Are machines taking over the markets? Does an individual intraday trader even have a chance at getting a return when algorithms snatch away profit?
Surprisingly, the answer is yes to both questions.
The rise of algorithms or programs that look for patterns and execute trades in and out of stocks in milliseconds (High Frequency Trading or HFT) has created a choppy sea that even technical signals can’t keep up with.
Today’s stock market is highly influenced by algorithmic trading. But because of enforcement to reduce spoofing (fake orders), glitches at major firms that caused big trading losses and the SEC’s analytics screening system, even proprietary traders are experiencing reduced profits.
The search is on for the next evolution in computing power that delivers more reliable returns.
So as an individual trader without the resources to build your own proprietary trading system or without the desire to sit back and watch the algorithmic trading system do your trading for you, can you jump on the evolutionary bandwagon and benefit?
Yes. A company founded by active traders who were tired of being scooped by the effects of HFT and algorithms decided to fund the development of a predictive analytic platform using machine learning, a form of artificial intelligence.
EOTPRO Developments are motivated to give individual traders like them an edge over the effects of proprietary algorithms.
It’s taken their team of data scientists and traders six years to get DeepStreet EDGE V 2.2 right. It was, as you might expect, a far more difficult challenge than they first expected.
Here is the how and why behind why it works.
The Trading Power Behind Predictive Analytics
What is predictive analytics and how is it any different than what you or proprietary traders are using today to select trades and time your entry and exit points?
An optimized machine learning model today can comb through big data, and determine what price and direction a stock or an index is about to move to next, before it actually happens. This predictive output provides a true leading indicator.
To evaluate whether you should add a predictive analytics platform to your trading system, you will want to understand how machine learning works and how far the development of predictive applications have come.
Before we dive into the mechanics, you may be curious to see DeepStreet EDGE’s predictive power in action.
Case in point, global macro events always destabilize the markets when you don’t expect it.
Central bankers from around the world issue opinions on monetary policy randomly, without notice and the markets react. Your long trade is suddenly moving against you and your signals did not give you a heads up until the move has been made.
Algorithmic trading won’t give you a heads up either before or during such events. A computerized trading strategy also relies on lagging indicators to spot patterns that meet its criteria. The instruction may or may not have been looking for the global macro pattern.
In Figure 1 below, DeepStreet EDGE recognizes that a central banker is about to speak (a global macro event) and provides a subscriber a notification BEFORE the market reacts.
Figure 1 – DeepStreet EDGE NQ Chart October 24, 2016 showing the power of knowing in advance that a central bank governor (Charles Evans) has just released a speech that is about to move the market. How many trading opportunities can you see in this chart as a result of this prediction?
How is this prediction even possible if the governor has not yet spoken? Read on so you get a better perspective on how machine learning and predictive analytics can give the individual trader an edge without having to hire a team of data scientists to interpret the results.
It’s important to think about the time line that traders use to look for an entry or exit signal.
Traditional technical analysis uses signals to let the trader know when patterns that have produced a reward in the past are present again in the moment. The theory is that if the same pattern is present now, then entering a specific trade will deliver the same result as before.
Algorithms are designed to find these patterns in the past too.
Before Machine Learning, there was no realistic alternative to this rear view.
While algorithmic trading used to provide an edge, it may surprise you to hear that today even Hedge Funds struggle to find strategies that beat the S&P or simple index funds, despite the investments they have made in algorithmic trading.
Further many hedge funds are chasing the same strategies so there is little to exploit now. So if the pros can’t find returns, why would you want to bother with algorithmic strategies either, or even trust your technical trading signals?
Your trading system may need an overhaul too just like proprietary traders are discovering.
Let’s look at the MACD, in figure 2 below, this signal is marketed as a ‘leading indicator’ which is supposed to predict when to enter a trade.
But look more closely as to how all the volatility from algorithmic trading and HFT has made the signal late to the party.
Figure 2 – the MACD has become an unreliable signal due to algorithmic and HFT produced volatility.
