Foundations of Alpha: The Theory Behind Algorithmic Trading

This lecture presents the topics that will be covered and explains some preliminary concepts necessary for understanding the section.

The text explains two approaches to developing alpha models: theory-driven (hypothesis testing) and data-driven (pattern recognition) based on the quant's scientific background.

Common quantitative models test hypotheses on market behavior using price or fundamental data. These models fall into six main strategies, also used by discretionary traders.

This text describes three quantitative trading strategies based on price and volume data: trend following (momentum), mean reversion (counter-trend), and technical sentiment.

This text classifies quantitative value/yield, growth, and quality strategies based on fundamental metrics (ratios) to buy undervalued instruments or those with high growth potential.

Data-driven alpha models use complex algorithms to find trading patterns in historical data, but require careful data selection and model updates to avoid unreliable results.

This lecture outlines six key elements for understanding quantitative trading strategies: forecast target, time horizon, bet structure, investment universe, model definition, and run frequency.

This lecture discusses combining multiple quantitative trading strategies (alpha models) using linear, non-linear, and machine learning models.

It is the summary of the lectures of this section.

In this lecture we discuss investment universes, why and how to restrict assets/securities traded, and what considerations we need to keep in mind when constructing investment universes.

In this lecture we take a look at the Sharpe Ratio, including both ex ante and ex post definitions, and what considerations are needed regarding the choice of risk-free rate.

Carrying on with the Sharpe Ratio, in this video we demonstrate some better ways to visualize the Sharpe Ratio, and discuss some limitations of the ratio.

We begin the lecture on the Fundamental Law of Active Management with an introduction to the Information Ratio, and why it's important to understand when it comes to adding value by maximizing return and reducing risk.

In this video we conclude the lecture on the Fundamental Law of Active Management, and show how we can use it as a framework to improve our risk-adjusted returns by focusing on investment skill and breadth.

This lecture presents the topics that will be covered and explains some preliminary concepts necessary for understanding the section.

Quantitative risk models limit exposure by constraint (hard limits or penalties), measure risk (longitudinal volatility or cross-sectional similarity), and can be applied to individual positions, sectors, or overall portfolio leverage.

The lecture explains risk models that aim to eliminate unintentional exposures to systematic risks, like market direction, by focusing on relative performance within a sector.

The lecture contrasts two risk modeling approaches: theory-driven models use established factors to measure systematic risks, while data-driven models use statistics to find (and potentially overfit) future risks.

The lecture discusses the pros and cons of theory-driven (explainable but inflexible) and empirical (adaptable but potentially inaccurate) risk models, suggesting that combining them or using pre-made models can be good options.

It is the summary of the lectures of this section.

Abstract:

The central problem for gamblers is to find positive expectation bets. But the gambler also needs to know how to manage his money, i.e., how much to bet. In the stock market (more inclusively, the securities markets) the problem is similar but more complex. The gambler, who is now an “investor”, looks for “excess risk adjusted return”.

In both these settings, we explore the use of the Kelly criterion, which is to maximize the expected value of the logarithm of wealth (“maximize expected logarithmic utility”). The criterion is known to economists and financial theorists by names such as the “geometric mean maximizing portfolio strategy”, maximizing logarithmic utility, the growth-optimal strategy, the capital growth criterion, etc. The author initiated the practical application of the Kelly criterion by using it for card counting in blackjack.

We will present some useful formulas and methods to answer various natural questions about it that arise in blackjack and other gambling games. Then we illustrate its recent use in a successful casino sports betting system. Finally, we discuss its application to the securities markets where it has helped the author to make a thirty year total of 80 billion dollars worth of “bets”.

This lecture presents the topics that will be covered and explains some preliminary concepts necessary for understanding the section.

Transaction costs include all expenses throughout the trade lifecycle, from idea generation to execution, and pre-trade analysis helps estimate these costs to improve investment return.

A pre-trade analysis, part of the Transaction Cost Model, estimates transaction costs using data like liquidity, risk (volatility and beta), and historical costs to optimize trade execution.

Post-trade analysis helps assess trade execution quality and cost by analyzing data like slippage and execution time, informing future trade decisions.

