A Computing Platform For Pairs

Forex Featured020

Table of Contents

If the price of Pepsi rose to close that gap in price, the trader would make money on the Pepsi stock, while if the price of Coca-Cola fell, they would make money on having shorted the Coca-Cola stock. The difficulty comes when prices of the two securities begin to drift apart, i.e. the spread begins to trend instead of reverting to the original mean. Dealing with such adverse situations requires strict risk management rules, which have the trader exit an unprofitable trade as soon as the original setup—a bet for reversion to the mean—has been invalidated. This can be achieved, for example, by forecasting the spread and exiting at forecast error bounds. A common way to model, and forecast, the spread for risk management purposes is by using autoregressive moving average models. Statistics play a crucial role in the first challenge of deciding the pair to trade. The pair is commonly chosen from the same basket of stocks, for instance, Microsoft and Google or ICICI & Axis or Nifty Index and MSCI index .

Ok so it looks from the chart as if we have around price data downloaded for around stocks; this should be more than enough to find at least a couple of co-integrated pairs to run our backtest over. I started this blog a few years ago, and one of my very first blog series was on this exact subject matter – mean reversion based pairs trading. Well this time I am going to add a few more elements that were not present in the initial blog series. The first task is to set the time and invested members to be equal to None, as they will be updated as market data is accepted and trade signals generated. latest_prices is a two-array of the current prices of TLT and IEI, used for convenience through the class. In our next stage, we want to pre-select eligible stocks that enable us to sail through further steps. First, we removed stocks that were delisted, exchanged, or merged during our sample period since those stocks are no longer tradable.

Trading Strategy Performance Report In Python

In a production environment it would be necessary to adjust this depending upon the risk management goals of the portfolio. We also need to create a backtest file to encapsulate all of our trading logic and class choices. The particular version is very similar to those used in the examples directory and replaces the equity of 500,000 USD with 100,000 USD.

What is the best algorithmic trading software?

Best Automated Trading SoftwareBest Overall: MetaTrader 4.
Best for Options Trading: eOption.
Best for Stock Trading: Interactive Brokers API / FIX CTCI.
Best for Forex: MetaTrader 4.
Best for Premium Access: Zen Trading Strategies.

By default, the Alpaca API will return today’s candle even though it is not closed. We will be checking the last 20 days to see what the maximum divergence has been over that period. A nice feature of the Alpaca library is that it will automatically attach a Pandas DataFrame to the object that’s returned when we query for historical data. Next is a simple custom function that uses the API to determine when the market opens next. If the market is not open, the script will sleep until it opens.

Step 5: Model Selection

In our previous post, we looked into implementing a Kalman filter in R for calculating the hedge ratio in a pairs trading strategy. Pairs trading is considered a market neutral form of trading that is uncorrelated to common wealth enhancing investing such as stock and bond trading.

How do you develop a trading algorithm?

Here are the steps for coding an algorithmic trading strategy: 1. Choose product to trade.
2. Choose and install software.
3. Set up an account with a broker.
4. Understand our strategy.
5. Understand and setting up your MT4.
6. Understand the parts of a MT4 trading algorithm.
7. Code the rules for entering and exiting trades.
More items•

In the new approach, the number of parameters will more than double if the number of regimes is increased to three. No explanation, however, from data sources such as Bloomberg and the Internet can be found. The occurrence of spikes appear data-specific, and looking at the same time series of spread from 2002, the spikes happened again quite frequently. To implement the rest of our pairs trade, we need to decide the z-score levels at which to trade and implement the logic for buying and selling the spread. Anyway, that’s for you to mull over and keep in mind as you read this series. Right now, we’re just going to present one version of a pairs trade in Zorro.

The strategy has a low drawdown, i.e. the max loss in a trade is significantly low, allowing flexibility in the use of leverage. For the remaining price ratios, we formulate a z-score table to understand the distribution and deviations from the mean. We will be using a time horizon of 60 days to calculate the moving average and Standard deviations. 3.Arbitrage strategies –These are event-driven strategy, and are thus market movement-neutral. Arbitrage can be across different Stock Exchanges, geographies, currencies etc.

Data Imputation

Join the QSAlpha research platform that helps fill your strategy research pipeline, diversifies your portfolio and improves your risk-adjusted returns for increased profitability. Parameter Optimisation – Varying the parameters of the Kalman Filter via cross-validation grid search or some form of machine learning optimisation. However, this introduces the distinct possibility of overfitting to historical data. It also has a long maximum drawdown duration of 777 days – over two years! Note that this strategy is carried out gross of transaction costs so the true performance would likely be worse. The equity curve begins relatively flat for the first year of the strategy but rapidly escalates during 2011. During 2012 the strategy becomes significantly more volatile remaining “underwater” until 2015 and reaching a maximum daily drawdown percentage of 15.79%.

This is because our original choice was clearly overfitted to the sample data. In our model, we used rolling parameter estimates and may wish to optimize window length. We may decide to simply iterate over all possible, reasonable window length and pick the length based on which our model performs the best . Below we write a simple loop to to score window lengths based on pnl of training data and find the best one. Notice how we sometimes make money on the short leg and sometimes on the long leg, and sometimes both. We’ve seen that our two securities are cointegrated so the ratio tends to move around and revert back to the mean. It seems our features should be certain measures for the mean of the ratio, the divergence of the current value from the mean to be able to generate our trading signal.

