HFT, the three letter word that garners a great deal of lip service on the networks of CNBC and Bloomberg, remains to most people, a three letter word. It stands for High Frequency Trading and has been the primary strategy for a new set of hedge funds over the past decade. They have been very successful, in fact, according to a recent (and one of the few disclosures of hedge fund performance) HFT led to $1 billion alone in profits for Citadel Investment Group. This is merely one of the very few funds running similar trading strategies, which leads one to believe the total profits, must be significantly larger. At first thought it is easy to assume that any computer based strategy can be qualified as high frequency though this is not the case; the following content aims to reveal how HFT does not qualify as legitimate quantitative strategies.
What is HFT, or better yet, what specific strategies comprise the whole industry? One of the most prominent is the liquidity rebate trade, which is based around utilizing exchange’s credit structure or in the case of the NYSE, their SLP (supplemental liquidity program). The primary concept within a credit structure – as opposed to a classic structure – is the exchange provides a rebate to traders who provide liquidity by either posting bids or offers. At the same time they charge a near identical fee, typically a few tenths of a penny per share to those who remove liquidity from their order books. How a liquidity rebate algorithm works is that it is able to spot institutional algorithms at work - trying to divide up large transactions into smaller blocks - and when it detects this, essentially steps in front. Say the institutional algo – short for algorithm – is bidding 60.00 for a few hundred shares; the rebate algo posts an order to buy 200 shares at 60.01 which due to its higher bid is filled before the institutional algo. Then immediately is posts the stock as “liquidity” again and sells those shares back to the institution at 60.01, effectively netting no profit or loss on the exchange of stock. However, at the same time they just collected a rebate of .0025 cents per share, which equates to a profit of 1 dollar for assuming essentially no risk – unless the initial buyer steps away – within less than a second. Likewise, they generated what appears to be more volume or interest that actually exists.
Another component, which is more often involved in moving prices, is the “predatory” algorithm. This version works essentially in the same method as the rebate trader, yet finishes off the deal differently. While observing the institution’s order the algo continues to raise their bid or lower their offer until the price exceeds that of the customer’s original intentions. When that happens, immediately the shares that the predatory algo had posted are not taken, the algo notices and reverses direction. If they were initially bidding, they then sell the stock short until the price falls. Now the price should fall because theoretically there is now new supply (more shares) and at the same time they were the only one that kept paying the offer. When the price falls a cent they cover their own shares and make a penny a share in the transaction along with a series of liquidity rebates. A brief example of this is perhaps that you place an order to buy 200 shares of Apple which is at 200.00 with a limit of 200.02; you might get a fill on 100 at 200.00 but for the next 50 - 200.01, and the last 50 - 200.02. A predatory algo would have collected the rebate on the first 150 shares then realizing you won’t didn’t take its offer of 200.03, sold the stock short until you bought at 200.02 along with covering its exposure at the market price. Advantages to a predatory algorithm are that it can generate additional income for a hedge fund; however at the same it takes more risk by holding a position.
Another example and probably the most disputed in current politics, is that of the automated market maker firm and or strategy. This is justified as supposedly the most beneficial liquidity provider by participants, but displays the most evident case of front running. AMM’s specialize in sending out fill or kill orders in an attempt to determine the limit of incoming order flows while at the same time optimizing price execution from the order books of ECN’s (Electronic Communication Networks), the NYSE, and NASDAQ. A way to grasp this concept is to note how they use the fill or kill: they – working against a long order – offer higher prices in ARCA or Instinet (both ECN’s) and if it is not filled immediately it is canceled. Then, progressively they step down their offer until it is accepted by the original customer. Once the AMM has determined the limit price it immediately seeks to find shares at a lower price in another ECN or exchange, and sell those shares back to the customer at their limit. The ending result is a net of few cents for being able to see order flow ahead of retail investors and normal execution. Surprisingly and the most critical aspect for AMM is the focus on speed, all of the above process happens within a few milliseconds. Also noteworthy is strategies like this are not subject to normal market maker requirements of having to post both bid and offers. Connection and execution times are the most important component for the automated market maker so to reduce times they pay to locate their servers next to the exchange.
These are the essential components of High Frequency Trading and the only problem is that in my opinion, none of these have a beneficial aspect to improving market quality or liquidity. All provide fake liquidity at the expense of the retail or institutional investor while taking advantage of current exchange initiatives. At the same time, it has become near impossible to see what the true volume of shares traded really is, but it has to be lower than quoted numbers. Likewise the visible volume is not the true volume because currently shares exchanged that are reported do not contain figures from the ECN’s, and the large amount of dark pools which now exist (another topic all together). Even worse, the NYSE enacted in 2008 their Supplemental Liquidity Program that created a credit structure for providing liquidity with no intention of reversion.
Lastly, in this current market we have seen volume decline substantially from the March lows indicating that conditions are “technically weak”. If both price and volume were increasing at the same time it could be considered “technically strong” by most participants. A look at the S&P 500 ETF, SPY, has seen average daily volume decline from 350 Million to 185 Million currently. This does not seem very promising and when you start to count in the effect of HFT, it presents a scary picture of the current market quality and state. Today, the 27th, conditions were extremely thin in both in the equities and currency markets, especially considering the day old news from Dubai. Regardless of all these issues, I find nothing wrong with computer based systems, but when those systems exist to cost the investor more money on their entry and exit something should be said.
Would you continue to shop at a store if you expected to find the item in stock, yet when going to the store they informed you the item was backordered – but, you could purchase it from the man outside for slightly bit more? Ironically he always seems to walk in ahead of you. If the answer is no, then you should begin to understand why there is little use for the HFT strategies explained here.