The Relative Performance Measure (RPM) assesses trade performance by comparing it to the overall market activity for that specific asset, instead of a single price point.

Transaction costs are not uniform and depend on various factors like broker, execution efficiency, and trading volume. They can be categorized into investment-related (before order placement) and trading-related (after order placement) costs.

Trading costs include explicit fees like commissions and exchange fees, but also hidden implicit costs that can significantly impact returns.

Commissions and fees, paid to brokers, exchanges, and regulators, are a known transaction cost included in trading models.

The lecture explains that the slippage is the difference between the intended price of a trade and the actual execution price due to market movements and order processing time.

The lecture shows that the bid-ask spread is the difference between the buy and sell price, representing a cost for immediate trade execution.

The lecture explains that Market impact is a cost caused by large orders moving the market price when buying or selling an asset. Also shows how to deal with it.

Transaction costs aren't fully preventable but strategic planning (like order sizing) can help manage them.

Upward price trends increase buying costs and decrease selling costs, while downward trends do the opposite.

Timing risk arises from asset price volatility and liquidity, and can be mitigated with a good understanding of market conditions.

Opportunity cost, unlike fees, is the potential profit lost due to incomplete order execution from price movements or lack of liquidity.

Transaction cost models aim to predict how much a trade will cost. The lecture explains some transaction cost models, with quadratic models being the most accurate but also the most complex.

The lecture provides some final considerations regarding transaction costs in general, as well as elements related to transaction cost models.

It is the summary of the lectures of this section.

This lecture presents the topics that will be covered and explains some preliminary concepts necessary for understanding the section.

Modern Portfolio Theory helps construct portfolios by considering overall risk and return, not just individual assets.

The efficient frontier shows the optimal investment portfolios that balance risk and return.

There are many ways to construct portfolios, with quants focusing on systematic and scientific approaches.

Portfolio construction models for quants act like active managers by overweighting and underweighting assets relative to a benchmark.

Rule-based portfolio construction models rely on pre-defined rules, like equal weighting or using alpha model strength, to determine investment positions.

Portfolio optimizers are algorithms that use mean-variance analysis to find portfolios with the best risk-return ratio.

The main inputs to portfolio optimization are expected return, expected volatility, and correlation matrix, but they can be difficult to estimate accurately.

Portfolio optimization uses metrics like Sharpe ratio to compare risk and return, but the results can be complex due to interactions between different models within the system.

Quantitative portfolio models recommend a targeted portfolio with specific weights for each instrument, requiring rebalancing to maintain the desired holdings.

There are two approaches to quantitative portfolio construction: intrinsic (independent trades) and relative (using optimizers based on correlations).

It is the summary of the lectures of this section.

This lecture presents the topics that will be covered and explains some preliminary concepts necessary for understanding the section.

An order book is an electronic list of buy and sell orders for a financial instrument, used by traders to see buy/sell pressure, depth of market, and potential price direction.

Direct market access (DMA) is how traders electronically access exchanges to buy and sell instruments.

Order execution algorithms aim to minimize costs and achieve the desired portfolio holdings when trading to reach a target portfolio.

Execution algorithms consider order types (aggressive vs passive), size (large vs small), and routing to minimize costs and achieve target portfolio holdings.

Trading infrastructure involves choosing a connection method (DMA, colocation) and communication protocol (FIX) to minimize latency.

It is the summary of the lectures of this section.

This lecture will consider the different types of orders.

Market orders will be examined.

Limit orders will be examined.

What are trading algorithms as well as the difference between algorithms with a predetermined schedule vs algorithms that adjust dynamically.

We will examine common algorithmic features such as duration and execution styles.

The main focus will be on Time Weighted Average Price.

The main focus will be on Volume Weighted Average Price.

The main focus will be on Percent of Volume.

The main focus will be on Minimal-Impact Algorithms.

Understand the difference between fundamental & price data. We also investigate alternative data and the information lag.

Examine the primary sources of data, sentiment analysis as well as the security master.

Understand how to handle missing values, erroneous values as well as the spike filter.

Examine the difference between relationship and flat-file databases. Using the arctic database in practice.