Quantitative hedge funds do this and they might have thousands of stocks and make thousands of trades in their high-frequency strategy. However, if you choose to trade 2 stocks, consider doing it over a limited time period (e.g. during the COVID-19 crisis, cruise stocks move together) or use another qualitative layer of analysis. Spread here is based on the hedge ratio which is updated on daily basis. So the daily mark to market pnl should be based on spread by t – 1 hedge ratio but not on t ratio, I.e settle your existing pair portfolio before getting into a new one. That scrapes all the NYSE stock tickers – its a LOT of tickers so you may like to be a bit more selective as it will take quite a long time to run that many stocks through the data download.

Constructing A Beta Neutral Hedge In A Pairs Strategy

We performed such test to all pairs and select at least one pair in each cluster to diversity our portfolio. Then, a strategy that observed based on the movement of the spread can be designed and executed well in the later part. Thus, we are over 95% confident to say that the spread of WDFC and HSIC’s stock price is stationary and they are valid pair. The resultant reduced datasets are then concatenated to create a 20 dimensional training dataset which we then use for clustering analysis. If one has an up day, the other will probably have an up day, and vice-versa.

Since QSTrader handles the position tracking, portfolio management, data ingestion and order management the only code we need to write involves the Strategy object itself. This data will need to placed in the directory specified by the QSTrader settings file if you wish to replicate the results. $e_t$ represents the forecast error or residual error of the prediction at time $t$, while $Q_t$ represents the variance of this prediction at time $t$. “Longing the spread” here means purchasing $N$ units of TLT and selling $lfloor$, where $lfloor$ is the “floor” representing the highest integer less than $x$. The latter is necessary as we must transact a whole number of units of the ETFs.

In “HMM-extended Kalman filters”, the construction of the filtering algorithms for the parameter estimation as well as the trading strategy is outlined. The effectiveness of the proposed approach is investigated in “Numerical demonstration” by setting trades based on simulated and historical data, where the spread process exhibits a strong non-normal behaviour. We apply our trading strategy to the real stock market and check how much we can earn based on our approach. For the training data, we used the previous 700 days stock prices.

By default, we’re trading 100 units of the spread at each level. We’re trading in and out of the spread as the z-score moves around and crosses our levels. If the z-score crosses more than one level in a single period, we’d be entering positions for each crossed level at market. If you want to use options on two different instruments, you need to completely understand the risks that you are taking before you enter each transaction.

We calculate the variance-covariance matrix R or set it to the zero-matrix if it has not yet been initialised. Subsequently we calculate the new prediction of the observation yhat as well as the forecast error et. $y$ is set equal to the latest price for IEI, while $F$ is the observation matrix containing the latest price for TLT, as well as a unity placeholder to represent the intercept in the linear regression. The Kalman Filter is subsequently updated with these latest prices.


While some brokers offer options on pairs, the process of buying and selling calls and puts on two different pairs generates multiple risks that are not always easy to manage for an inexperienced trader. This strategy takes advantage of pairs that are co-integrated where the relative value changes intra-day. This could be a standard pair, such as gasoline versus crude oil or even a pair between a futures contract of a currency and a cash currency pair.

A co-integration of 1, means that the pair is perfectly co-integrated, while a co-integration of -1 means that there is absolutely no co-integration. You can measure the co-integration of a cross currency pair by breaking it down into currency pair versus the US dollar. For example, you could take the GBP/USD and run a co-integration study versus the USD/JPY, to determine if the GBP/JPY is co-integrated. While correlation describes the returns of two different assets, the statistical measure co-integration describes how well each assets returns are linked, and the strength of their correlation.

However, they are nothing but a bunch of rules that are used on a Stock Exchange to automate the execution of orders without any human intervention. Algorithmic Trading Strategies require a quantitative model which runs upon live market data parameters such as price, volume, etc. and ensure apositive Alphaat the end of the trade. This is a general multi-asset, multi-currency pair and spread trading algorithm for two instruments. The term “pair trade” describes a trade in which two instruments are traded in equal currency units and a “spread trade” refers to a trade in which the two instruments are traded in a ratio of shares or contracts. This trading strategy leads to a substantial decrease in the overall risk exposure. This post presents a script for a pairs trading algorithm using Zorro.

Further examination is required on how to produce filtered parameter estimates within their specified constraints without, or at least with minimal, additional numerical optimisation procedures. The results in Table 3 show that the aggressive strategy does not necessarily produce a significant increase in profits. In investing all earned capital in the spread portfolio, the main part of the profit is collected whilst the investor takes a long position on the portfolio.

However, losses are incurred when prices reach the stop-loss boundaries after the portfolio is opened and do not return to the average. Furthermore, after the portfolio is opened, if the trading signal is not reversed to mean during the trading window, the portfolio is closed by force; this is called the exit position of the portfolio. Before designing the model, it is essential to determine which stocks will be eligible for trading. Only tickers with complete set of data points over the time horizon and a minimum daily trading volume of 15 million INR are eligible for trading.

Pairs Trading (cointegration Approach)

But it’s not unusual to see the relationship between two assets suddenly change. An example of this is the correlation between gold and equities. But as of the summer of 2019, they’ve had a positive correlation. Text is available under the Creative Commons Attribution-ShareAlike License; additional terms may apply. By using this site, you agree to the Terms of Use and Privacy Policy. Wikipedia® is a registered trademark of the Wikimedia Foundation, Inc., a non-profit organization.

Mary Davis
My name is Mary Davis. I am successful broker. I want to share my experience with you through tutorials and webinars. For any questions of interest, please contact us by e-mail: [email protected]. +1 973-709-5130


No comments yet. Why don’t you start the discussion?

    Leave a Reply

    Your email address will not be published. Required fields are marked *

    Scroll to Top