You will notice in the chart above when the MACD give you a signal to the down side (note the red bars), the market moves up and against you first. You may have gotten out of your trade having lost before the market actually moves down again. Looking at the green bars, the same dynamic occurs, the market drops, then moves up and later, doesn’t let you know about the next drop before moving back to the previous price.
The algorithmic trading instructions exploited the opportunity first, so the individual trader is working with stale information that the MACD can’t compensate for.
So what do you do when you can’t find alpha, don’t have the funds to beat the computerized systems and can’t rely on traditional signals anymore?
Some are going to index funds. Others are looking to new forms of big data and machine learning to super charge their strategies and returns again.
Chasing Alpha Drives Evolutionary Changes In Trade Selection Technology
If algorithmic trading is now failing the market, what’s next? Machine Learning mining big data for other patterns, or artificial intelligence.
One source of big data is social media. Scoring the sentiment analysis provides the ability to determine how a social media tweet or news story would affect a stock.
There are several companies like DataMinr that have commercialized tools to speed up the predictive capability of sentiment analysis. They have raised millions of dollars from Wall Street and Silicon Valley in their bid to be the first to predict stock movement.
But is determining whether Starbucks stock will move because a series of tweets came out with the name of the coffee company in the tweet actually Artificial Intelligence at work?
Not quite. Trading social media scored with sentiment analysis powered by natural language processing and machine learning is deceptive and possibly even confusing. There is a lot of ‘noise’ in those tweets.
Let’s examine what you would need to have in place to actually reduce the risks that sentiment analysis alone introduce to the intraday trader.
You have to give the model and the machine the right base of data to learn from. The creation of your model starts with having the right data sources. What is the primary mover of stock prices? It is not historical data.
It is News. And not any news. You need news streams that have yet to impact the market.
In simple terms, to build your predictive model you feed in vast streams of news, not just social media and tweets. Then show the machine what happened to every stock and index as a result of the news. This feedback loop increases the machine’s confidence that it can predict price movement even more accurately with the next piece of news on any stock.
Then you multiply all that activity by thousands of news items every day. Social media is only one source of activity that could, or may not move a stock.
You also need to then aggregate those thousands of changes in price movement of every stock so you can make the prediction as to how all that news will move an entire index.
You are now talking about collecting all the inputs that affect price and movement so that your particular machine learning model can spot thousands of patterns and relationships between disparate data that humans can’t, and turn these relationships into increasingly accurate predictions.
Up until now, this was an immense computation problem as a well as a data source problem. It is very difficult to get all the right input feeds that contribute to stock price movement… and house that data in one massive location so it can be analyzed.
For the active trader, having access to a tested machine learning model with the right data sources can level the playing field against the effects of algorithmic and high frequency trading.
Every iteration in the learning model increases the accuracy of the probability. Or in simple terms, the platform becomes more capable over time about predicting that a stock or index will move to a certain price.
Up until now, you needed millions of dollars and access to data scientists to develop such a learning model. The average retail investor has not been able to tap into the power of that predictive engine.
But even those well-funded hedge funds able to employ data scientists have struggled with interpreting how to best interact with such enormous data sources and pull out actionable, immediately useful trades and signals to time entry and exits.
As an individual active trader, should you even care about the struggles the pros are having with developing their own machine learning models?
Yes.
For two reasons: First, developing a reliable machine learning model means managing a team of software, math and data scientists while evaluating the merits of their latest models. These are not skill sets easily available on the market. Traders, not software project managers or data scientists run the average Hedge Fund.
The models that have been created so far inside hedge funds are friendly to data scientists, but unreadable to active traders.
While the big institutions have been trying to solve the problem of what data to include, what to exclude and how to interpret, train the model and forecast a news alert for their own benefit, EOTPRO has been quietly developing a solution to give Active Traders their own advantage.
