Michael Lewis' new book, "Flash Boys", has generated significant interest in the trading community. It's about high-speed trading. I haven't read Mr. Lewis' book (it's not out until Tuesday), but I have done something almost no one else has done: I've read the S-1 for Virtu, which is a high-speed trading firm that is slated to go public, likely in the next few weeks.
The S-1 is over 200 single-spaced pages, but it is an interesting window into the world of high-speed trading.
It has generated a lot of publicity because of this claim: "As a result of our successful real-time risk management strategy, we have had only one losing trading days ince January 1, 2008."
Only one losing day since 2008? How can that be? It's impossible, isn't it?
No, it's not. I don't claim to be an expert on high-frequency trading, but I have spent a lot of time in the last five years talking to these traders, trying to understand what kind of strategies they use and how their models work.
And it seems fairly simple: The tiniest of advantages, done millions of times, combined with payments from stock exchanges for posting bids and offers, equals consistent money making.
Let's start at the beginning.How do they make money? They are market makers: "...we generate revenue by earning small bid/ask spreads on large trading volumes."
In other words, they buy low and sell high, but they do it in very large volumes. They make markets, they say, in 10,000 listed securities on more than 210 exchanges in 30 countries.
Think about that. 210 exchanges in 30 countries.
But they spread out the bets: No single asset class or geography constituted more than 30 percent of their income. Only 27 percent of their income comes from trading U.S. equities, for example.
They attribute their success partly to this diversification: "Our diversification, together with our revenue generation strategy of earning small bid/ask spreads on large trading volumes across thousands of securities, enables us to deliver consistent Adjusted Net Trading Income under a wide range of market conditions."
This is an important point: No losing days OVERALL is different than no losing trades. They are making hundreds of thousands of trades a day (I would not be surprised if they made over a million trades a day). They have likely had many losing days in individual asset classes or markets, like Japan or Europe or bonds, but overall those losses get cancelled out by the winners.
But other people also employ macro strategies, and it's unlikely they make money with that kind of consistency.
So there's something more, and they name it: "We believe that our success in the past has largely been attributable to our technology, which has taken many years to develop."
So they've developed technology they think is better than anyone else's. And, apparently, this is very expensive, and they seem to be in a constant technological arms race:
"If technology equivalent to ours becomes more widely available for any reason, our operating results may be negatively impacted. Additionally, adoption or development of similar or more advanced technologies by our competitors may require that we devote substantial resources to the development of more advanced technology to remain competitive."
Aside from the money spent on technology, they are big on keeping costs down. They are self-clearing, so they don't pay clearing fees. Because they trade such large volumes, they get favorable pricing for trade processing. And they do it with very few people: 151, to be exact.
So let's review what they say:
What they don't say is exactly what their winning percentages are, only that they win on aggregate almost every day.
But let's say they have a one percent edge--51 percent of trades they make money, 49 percent they lose.
Ask any mathematician, and they will tell you that the law of large numbers comes into play. If you do one trade a day, you're not going to make money consistently with those odds.
But if you do 100,000 trades a day, you are going to win almost every day, on aggregate, even with a small 51 percent advantage. It's the law of large numbers.
A few other things stick out:
First: On most days they make between $1 and $2 million. What this tells me is that they do not have a lot at risk at any time in the market. In fact, their core skill seems to be that they are very efficient at managing risk.
That's not even speculation. They said that, here: "Our market making strategies are designed to put minimal capital at risk at any given time by limiting the notional size of our positions."
They do this by hedging: "Our strategies are also designed to lock in returns through precise and nearly instantaneous hedging, as we seek to eliminate the price risk in any positions held."
What's that mean? At any one moment, they might be long, say, $60 million, and they might be short, say, $58 million. But the value at risk at any one moment is only $2 million.
This is consistent with what other traders who engage in high-speed trading have told me. They trade often, but the total amount of money in at any one time is relatively small.
Remember, most of these trades are in 100 or 200 share lots. We are not talking about big block trades.
When you are trading in such small quantities, even if you are turning over the inventory quickly, it's much easier to avoid getting caught with large amounts of risk.
