{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#Long-Short Equity Strategies\n",
"\n",
"By Delaney Granizo-Mackenzie\n",
"\n",
"Part of the Quantopian Lecture Series:\n",
"\n",
"* [www.quantopian.com/lectures](https://www.quantopian.com/lectures)\n",
"* [https://github.com/quantopian/research_public](https://github.com/quantopian/research_public)\n",
"\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Long-short equity refers to the fact that the strategy is both long and short in the equity market. This is a rather general statement, but has over time grown to mean a specific family of strategies. These strategies rank all stocks in the market using some model. The strategy then goes long (buys) the top $n$ equities of the ranking, and goes short on (sells) the bottom $n$ while maintaining equal dollar volume between the long and short positions. This has the advantage of being statistically robust, as by ranking stocks and entering hundreds or thousands of positions, you are making many bets on your ranking model rather than just a few risky bets. You are also betting purely on the quality of your ranking scheme, as the equal dollar volume long and short positions ensure that the strategy will remain market neutral (immune to market movements)."
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": true
},
"source": [
"##Ranking Scheme\n",
"\n",
"A ranking scheme is any model that can assign each stocks a number, where higher is better or worse. Examples could be value factors, technical indicators, pricing models, or a combination of all of the above. The Ranking Universes by Factors lecture will cover ranking schemes in more detail. Ranking schemes are the secret sauce of any long-short equity strategy, so developing them is nontrivial."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"##Making a Bet on the Ranking Scheme"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Once we have determined a ranking scheme, we would like to be able to profit from it. We do this by investing an equal amount of money long into the top of the ranking, and short into the bottom. This ensures that the strategy will make money proportionally to the quality of the ranking only, and will be market neutral."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"###Long and Short Baskets\n",
"\n",
"If you are ranking $m$ equities, have $d$ dollars to invest, and your total target number of positions to hold is $2n$, then the long and short baskets are created as follows. For each equity in spots $1, \\dots, n$ in the ranking, sell $\\frac{1}{2n} * d$ dollars of that equity. For each equity in spots $m - n, \\dots, m$ in the ranking, buy $\\frac{1}{2n} * d$ dollars of that equity. \n",
"\n",
"####Friction Because of Prices\n",
"\n",
"Because equity prices will not always divide $\\frac{1}{2n} * d$ evenly, and equities must be bought in integer amounts, there will be some imprecision and the algorithm should get as close as it can to this number. Most algorithms will have access to some leverage during execution, so it is fine to buy slightly more than $\\frac{1}{2n} * d$ dollars per equity. This does, however, cause some friction at low capital amounts. For a strategy running $d = 100000$, and $n = 500$, we see that \n",
"$$\\frac{1}{2n} * d = \\frac{1}{1000} * 100000 = 100$$\n",
"This will cause big problems for expensive equities, and cause the algorithm to be overlevered. This is alleviated by trading fewer equities or increasing the capital, $d$. Luckily, long-short equity strategies tend to be very high capicity, so there is for most purposes no ceiling on the amount of money one can invest. For more information on algorithm capacities, refer to the algorithm capacity lecture when it is released."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"###Returns Come From The Ranking Spread\n",
"\n",
"The returns of a long-short equity strategy are dependent on how well the ranking spreads out the high and low returns. To see how this works, consider this hypothetical example."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import numpy as np\n",
"import pandas as pd\n",
"import matplotlib.pyplot as plt"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/html": [
"\n",
"
\n",
" \n",
" \n",
" | \n",
" Factor Value | \n",
"
\n",
" \n",
" \n",
" \n",
" | Equity 0 | \n",
" 1.894223 | \n",
"
\n",
" \n",
" | Equity 1 | \n",
" -0.211479 | \n",
"
\n",
" \n",
" | Equity 2 | \n",
" -0.391872 | \n",
"
\n",
" \n",
" | Equity 3 | \n",
" 0.027782 | \n",
"
\n",
" \n",
" | Equity 4 | \n",
" 0.524842 | \n",
"
\n",
" \n",
" | Equity 5 | \n",
" -0.573875 | \n",
"
\n",
" \n",
" | Equity 6 | \n",
" -0.452927 | \n",
"
\n",
" \n",
" | Equity 7 | \n",
" -1.785987 | \n",
"
\n",
" \n",
" | Equity 8 | \n",
" -0.384685 | \n",
"
\n",
" \n",
" | Equity 9 | \n",
" -0.716920 | \n",
"
\n",
" \n",
"
\n",
"
\n",
"
\n",
" \n",
" \n",
" | \n",
" Returns | \n",
" Factor Value | \n",
"
\n",
" \n",
" \n",
" \n",
" | Equity 0 | \n",
" 1.917594 | \n",
" 1.894223 | \n",
"
\n",
" \n",
" | Equity 1 | \n",
" 0.298692 | \n",
" -0.211479 | \n",
"
\n",
" \n",
" | Equity 2 | \n",
" 1.002879 | \n",
" -0.391872 | \n",
"
\n",
" \n",
" | Equity 3 | \n",
" 1.284868 | \n",
" 0.027782 | \n",
"
\n",
" \n",
" | Equity 4 | \n",
" 1.159811 | \n",
" 0.524842 | \n",
"
\n",
" \n",
" | Equity 5 | \n",
" -2.483890 | \n",
" -0.573875 | \n",
"
\n",
" \n",
" | Equity 6 | \n",
" 1.647554 | \n",
" -0.452927 | \n",
"
\n",
" \n",
" | Equity 7 | \n",
" -1.889101 | \n",
" -1.785987 | \n",
"
\n",
" \n",
" | Equity 8 | \n",
" -0.121327 | \n",
" -0.384685 | \n",
"
\n",
" \n",
" | Equity 9 | \n",
" 0.190503 | \n",
" -0.716920 | \n",
"
\n",
" \n",
"
\n",
"