitcoin days destroyed. In this paper, it is demonstrated that beyond being a useful metric, bitcoin days destroyed has mathematical properties that allow you to calculate the average dormancy (time since last use in a transaction) of the bitcoins used in transactions over a given time period. In addition, bitcoin days destroyed is shown to have another unexpected significance as the average size of the pool of traded bitcoins by virtue of the expression Little's Law, though only under limited conditions.

Financial exchange operators cater to the needs of their users while simultaneously ensuring compliance with the financial regulations. In this work, we focus on the operators' commitment for fair treatment of all competing participants. We first discuss unbounded temporal fairness and then investigate its implementation and infrastructure requirements for exchanges. We find that these requirements can be fully met only under ideal conditions and argue that unbounded fairness in FIFO markets is unrealistic. To further support this claim, we analyse several real-world incidents and show that subtle implementation inefficiencies and technical optimizations suffice to give unfair advantages to a minority of the participants. We finally introduce, {\epsilon}-fairness, a bounded definition of temporal fairness and discuss how it can be combined with non-continuous market designs to provide equal participant treatment with minimum divergence from the existing market operation.

Prices in financial markets exhibit extreme jumps far more often than can be accounted for by external news. Further, magnitudes of price changes are correlated over long times. These so called stylized facts are quantified by scaling laws similar to, for example, turbulent fluids. They are believed to reflect the complex interactions of heterogenous agents which give rise to irrational herding. Therefore, the stylized facts have been argued to provide evidence against the efficient market hypothesis which states that prices rapidly reflect available information and therefore are described by a martingale. Here we show, that in very simple bidding processes efficiency is not opposed to, but causative to scaling properties observed in real markets. Thereby, we link the stylized facts not only to price efficiency, but also to the economic theory of rational bubbles. We then demonstrate effects predicted from our normative model in the dynamics of groups of real human subjects playing a modified minority game. An extended version of the latter can be played online at this http URL.

The success of a cross-sectional systematic strategy depends critically on accurately ranking assets prior to portfolio construction. Contemporary techniques perform this ranking step either with simple heuristics or by sorting outputs from standard regression or classification models, which have been demonstrated to be sub-optimal for ranking in other domains (e.g. information retrieval). To address this deficiency, we propose a framework to enhance cross-sectional portfolios by incorporating learning-to-rank algorithms, which lead to improvements of ranking accuracy by learning pairwise and listwise structures across instruments. Using cross-sectional momentum as a demonstrative case study, we show that the use of modern machine learning ranking algorithms can substantially improve the trading performance of cross-sectional strategies -- providing approximately threefold boosting of Sharpe Ratios compared to traditional approaches.

Distributed ledger technologies replace central counterparties with time-consuming consensus protocols to record the transfer of ownership. This settlement latency slows down cross-market trading and exposes arbitrageurs to price risk. We theoretically derive arbitrage bounds induced by settlement latency. Using Bitcoin orderbook and network data, we estimate average arbitrage bounds of 121 basis points, explaining 91% of the cross-market price differences, and demonstrate that asset flows chase arbitrage opportunities. Controlling for inventory holdings as a measure of trust in exchanges does not affect our main results. Blockchain-based settlement without trusted intermediation thus introduces a non-trivial friction that impedes arbitrage activity.

We propose a method for detection and prediction of native and synthetic iceberg orders on Chicago Mercantile Exchange. Native (managed by the exchange) icebergs are detected using discrepancies between the resting volume of an order and the actual trade size as indicated by trade summary messages, as well as by tracking order modifications that follow trade events. Synthetic (managed by market participants) icebergs are detected by observing limit orders arriving within a short time frame after a trade. The obtained icebergs are then used to train a model based on the Kaplan--Meier estimator, accounting for orders that were cancelled after a partial execution. The model is utilized to predict the total size of newly detected icebergs. Out of sample validation is performed on the full order depth data, performance metrics and quantitative estimates of hidden volume are presented.

We propose a formula of time-series prediction by means of three states random field Ising model (RFIM). At the economic crisis due to disasters or international disputes, the stock price suddenly drops. The macroscopic phenomena should be explained from the corresponding microscopic view point because there are existing a huge number of active traders behind the crushes. Hence, here we attempt to model the artificial financial market in which each trader i can choose his/her decision among `buying', `selling' or `staying (taking a wait-and-see attitude)', each of which corresponds to a realization of the three state Ising spin, namely, S i =+1 , -1 and S i =0 , respectively. The decision making of traders is given by the Gibbs-Boltzmann distribution with the energy function. The energy function contains three distinct terms, namely, the ferromagnetic two-body interaction term (endogenous information), random field term as external information (exogenous news), and chemical potential term which controls the number of traders who are watching the market calmly at the instance. We specify the details of the model system from the past financial market data to determine the conjugate hyper-parameters and draw each parameter flow as a function of time-step. Especially we will examine to what extent one can characterize the crisis by means of a brand-new order parameter --- `turnover' --- which is defined as the number of active traders who post their decisions S i =1,−1 , instead of S i =0 .

We propose a model for price formation in financial markets based on clearing of a standard call auction with random orders, and verify its validity for prediction of the daily closing price distribution statistically. The model considers random buy and sell orders, placed following demand- and supply-side valuation distributions; an equilibrium equation then leads to a distribution for clearing price and transacted volume. Bid and ask volumes are left as free parameters, permitting possibly heavy-tailed or very skewed order flow conditions. In highly liquid auctions, the clearing price distribution converges to an asymptotically normal central limit, with mean and variance in terms of supply/demand-valuation distributions and order flow imbalance. By means of simulations, we illustrate the influence of variations in order flow and valuation distributions on price/volume, noting a distinction between high- and low-volume auction price variance. To verify the validity of the model statistically, we predict a year's worth of daily closing price distributions for 5 constituents of the Eurostoxx 50 index; Kolmogorov-Smirnov statistics and QQ-plots demonstrate with ample statistical significance that the model predicts closing price distributions accurately, and compares favourably with alternative methods of prediction.