Lokesh Madan Lokesh Madan is a strategy business consultant for various high frequency trading companies worldwide with more than 12 years of experience in financial technology, research work and business development

What High Frequency Traders Do – BAD or GOOD for Markets?

9 min read

What High Frequency Traders Do – BAD or GOOD?

Over the past few Year’s, there has been a quick shift towards algo / Quant HFT based trading, Where as Asset managers make 24% return in market & HFT traders make 300% Return. Both among long-term investors using execution algorithms to lower trading costs and short- term investors automating market making and statistical arbitrage strategies. These short-term investors, popularly known as high-frequency traders (HFTs), account for a substantial fraction of total equity market trading volume. However, there is fairly little known about how their trading effects liquidity. To the extent that HFTs act simply as market makers, they will tend to improve liquidity. But HFTs also search trade and order data for clues about where prices will go in the future, and when they trade on this information, they may compete with long-term investors for liquidity, thereby increasing those investors’ trading costs.

In this article we also try to examines the extent to which HFTs profit from traditional asset managers’ trading by anticipating their future order flow & Weather HFT traders provide liquidity real time or fake way. Traditional asset managers, such as mutual fund and pension fund managers, typically split their large trades into a series of orders executed over the course of one or more trading days. They split their trades, because a series of small trades move prices less than a single large one. But in splitting their trades, traditional asset managers may reveal their trading intentions to other Trading Technology Experts investors who may end up trading ahead of or alongside the traditional asset manager.


To examine these issues, I examine return and trade patterns around periods of aggressive buying and selling by HFTs using trade data from the NSE Stock Market. Specifically, I focus on HFTs’ aggressive trades, that is, trades where an HFT initiates the transaction by submitting a marketable ( IOC or bidding process) buy or sell order, which are functionally equivalent to market orders, because it is a simple way to screen out liquidity providing trades ( Limit Order traders). I test whether HFTs’ aggressive share purchases predict future aggressive buying by non-HFTs, and whether HFTs’ aggressive sales predict future aggressive selling by non-HFTs.

HFT user react more fastly as compared to non HFT users.

I find evidence consistent with HFTs being able to anticipate order flow from other investors. In tests where Stock Futures are sorted by HFT net marketable buying at the some micro second horizon, the Stock Futures bought most aggressively by HFTs have cumulative standardized non-HFT net marketable buying of 0:66 over the following thirty micro seconds,) and stocks that are sold most aggressively have cumulative standardized non-HFT net marketable buying of 0:68 over the next thirty micro seconds. Moreover, the stocks Futures HFTs buy aggressively have positive future returns, and the stocks Futures they sell aggressively have negative future returns.( By using scalping Calculation various Time series based methods – HFT traders do so).

Taken together, these two results suggest HFTs’ aggressive trades forecast price pressure from other investors.( Actually HFT trades use CEP to analysis per Micro second Market Future prediction method by using various methods).

I consider and reject the most likely alternative explanations for these results.

To rule out the possibility the results are driven by HFTs responding to news faster than other investors, I rerun the sort tests excluding periods within few second of the publication of intra-day corporate news by Bloomberg Terminal. The results excluding periods around intra-day news are nearly identical to those for the full sample.

A second explanation is that HFT and non-HFT trading are driven by the same underlying serially correlated process (i.e., same trading signals), so HFT trading predicts non-HFT order flow only because it is a proxy for lagged non-HFT trading. A final alternative is that if non-HFTs chase price trends, HFTs might actually cause future trading by non-HFTs through their trading’s effect on returns. However, the lead-lag relationship between HFT and non-HFT trading is robust to controls for lagged non-HFT trading and lagged returns, which is inconsistent with the second and third alternative explanations.

I also examine whether there are cross-sectional differences in how well different HFTs’ trades forecast future order flow. Perhaps some HFTs are more skilled or focus more on strategies that anticipate order flow, while others focus on market making or index arbitrage. My evidence indicates that there are indeed differences among HFTs. Trades from HFTs that were the most highly correlated with future order flow in a given month have trades that also exhibit stronger than average correlation with future non-HFT order flow in future months.

DARK POOL and HFT Traders.

There are two types of market’s for trading one is displayed order book type & second one is non- displayed order book type. In 2009, NASDAQ’s market share was roughly 35% in NASDAQ-listed securities, and 20% in securities listed on the NYSE. The remainder of U.S. equity trading is spread among trading venues with displayed order books, such as the NYSE and BATS, and trading venues with non-displayed order books, such as ITG’s POSIT Marketplace, Credit Suisse’s Cross finder, and Knight Capital. Much of U.S. trading occurs on non-displayed trading venues. This part is not valid for Indian Exchanges – it also have both market type’s but large or most volumes are with displayed only order book where as some bulk deals are not visible to Order book in some cases. As you know dark pools are not allowed at India.