For instance, as a trader, you want to know how useful any social media source is to any stock’s movement. However, unless you have the right data sources and model to compare social media to, you can’t actually prove the predictive value of Twitter. A single data source like social media seems tantalizing but leads you down the garden path. You can produce cool charts and visuals showing the sentiment score that excite people about the potential. But the active trader would not know if the scores have high predictive value or not, until the trades don’t produce any better returns than other indicators.
EOTPRO proved this year that social media and sentiment analysis has a 2% positive predictive effect compared with numerous other data sources. What this research means is that if your model is exposed to a lot of social media you have to be able to filter out what is noise and what is market moving.
How would you know if your technology platform has the correct filters? The data scientists know, but the hedge fund manager might not unless they know to ask such questions.
How come more data comparison to prove models is not done? With the state of computing power on the planet today, it can take months to prove the value of a single model.
Given that you could test thousands of permutations in any given model, the chances of discovering what is noise and what is worth paying attention to are low unless you’ve tested discrete data sources one at a time before bundling them altogether in various combinations.
A systematic validation of data streams before throwing it into the hopper with the rest is an essential discipline that takes time. Hedge funds are losing money now. They need new strategies. They may skip essential, but not well understood steps.
So as a trader, if you have doubt about your trading system then you don’t have a credible user experience. There may be little value being extracted from those millions being invested in machine learning.
The team at EOTPRO didn’t just want to ask big data questions to uncover the sentiment of thousands of news stories. They wanted to glean advanced knowledge (true leading indicators) before any other trader gets to see that news, so that their fellow individual traders had an equal playing field against the supercomputers.
So they measure the effect of each data source so they know what has predictive value and what needs filtering out.
In developing DeepStreet EDGE, the EOTPRO founders wanted to know:
- When to trade any stock or option in the Dow 30 or top 100 stocks on Nasdaq, the S&P 500 or the Russell.
- When to trade futures or ETFs
- What price change to expect from the statistical probability assessed before stepping into a trade.
- What level of confidence they could expect from the trade, based on the statistical probability.
- How long it would take to get to that price.
- When to get out.
- When not to trade at all.
Seven seemingly simple goals that would replace out-dated, out-maneuvered, trading signals that computerized trading have made irrelevant and in fact risky for retail traders to rely on.
There were two failed attempts before EOTPRO found the correct way in to wrestle big data from 45 different data sources to deliver on their goals. It’s how they discovered the poor predictive value of social media!
The 8 Goals That Give Traders a Reliable Edge Even During Volatility...
The Crucial Difference Between Algorithmic Trading & Machine Learning
Machine Learning actively interprets and scores numerous data sources to predict a trend.
But all Machine Learning technology is not created equal.
The data sources the machine learns from and the visual analytics that tell you what to do with that prediction trend is what creates a reliable leading indicator.
Remember the global macro event? How does DeepStreet EDGE know in advance what a central banker anywhere in the world is about to say before they speak? EOTPRO has secured access and tested the right sources of early news, before it’s released to the web.
The real question you would want to ask yourself is, how reliable are any of these technologies at modifying and managing your risk in a way that makes trade selection, entry and exit fulfill your expectations?
The 8 Goals Your Technology Platform Should Deliver To Equal The Playing Field to Counter The Effects of Algorithmic Trading
To evaluate whether any of the many platforms offering predictive reliable accuracy that could provide you with an edge, these 8 goals must all be part of the predictive analytics platform so you have useful and actionable intelligence:
- Early News You want deep data from news sources that arrive into your big data stream EARLIER than journalists receive a news alert, so you know that the data source has yet to be priced into the market.
- Accurate Scoring The Machine Learning Model then has to have an effective ‘scoring’ system so you know that it is rating the value of that news to move a stock in a direction AND filtering out the irrelevant noise.
- Price Prediction By Stock Prices change fast. In fact HFT is designed to exploit latency in stock prices, which the average human can’t compete with. You need a model that tells you how much a stock will move so that you know whether the trade is worth taking BEFORE HFT acts.
