Hacker News Reader

Comment by bachback | Hacker News Reader

The HFT debate is bound to be misunderstood... It rose to prominence last year with the book by Michael Lewis "Flash boys". A discussion between two exchange operators Bats and the new IEX: https://www.youtube.com/watch?v=RcpmHyPD_PY

This issue is not about frequency per se at all. Its about who has access to what information at what price. For instance one thing that happened was that data providers (e.g. Reuters) sold information to a group of traders, before it was available to the public.

The question is not at all whether software should be used to trade (we're not going back to pen & paper / pit trading). Rather the question is how software should be used in financial markets. HFT is not evil, it's a difficult problem to be solved.

View on HN

I think your viewpoint is fair. The aspects with hft that strike me as unfair are generally the result of bad exchange practices, such as lack of transparency in customer segmentation on their platforms. If it was all out in the open with price tags on display, I think the problems would fade away. Instead, the exchanges are running their businesses with about as much integrity as a bucket shop.

Can you identify a particular way in which an actual US exchange implements "customer segmentation on their platform" in a way which is tradable? This is not me trolling, this is a sincere question. If that actually exists I want to make a level in a trading simulation where the playing programmer would have to abuse that.

Not a US exchange, but much more straight forward example is Eurex.

They have 2 different electronic connections based on your latency/throughput requirements. One is more expensive than the other (http://www.eurexchange.com/exchange-en/technology/connecting...). If I remember correctly, the HFT version also requires more onerous legal hurdles as well.

Personally, I view this as a good thing. Why wouldn't we want segmentation in exchange technology?

One of the things that gets some of the anti-HFT types going is that it's possible to subscribe to direct exchange data feeds.

All exchanges are obliged to contribute to the SIP, which offers a market-wide view of quotes (prices) and historical trades. However, because the SIP must aggregate data feeds from all exchanges and then republish this information, it's possible to "predict the future" if you subscribe to the direct exchanges feeds (which are more-or-less what they send to the SIP) and avoid the added latency of a hop via the SIP's aggregator and redistribution.

These direct exchange feeds are expensive, and getting more so (ie. NYSE UltraBook is $11,000/month for the data, plus port fees, plus connectivity). The argument is that this segments the customers such that unfair advantage is available to those who can afford to subscribe to the direct feeds.

More recently (last two? years) it's been possible to get a microwave transmission of the data, rather than fiber, which is both lower-latency and more expensive again.

Direct feeds are available to all market participants. Expensive really sounds like an excuse. It costs what it costs, it's not clear why it should cost any less than it does.

Right. If a direct feed is too expensive, then a SIP feed being more than the end-of-day prices in the newspaper could fall foul of the same argument.

I think he is lumping dark pools in with exchanges. Technically there are only 11 registered US equity exchanges and any displayed quote has to be accessible to all market participants under nondiscriminatory terms.

Dark pools OTOH tend to have all sorts of gimmicks with respect to whose orders are allowed to match against other orders and under what circumstances. E.g. if I connect to a bank-run dark pool, I may be able to opt out of matching against principal flow from any of the bank's prop desks. The matching rules should be described in a dark pool's "form ATS" which some dark pools publish (and some don't).

Reply on news.ycombinator.com