There was once a large difference between the market structures of different displayed markets such as the NYSE and NASDAQ. Trading on the NYSE was conducted in an open outcry trading pit, and while NASDAQ had an electronic order book with market maker quotes, brokers had to call the market maker on the phone to execute a trade. Now, displayed markets are all structured as electronic automatic execution limit- order books, and they largely compete on price. Though the NYSE still has specialists, now known as Designated Market Makers (DMM), they are for the most part the same electronic trading firms that make markets on other exchanges. For example, GETCO LLC, a large electronic market maker, became a DMM for 350 NYSE stocks in early 2010 (Kisling 2010) and is also a registered market maker on NASDAQ, BATS, NYSE ARCA, and the CBOE (McCarthy 2010).

Executions in displayed markets predominately come from professional traders. Few retail orders reach the displayed markets directly. Most retail brokerages have contracts with market making firms who pay for the right to fill retail orders. For example, in the third quarter of 2009, Charles Schwab routed more than 90% of its customers’ orders in NYSE-listed and NASDAQ-listed stocks to UBS’s market making arm for execution. Similarly, E*Trade routed nearly all its customers’ market orders and over half its customers’ limit orders to either Citadel or E*Trade’s market making arms. However, when there is a large imbalance between retail buy and sell orders in a stock, market making firms likely offload the imbalance by trading in displayed markets, so there is some interaction between retail trading demand and the displayed markets.

High-frequency traders account for a substantial fraction of equity market trading

volume. High-frequency traders are proprietary trading firms using high-turnover auto-mated trading strategies. Examples of such trade ring firms include Edelweiss Capital, Way2Wealth, Motilal Oswal Financial Services, Ltd. Angel Broking, India Infoline (IIFL), SMC Global Securities, Anandrathi Financial Advisor & many more. Such firms likely engage in some combination of market making and statistical arbitrage.

Data used By HFT Traders

Primarily HFT Traders Use High Speed TBT data at Colocation & for testing purpose they used Level 2 historical data.

How we Identify high-frequency traders ( Or HFT Trading Firms )

Mostly Prop shops with Colocations Setup with better low latency hardware’s / Low latency networking / Ultra Low latency OMS engines & more Interactive & TBT Links from exchange ( 400 messages links ) are fall under this category.

Or Trade records on exchanges include a Market Participant Identifier (MPID) indicating the broker/dealer making the trade. A broker/dealer may have multiple MPIDs that are used by different business lines or customers. A typical reason for a customer to have their own MPID is that they have sponsored access ( Colocation ID with Market making expertise). In a typical sponsored access arrangement, the customer handles connectivity with an exchange and has limited interaction with the broker/dealer’s trading system. Sponsored access arrangements are motivated by exchanges’ tiered pricing schedules that give better pricing to higher-volume broker/dealers.

Customers accessing exchanges through sponsored access agreements get direct market access and the lower fees of the larger sponsoring broker/dealer. These sponsored access customers tend to be large, active traders or smaller broker/dealers. One consequence of these arrangements is that it is possible to observe activity of sponsored access customers directly rather than at the aggregated broker level.

The data from NSE classifies market participants as either a HFT or a non-HFT.

Firms were classified as HFT firms using a variety of qualitative and quantitative criteria.

The firms ( Prop Shops with Colocation setup ) classified as HFTs typically use low-latency connections and trade more actively than other investors. Their orders have shorter durations than other investors, and they show a greater tendency to ip between long and short positions in a Market during a day. They are Volume driver’s.

How HFT Traders find the Asset class for its working.

A reasonable definition of the set of Asset classes traditional HFT Traders invest in is those in either most liquid products i.e Nifty 50 or bank Nifty or products those are not touch by most come Traders i.e Or Shares fall under those Index’s or they do Trading based Corporate news.

Or they do trading depends upon the Volatility on Asset classes – Small-cap stocks are slightly more volatile than mid and large-cap stocks. The standard deviation of small-cap stocks’ daily returns is 4.5%, compared to 3.5% for mid-cap stocks and 2.5% for large-cap stocks.

HFTs are relatively more active in large-cap Futures Stock & Index’s. Their median share of total volume is 14:8% in small-cap stocks, 29:2% in mid-cap stocks, and 40:9% in large-cap stocks. It is conceivable that since HFTs’ comparative advantage is reacting quickly to market events, they find more profit opportunities in Futures & Options for which quoted prices and depths update frequently.

How HFT Trades Take advantages of Trade imbalances.

There are two types of trade imbalances: marketable imbalances and buy- sell imbalances. The marketable imbalance is a common measure of buying and selling pressure .The buy-sell imbalance is simply shares bought minus shares sold and has been used to measure position changes of different investor groups

The simplest explanation of marketable imbalances is that they are essentially the number of shares bought with market orders minus the number of shares sold with market orders. There is actually no such thing as a market order in the NSE system, but any order to buy with a limit price at least as high as the best ask or order to sell with a limit price at least as low as the best bid is essentially the same thing as a market order. The party submitting such an order is said to have submitted a marketable order, and the trade executes immediately. If the marketable order was a purchase, the trade can be said to be buyer-initiated, and if it is a sale, the trade can be said to be seller-initiated. Subtracting shares sold with marketable orders from shares bought with marketable orders gives the marketable imbalance measure. If investors with resting limit orders in the order book are passive liquidity suppliers, then the marketable imbalance is an intuitive measure of trading demand.

