Not exactly, a firm like Virtu makes money due to volume and volatility. If the market goes up, they make money, if it goes down, they make money. They don't make money when they break things, or when the volume (and volatility) is low.

Source: Worked at Madison Tyler / Virtu for over 4 years, but left before their IPO.

They make money NOW either direction; this doesn't mean they will always make money. If a competitor suddenly starts being faster, they are going to start losing money no matter which direction the market goes.

Note that I'm not saying that market maker profit is correlated to the performance of the wider stock market, just pointing out that market maker performance can and will be affected by market-related factors that can cause all their market bets to go wrong at the same time, so the actual risk of losing money will always be much greater than what just naively applying the LLN says.

Here's a couple of plausible factors:

1. A flash crash is just a huge coin flip -- even if your market making firm has circuit breakers that stop it from making more trades (and successfully kept its position small), if the exchange ends up ripping up trades en masse you may suddenly have a large position that loses you a lot of money

2. Someone screws up an algo [1] and you burn through more than your entire market cap in half an hour

[1] https://www.bloomberg.com/news/articles/2012-08-02/knight-sh...

Not sure I understand why we can assume that its all trades are entirely "independent" here.

Agree with you that HFT performance is not correlated with market direction. But given the volume of trades it would imply that many of the trades are occurring on a smaller pool of equity instruments. I'd think that given liquidity constraints and competition that there's a sweet spot in terms of number of stocks that a given strategy is actually efficient on.

So not entirely sure why we consider every trade to be iid - rapid shocks (flash crash), equity specific news etc would affect a number of trades at once changing the 51% probability?

Also does anyone know if the profitability numbers include payments to be a market-maker?

Market makers typically have a requirement to be in the market a certain (high) percentage of the time. In return, they get rebates (which can be a significant source of revenue) and protections from the exchange against overfilling. This can help them for certain shock events.

Most likely, though, their profitability % is tuned by edge and risk parameters. They can adjust size, widen out, or tighten up the spread to tune trade frequency and profitability. They are finding a sweet spot between # of trades and profitability.

Those number absolutely include payments. It's rare (in my experience) for an equity market making strategy to be positive in gross terms (prior to the maker-taker rebate). These numbers you hear from Virtu/KCG should be net of exchange trading costs/rebates.

My friend at a different hit firm says they tend to make MORE profit in high volatility regimes

That's exactly what SEJeff is saying (profit is positively correlated with both trading volume and price volatility).

Correct sir! And I've spent the past 9 years of my career as a Linux monkey in electronic trading / HFT / whatever the hip term for it is today.

Yeah, the analysis in this article is not really answering the question of WHY they are profitable every day; in reality, it is just explaining the law of large numbers.

The whole thing rests on the premise that the company makes a profit on 51% of the trades, a loss on 24%, and break even on 25%. It is easy to show from those assumptions that you will never have a losing day if you make enough trades.

The REAL interesting question is "How do they make money on 51% of the trades they make, and are those results sustainable?"

Is this question as interesting as it might seem to someone unfamiliar with trading? Electronic trading firms do more than make markets, but stipulate that market-making is the sort of baseline approach they take to making money. Are you really asking "how do market makers make money on 51% of their trades"? Because that question is easy to answer; it's the premise of a market maker, whether done by a computer or by a person in a pit.

I probably don't know enough about market making and trading to really say whether this question is interesting to people who are.

My thinking, and why I found it interesting, is that the connotation behind showing that the company 'never has a losing day' because of the law of large numbers is that the company is taking practically no risk. I was trying to ask a question about that point; do these companies really have no risk? Are they basically printing money?

I am sure the answer is no, because no company has zero risk. I want to know where the risk is, and I figured the answer to that question is answered by answering the 'how do they make money on 51% of their trades' question.

That might not be the right question, but I don't think my question was equivalent to asking 'how do MARKET MAKERS make money on every trade', because traditional market makers (i.e. the stock exchanges) are not the same as HFTs. A stock exchange has inertia on its side to protect their profits; a competitor has to fight against the network effect to take them on. An HFT has none of that protection.

HFT firms make money by having sufficiently more accurate predictions of the future price than those they trade against -- here, for prediction, you can take any reasonable metric, such as the change in price in the next 10 seconds. Note that this can be done by simply being faster, as having the same prediction in 400ns that took everyone else 2000ns to discover is still having better predictions.

HFT firms typically have anywhere from little to very, very little trading risk, so long as the above conditions hold. Their actual risks are operational (How Knight blew up) or the above situation no longer holding (How Getco, Teza, etc. died).

Also, one thing I have to mention in your post -- stock exchanges were not the market makers of the past. That role was played by someone called a specialist; you can think of them as doing exactly what an HFT market maker does these days, except far, far dumber and far, far slower. Those people were typically (always?) not employed by the exchange. An exchange is simply a centralized place to display orders, not a counterparty that you trade against.

They're minimizing their market risk, but the trick is doing that while making payroll. You can trade practically risklessly by quoting an unrealistically wide spread. Practically no one will trade with you, but none of your orders will sink the company.

Correlations will tend towards 1 when using a large-enough time frame. On a microstructure scale, correlations go haywire in liquidity distress.