What I learned about HFT from day trading

Years ago, I briefly tried my hand at day trading with a company called Title Trading.  Title Trading is a proprietary day trading firm similar to Swifttrade (the pioneer) and somewhat similar to SMB Capital (SMB has an excellent blog).  These firms will take almost anybody, sit them in front of a computer and get them to trade very small amounts of money.  The end game is to find talented traders to day trade the company’s capital.

An example training exercise would be to buy shares of C at $3.00/share and try to sell them for $3.01 or $3.02.  The reason for trading C is because it has very high volume and stock is unlikely to quickly move by a few cents.

There is some skill involved in considering other factors when day trading (e.g. technicals, time of day, volume, time and sales, the size of the bid and the ask, etc. etc.).  This trading strategy used to work very well in the past.  After I joined Title, I quickly discovered that this strategy would not work.  For example, I would place an order to buy C at the bid an hour after the market opened.  Hours later, millions of shares of C would trade at the bid and the ask.  My order would not get filled.

This was very strange when I ran the math.  The bid size was a certain amount.  I don’t remember the exact figure but let’s suppose it was 500k.  On the time and sales, I could see that several million shares were traded at the bid price.  Yet, for some reason, my order continued to sit there unfilled.

Eventually I did some research on Google and discovered something called “sub-penny pricing”.  Apparently, some market participants are allowed to buy 100 shares of C at $3.0001/share.  On 100 shares, the buyer is paying an extra penny.  Somebody can jump in front of other orders by paying an extra penny per 100 shares.  There is a whole trading strategy where traders buy at $3.0001 and sell at $3.0099.  I will refer to this strategy as scalping.  If you set your trading software to show four decimal places in the time and sales, you will see these trades happen.  You can also see an example below for BAC:

BAC-time-and-sales-subpenny-example

The reality is a little more complex.  Exchanges and ECNs have a rebate structure.  They’re effectively another form of sub-penny pricing.  If you post an order on EDGEA, you can more or less bid for C at $3.0003 (Direct Edge keeps $0.0001, and you keep $3.0002).  So you can try to repeatedly buy shares at $3.0003 and to sell them at $3.0097, generating $0.0094 in revenue before various trading costs (exchange fees, commission, and SEC fees).

On top of this, there is something called “price improvement”.  Suppose that I sell 100 shares of C at the bid.  If there are orders on EDGEA and if EDGEA offers the best negative rebate (I believe it did at the time), then EDGEA would have the best displayed price.  I would expect to sell my shares for proceeds of $3.0003/share.  But this is not what happens most of the time.  I might get a price such as $3.0029/share.  Some party was willing to step in and pay me the NASDAQ rebate.  What’s happening is that there is intense competition over repeatedly collecting the bid/ask spread on C.  Some market participants are willing to offer better pricing than their competitors.  The real bid/ask spread was much smaller than 1 cent.

SEC Rules

SEC rules specifically PREVENT investors and institutional investors from engaging in price improvement and sub-penny pricing.  This gives broker-dealers, market makers, and other “intermediaries” a massive edge over everybody else.  They can play the scalping game and front run everybody else.  At Title, the tightest effective bid/ask spread I could offer was $3.0003/$3.0097.  Other market participants could offer a tighter spread and front run me.

This is why my C limit orders would rarely fill.  If my order filled, the bid/ask would almost always move against me.

Title Trading had an outdated business model

Title’s business was largely built around the scalping strategy and their training materials encourage trainees to start off with that strategy.  However, because of the front running that was occurring, this strategy was terrible.

At that time, I believe that many franchises of Title and Swift Trade were closing down because their old bread and butter no longer worked.

My misconceptions about “flash” trading

At the time, there were other things going on in the market that was very weird to me.

On a stock such as INTC, there would often be at least 500k shares on the bid and 500k shares on the ask.  In rare occasions, the bid/ask would move up a penny on no volume.  On the time and sales, there wouldn’t be a single trade at the ask price.  Of the 500k+ shares I saw offered a second ago, none of those 500k+ shares were filled.  Orders for half a million shares disappeared.