It's true you can get slammed unexpectedly if the market suddenly heads south; then you can take losses. But most of the time If things head south, you can readjust very quickly.
The criticism of these kinds of traders has been that they are not real market markets. That anyone who sets up their risk profile to get out of the market immediately if things go south is not a true market maker. There's some truth to that, I think.
Do they provide liquidity? Well, they are on the bid and offer constantly, so they are certainly there, even if it is only for 100 shares at any one time. But it can vanish fast.
So Virtu is less about taking big risks and trying to lay it off, and more about:
Let's go back to this emphasis they place on their technology. Why does it give them an advantage? Their system, they say, "simultaneously searches for the best possible combination of prices available at the time an order is placed and immediately seeks to execute that order electronically or send it where the order has the highest probability of execution at the best price."
That's old-fashioned buy low and sell high, if it is done as described. If it's because they areusing their faster system to buy one security (like S&P futures) low and selling higher a split second later, that is perfectly legitimate. If they are doing an arbitrage...say, trading gold futures against the gold ETFs (a fairly common strategy) that is perfectly legitimate.
But there have been other allegations against high-speed trading that are not addressed in the S-1, that Michael Lewis is likely to discuss.
Here are the core allegations made against some of these high-speed traders:
1) that these firms use speed to detect when a large order is coming in, and either flee or stay and take it, and that this is a form of front running;
2) that some high-speed firms may be sending huge amounts of "phony" bids and offers to try to detect and "trigger" other orders, and that this is abusive and manipulative behavior.
3) that exchanges may be improperly selling direct feeds to some traders that allow them to "see" the tape a fraction of a second before the data is sent out through the "public" feed.
I have a simple answer to these allegations: if any firm is found to be engaging in abusive or manipulative market practices, as defined by the SEC, than they should be fined or arrested.
As a general rule, I also believe that exchanges should provide direct feeds that are provided to private clients at the same time that it is provided to "public" feeds.
What else to do? A number of suggestions have been made:
1) that a "speed bump" should be initiated to slow down trading. Some trading platforms already do this. We recently had Keith Ross from PDQ on our air. Keith's system allows traders to name what speed they want to allow.
Another automated trading system, IEX, mandates a 350-microsecond delay between requests to trade and executions. They went public in October, 2013, and are now trading 45 million shares. That's a drop in the bucket compared to the roughly 7 billion shares that trade each day, but it's a start.
They would like to become an exchange, but one with some different qualities. Besides a "speed bump", There is no "maker/taker": they will not pay to have traders post bids or offers. There is no payment to broker/dealers for order flow.
I support this idea, but I do not support making it a rule for everyone. If one trading system or exchange wants to offer colocation, great. Others may not want to.
2) that traders who put in huge amounts of bids and offers should pay for the cost by paying some kind of excessive messaging charge. I think this idea is worth discussing. The trading community is required to buy more pipes to handle the excess traffic, so charging some kind of fee does not seem crazy.
3) that the market has become too fragmented (13 exchanges, 40 dark pools) and that we need to lower the number of trading venues. Many have argued we should eliminate "maker/taker" whereby exchanges pay traders rebates to provide liquidity. Eliminating that would dramatically cut the number of trading venues. But others feel that a wide choice is important and that it is the technology that should improve.
4) that stock exchanges should be barred from providing "co-location" services to traders, which allows them to locate their servers next to the stock exchange servers for near-instantaneous access to quotes.
As a general principal, I believe exchanges should provide equal access to everyone. The truth is, almost everyone is co-located, including the discount brokers. Including Fidelity and Vanguard.
There are other issues, but the general principal that abusive or manipulative market practices should not be tolerated should be the guiding concept.
But eliminate trading with computers? You're not serious. Computerized trading lowers the cost of capital. Trading is more efficient. We are NOT going back to the 1970s. We are NOT going to have a bunch of guys trading on telephones. We are NOT going back to specialists who have access to information no one else has, or to sell-side desks that get the first call. We are NOT going back to trading in sixteenths of a dollar.
A friend of mine said that the idea that a trader could keep up with a machine went out the door as soon as we went to pennies and as soon as we went below 500 milliseconds to execute a trade.
That happened more than a decade ago.
—By CNBC's Bob Pisani