What I don't understand about HFT is why its efficiency does not seem to be limited by transaction fees. I've tried a bit in the past to play with some algorithms that buy and sell (on paper) and it seems like the faster you go, the more you need the market to move in order to scrape a few pennies of profit, because the movement has to be sufficient to cover the spread, the transaction fees, _and_ provide some profit. Of course with a simple example like "buy at $20, sell at $20.10", it seems simple enough to extract some profit, but in fact it would need to be enough profit that it covers the transaction fees, which seem to be a kind "friction" that limits the smallest change in price that is useful. From what I can tell this kind of movement just doesn't generally occur in the sub-millisecond time frame, so how does it work? Or is it just that they are trading so much $$ in one shot that a very tiny change in price covers the transaction fee without issue? (Meaning they are taking very large risks..)
They don't pay "transaction fees" in the way you think about them. The way it generally works is you pay when you "remove liquidity" from the book and you get paid when you "add liquidity" to the book.
What does that mean? Suppose the bid/ask is 20.00/20.10 as in the example given. The order book will show a bunch of people willing to pay 20.00, 19.99, 19.98, etc, and a bunch of people willing to sell for 20.10, 20.11, 20.12, and so on. All of those people have added "liquidity" to the book by making orders which have not yet been placed. The idea is that a huge buyer could come in and try to buy 1 million shares at 20.25, and that order could get immediately executed because of the backlog of non-executed orders.
If you bid 20.10 (i.e. offer to buy the stock at 20.10) the seller at the top of the book will have his order executed (with you) and you will have removed a tiny bit of liquidity from the book. On the other hand, if you bid 19.95, your order just gets added to the system, and you have added a bit of liquidity.
The exchanges will charge you slightly more for removing liquidity (say .05 / share) than they pay you for adding liquidity (say .04 /share) and so make a spread. However, high frequency strategies can be clever about the way they add and remove orders to the book in order to minimize their net transaction costs, and so they end up being rather low.
On the other hand, when a retail investor purchases a stock through Schwab or something, Schwab (or in many cases some other larger bank) is dealing with exchange fees directly, and simply charging their customer a flat commission on top of your trade.
Thank you, I haven't heard that explanation before, it makes a lot of sense. So the normal transaction fee that you or I would pay is an additional fee by the broker, rather than a direct market fee?
The sub-millisecond timeframe is part of a race to be at the top of the book or a race to pull orders. Very few HFT's literally trade a single security repeatedly on a sub-millisecond timeframe.
The general strategy is "our signal changed - move fast, get to the top of the book in milliseconds". Then you wait, and seconds to minutes later someone will fill your order sitting at the top of the book. Then you play the same game on the opposite side to unload your position.
The game of shaving milliseconds is solely about beating other HFT's to the punch, it's not about actual market movements.
Thanks for the article. It seems to me like a missing piece is a discussion of volatility. That is, I get it that HFT drives spreads lower, but spreads are the wrong direction to make money, even in volume. That is, if pre-HFT spread is that the best buy price is $10.00 and best sell price is $10.05, and post-HFT best buy = $10.03 and best sell = $10.04, you still lose a penny on every trade instead of a nickel -- you can't make that up on volume! :-)
It seems to me that volatility is required for any market maker to make money, and that's a fundamental conflict, isn't it? That is, "outside" traders would prefer the market to be smooth, whereas "inside" traders want it to fluctuate.
I can see the argument that, in actuality, HFT on the whole needs less fluctuation to extract enough profit to provide liquidity, so in theory, it would be expected to be a more stable market maker than human operators. Is that basically equivalent to what you're saying in the article? (And is there any data on that hypothesis?)
The transaction fees large players pay are negotiated to nothing. The exchanges want liquidity, especially in the derivatives markets. HFT does provide a ton of that, so they can negotiate away the transaction fees (in exchange to agreeing to maintain some minimum volume.
Yep. Most large market makers enter into obligations with exchanges to make markets in a wide spectrum of funds. Even those which are not that profitable. In exchange, the market maker gets much lower fees.
Minor point, but actual market makers are required by law to make markets. In fact, the market makers basically are the exchange.
This is different from hedge funds which act as market makers by earning rebates by providing liquidity. These guys (and gals) are have no regulatory requirements to make markets.
Transaction fees for most brokers are pretty high. Even "discount" brokers like Scottrade charge $7 for every stock trade.
There are other brokers that charge $0.01/share traded or less.[1][2] A $0.10 spread becomes easy profit if the buy/sell only costs $0.02.
The linked broker below requires you to be an "expert" before you can sign up -- you have to have made at least 100 trades -- but otherwise just about anyone can sign up.
I have NOT tried to use them for day trading (or at all, for that matter), so please don't consider this a full recommendation. Just trying to show that there ARE other options that are cheaper.
Yes, their trades are for much bigger amounts, and they also don't pay the hefty brokerage fees that casual investors like us pay. One example, http://en.wikipedia.org/wiki/Direct_access_brokers which cost around $0.004 per share. And maybe HFTers don't even pay that much.
I've also looked at this before, and with transaction fees of 3% of the total traded, I can't see how they make money. Unless they have a deal with brokers for lower transaction fees or something.
The OP provides a good technical explanation. Unfortunately he fails to mention another closely related term to high frequency which is what makes this type of trading highly lucrative: 'Front running'. See http://en.wikipedia.org/wiki/Front_running .
Front running is illegal, but if you look for successful cases of high frequency trading they are generally tied-to/accused-of front running. And as you might imagine, in order to do front running you need to be high-up on the food chain (i.e. be a market maker)
The missing key about front running in the article is the 'anonymous' bid-ask: "The matching engine takes his order and displays it (anonymized) to all other traders with a data feed." and "She places her orders, and it is again displayed to the world (anonymously) and stored.".
If you have forehand knowledge of the bid-ask (i.e. non-anonymous) the market maker can front-run and with high-frequency make a considerable profit.
Frontrunners have knowledge of specific trades that are about to occur. You frontrun a specific block trade. The definition you imply here suggests that all market makers are "frontrunners".
HFT is not front running. Search comments for my remarks on latency arb - that may seem like front running, but it's entirely different. Latency arb involves information that is publicly available. Front running does not.
My understanding is that front running has been around as long as there have been market makers - and there have been market makers as long as there have been markets.
The practice of placing a large order and canceling milliseconds later to induce an exploitable price movement accomplishes this, sanctimonious protest to the contrary notwithstanding.