Intra-day returns for HFT traders

Intra-day returns are calculated using bid-ask midpoints from two sources. The primary source for quotes is the National Best Bid and Best Offer (NBBO). The NBBO aggregates quotes from all displayed order books.This is the best definition of a market price, but one disadvantage is that due to latency between exchanges and data feeds, the time stamp on the NBBO feed is not guaranteed to be synchronized with the time stamp for NSE trade data. To ensure results are robust to misalignment of timestamps between the quote and trade data, I redo some results using the NSE best bid and best offer, or NSE BBO. The timestamp of quotes in the NSE BBO is precisely aligned with the timestamp of trade data. Creating the NSE BBO is computationally intensive, so I replicate only a subset of the results using the NSE BBO.

The quote feeds occasionally contain absurd price data. Every once in awhile, for a fraction of a second, the bid might be above the ask, or for a stock that typically trades at 125 Rs/, the best bid might be 0.25 Rs Tick . Returns calculated from such quotes do not accurately react changes in firm value. I use filters to screen out these prices. I remove quote updates where the bid is greater than the ask or where the bid-ask spread is more than 20% greater than the bid-ask midpoint. To fix an issue with bad pre-market quotes on the NSE, the last of which is used as a proxy for the opening price, I throw out the last price before the open if there is more than a 20% difference between the last pre-open bid-ask midpoint and the first post-open bid-ask midpoint. After the application of these filters at the tick level, the global minimum and maximum one-second returns across all stock days, reported as -37% and 70%. While extreme, given the number of one-second observations in the sample, they do not seem like they would meaningfully affect the results. One could apply a stricter filter, such as requiring that the magnitude of one-second returns be less than 20%, but I do not think this would meaningfully affect results.( Data Used 16 august 2013 for 5 NSE Stocks)

HFT Traders are Liquidity provider – Yes/No ..

HFT can play an important role as market makers, for example, generating trading volume on new electronic exchanges .Trade volume, however, is not liquidity but all too often mistaken for it. Liquidity means “there is a bid/offer on the other side when I need it, for the amount I need it (market depth) at a reasonable level (market breadth). Volume is not the same as liquidity, since volume is approximately like the product of liquidity x velocity, and a large volume does not necessarily imply a large liquidity. This is illustrated by the May 6 flash crash when a fundamental trader’s algorithm started selling based on previous trade volume, creating a positive feedback between its own selling and the trading activity of other market participants.

The same event also demonstrated that HF Traders can turn into significant liquidity takers; while they are liquidity providers when it suits them (they have no obligation to make quotes). This is also described as “flow toxicity”, when market makers provide liquidity at their own loss or when informed traders take liquidity from uninformed traders. In fact it seems HFT provides liquidity in good times when it is perhaps least needed and takes liquidity away when it is most needed, thereby contributing rather than mitigating instability.

A recent report showed that the frantic development of HFT has slowed down in developed markets, and there is a transfer of activity to emerging markets such as Russia, India, Brazil and Mexico where exchanges are beginning to revamp their systems to attract such players. Low market volumes and stiff competition have led to a sharp fall in “high-frequency”. This illustrates the fact that, as HFT market participants flock into a given market, the opportunities shrink, dispelling the possibility for further growth.

It is also conceivable that HFT liquidity is provided at the expense of other market participants. Short term traders may be specifically prone to herd to the same information, driving the price further away from its fundamentals .The more momentum traders there are in a market and the higher the diversion from fundamentals, the fewer fundamental traders survive, further strengthening momentum traders. Various equilibrium are possible between short and long-term investors. The question is what is the right mix of investment strategies and horizons that best serves the well-functioning of financial markets and ultimately social welfare?

by Lokesh Madan

Add me skype – lokesh.madan3

Lokesh Madan Lokesh Madan is a strategy business consultant for various high frequency trading companies worldwide with more than 12 years of experience in financial technology, research work and business development

Amibroker Vs Tradingview Vs Python Which One to Use…

When choosing a platform for systematic trading, there are a few key factors to consider, including the features and capabilities of the platform, the...
Rajandran R
1 min read

Mini Certification on Automated Trading using Amibroker

If you are new to Automated trading? This certification course will help you to kickstart your automated trading with your broker and get to...
Rajandran R
1 min read

Finfolab Technologies developed API Solutions for Goodwill

Finfolab Technologies developed Arrow API solutions for Goodwill. Our API solutions enable retail traders and Goodwill commodities partners to design and deploy advanced algorithmic...
Rajandran R
32 sec read

Leave a Reply

Get Notifications, Alerts on Market Updates, Trading Tools, Automation & More