At the time, I erroneously assumed that orders were delayed and that computerized market makers were able to get a sneak peak at incoming orders.  I thought that they were able to act on the knowledge of incoming orders before they were officially posted to the exchange.  This isn’t how it works.  I currently believe that this behaviour happens due to obscure order types.  For example, there is supposedly an order type that gets cancelled if the other order is above a certain size (see this excerpt from Michael Lewis’ Flash Boys).  Of course I could be wrong as I am not an expert in market structure.

Stocks with very large spreads

Another strange behaviour occurred in stocks with very large spreads.  If I were to place a bid at one cent higher than the current bid, suddenly there were a huge number of orders at the new bid price.  I did not see the number slowly go up.  It happened right away.  I erroneously thought that computerized market makers got a sneak peak at my incoming order and were allowed to line up ahead of me before my order was posted to the exchange/ECN.  (I currently believe that the mechanism for this is through some other means, such as orders who are pegged to the best bid/ask price.)

Retail order flow

There are many ways in which various companies skim money from retail investors.  Here are some of them:

  • Market orders.  I believe what happens is that the intermediary will manipulate the NBBO so that the market order is filled at an awful price.
  • NBBO exceptions.  For example, Etrade doesn’t obey the NBBO on the closing cross.
  • Manipulating the opening cross and profiting from market buy/sell orders on the opening cross.  Explained in this excellent WSJ article.
  • All or none orders.  These are banned in Canada.
  • Picking off limit orders on cross-listed securities or on options.  (e.g. currency arbitrage, option arbitrage)
  • Trying to repeatedly collect the bid/ask spread

Moving up the food chain

I am theorizing here, but I think that there is a class of traders that prey on HFTs.

Most HFTs out there play the game of collecting the bid/ask spread repeatedly (I’ll call this scalping).  This happens with retail orders, institutional orders routed to an investment bank’s dark pool (because the investment bank more or less wants to trade against its clients), and institutional orders routed to exchanges.  The strategy entails risk.  If a stock were to suddenly move in one direction, a market maker could be stuck with a position that they are taking huge losses on.

I think that there may be ways to specifically target people who are scalping.  For example, if you look at the time and sales, you can figure out that anybody trading in sub-penny increments is more likely than not scalping (retail and institutional investors can’t trade in sub-penny increments and rarely scalp).  Perhaps if you are smart enough, you can make educated guesses about what their position is and when they might be vulnerable.  (I haven’t tried this and I don’t know if it works.)

Another approach is to trade based on news.  You pay news wire services for faster access to their latest press releases (BusinessWire recently announced that they would no longer do this).  You design a computer program to quickly understand and process unstructured data (the press release).  If a company releases bad news, your computer algorithm will quickly take liquidity and hit the bids (sell at the bid price).  If you sell enough shares, you will move the bid/ask price and force a lot of market makers to cover their positions.  Their forced selling is a potential source of profit on top of what you make from acting quickly on news.  If you are wrong about the news, you might make a quick buck from the market makers anyways.

Risks to the current market structure

Because there are many parties playing the scalping game, there may be a lot of traders who suddenly need to exit the same positions.  As the flash crash demonstrated, traders may all try to close their positions at the same time.  There is some fear that we might see an even bigger flash crash.

Thomas Peterffy gave a very good speech on market structure in 2010 where he address this topic.The bottom lineThe lesson I took away from my day trading experience was that markets at the time were “unfair”.  Some market participants had special trading advantages over everybody else.  Of course, I don’t think that unfairness should be used as an excuse for not making money.  (Unfortunately, I was not a profitable day trader after commissions.)  Proprietary day traders were making money when the NYSE specialist structure was still around.The big picture is that most brokers and exchanges are skimming money from their customers.  Either they trade against their customers directly or they sell out their customers to other firms (Bernie Madoff and Knight Capital Group were pioneers of the latter business model).  I’m not a fan of this business model because it generates zero value for society and should rightfully attract regulatory scrutiny.  Historically, regulators have greatly reduced the amount of skimming that occurs and have reduced profitability for brokers, market makers and exchanges.  Trading costs for both retail and institutional investors have gone down dramatically over the course of history.

Advertisements

3 thoughts on “What I learned about HFT from day trading

  1. 1. Is IB the best brokerage for the retail client?
    2. What do you think about their “SMART” routing?

    I think they are the cheapest, fair and very transparent to their customers. So many features, great support and easy access to most of the world’s markets.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s