- Confidence in the Prediction You need to know the confidence the Machine Learning model has with respect to any stock price movement. You might not want to take trades that have less than 70% statistical probability of moving to a certain price. Lagging indicators can’t give you this kind of risk management for your pool of capital. Neither can cool sentiment analysis tools.
- Timing How long will it take for the stock to move to that price? You need to know how soon the price prediction will be reached, so that you don’t get stuck in a trade that’s retracing against you.
- When to Exit Knowing in advance that a stock will only move up to a certain price and within a specific time frame allows you to time your exit which is the second most important risk-reducing activity after selecting the right trade.
- When NOT to Trade Stocks don’t move up or down in isolation of other macro variables and news sentiment. There are times when a stock might have a great entry point but other uncontrollable variables (Brexit, Elections, The Fed, Social Unrest) are working against you. You need a platform that alerts you to this, BEFORE you take a trade so you can protect your pool of capital.
- A Visual Dashboard Because of the pace of the market, you need a visual graphic ‘at-a-glance’ dashboard that tells you which stocks or futures to pay attention to before it happens. You want to learn how it works quickly and easily without having to be a data scientist or read lengthy manuals.

Figure 3 – This is the DeepStreet EDGE dashboard. You can see that the machine learning model suggests that JPM will move (green line with boxed ends) up $0.09 in the next 44 minutes. The model has a 77% confidence in this prediction. Below the JPM chart are other trader alerts that the model says have predictive capacity to move to a certain price. Subscribers listen to EOTPRO’s traders interpret what’s going on as well as take traders for their own accounts. This real time live audio training (see Club Room Audio box at bottom right) at market open is how subscribers learn to use the platform quickly.
If you are a futures trader, you can see in Figure 4 below how DeepStreet EDGE is clearly pointing in the direction the market is about to move next (green arrow at 53% confidence).
Figure 5 – The green lines and red lines are the past predictions for the NASDAQ overlaid on the actual market trajectory. The green arrow is pointing out in front of price, before the market moves. Traders are able to add multiple other signals to these charts to customize their view. This means you can combine your traditional trading system with DeepStreet EDGE and have the best of both a predictive analytic system and traditional technical analysis.
Summary
Lagging indicators, employed by most traders today, are quickly becoming out-dated and un-useful as trading tools because of HFT and algorithmic trading. Even Hedge Funds can’t find opportunities to exploit using these tools.
Machine Learning technology can give the individual Active Trader a true equal playing field, and even an edge against the effects of algorithmic trading employed by hedge funds and big banks.
It is vital to use a platform that delivers actionable knowledge with a precise level of confidence you can rely on so you can manage your risk and capital.
Use these 8 goals to evaluate whether other innovative technologies give you what you really need before investing time and energy in changing or adding to your technical trading system.
THE SPECIAL OFFER
So how accurate has EOTPRO’s model become? Join us in the Club Room at market open and ask Bill Dennis the head trader that question. We update our accuracy scores every day.
DeepStreet EDGE is Machine Learning in action, which you can experience by getting a free 30-day trial made possible by the publishers of this book. If you are an Intraday experienced trader, go to this link to activate your trial subscription
Get an All Access Pass to Deep Street Edge Here!
ABOUT THE AUTHOR
Lorraine McGregor is the Vice President of Sales and Marketing for EOTPRO Developments. She is an active options trader and interprets how artificial intelligence and big data are starting to affect the trading experience for active traders.
EOTPRO has spent the last five years building an artificial intelligence platform using disruptive technology and collective intelligence that today is able to provide reliable statistical probability trading ideas for stocks, futures and options traders on the Dow, NASDAQ, S&P and the Russell indices. It is far more advanced than sentiment analysis.
Using the supercomputer and machine learning that powers DeepStreet EDGE provides answers for the following questions: How likely is the stock or index to go up or down? How far will the stock move? How long will it take to get there? The user interface is designed to make the complex simple: Current subscribers report they are able to use the platform easily within a few hours.