Populism on Dark Pools again? Nasdaq Innovates, and AI gets used for Trading

Are Populists back to picking on Dark Pools, now that IEX is an exchange?

Senator Warner’s letter, as reported by Reuters is potentially troubling, depending on the motivation behind it.    If the Senator is genuinely concerned with the routing practices of broker dealers, and the importance of resolving both conflicts of interest and improving investor ability to evaluate brokers, then I agree with him.  In that case, however, his time would be better spent pushing the SEC to implement enhanced routing disclosures that include execution quality statistics.   That is the most effective manner to expose conflicts of interest and provide investors with the information necessary to improve their execution quality. If, however, the Senator is playing political favorites to, once again, argue against non-exchange trading, that would be sad.  The evolution of broker owned and independent dark pools was spurred by a desire to minimize transaction costs for institutional clients.  It is one example where robust competition, as I discussed yesterday, has been a positive influence and has helped make the US equity market the envy of the world.

To be clear, a push to improve best execution in routing would be welcome, but it is vital that any disclosure or enforcement actions, such as the Senator calls for, also apply to the routers of Registered Stock Exchanges.  All the major exchange groups offer routing services and there is a substantial amount of routing from exchanges to dark pools.   IEX, in particular, markets their router aggressively and also operates as a dark pool for the majority of their trading.  Any regulatory push to examine routing should include IEX, considering that their own data, as I reported last week, shows that their approach of always (and exclusively) routing to their own pool might sometimes be against the interests of their clients.  The point here isn’t that IEX is doing anything wrong, but rather that their being “promoted” to exchange status should not lower the bar for regulatory oversight; if anything, exchange status should increase the level of best execution and routing surveillance as compared to traditional brokers.


Nasdaq pursues an innovative approach to institutional trading

Last week, the Nasdaq exchange filed a proposal for a midpoint Extended Life Order that is different from all the order types available on stock exchanges.  The proposal is to have a single order type, executable at the midpoint, that becomes eligible to trade once it has been resident at Nasdaq for ½ of a second.  The new order type can only trade against a contra-side mid-point ELO order that has also been held by Nasdaq for the full ½ of a second.   Unlike speed-bumps that apply to all orders or crumbling quote algorithms that seek to “out run” HFT firms to prevent orders from trading inside a “stale” quote, these orders are designed for a specific purpose and have no impact on the other orders handled by the exchange.  The midpoint ELO is designed exclusively for institutions that are not tick-by-tick price sensitive, and would prefer to trade against others with the same goal.   Whether or not this proposal is approved and succeeds, it demonstrates again that innovation is alive and well in the exchange space…


AI is becoming more than a buzzword for trading

JP Morgan’s new AI framework for executing trades is interesting and potentially transformative, but there are many questions not answered in the press coverage.   Notably, what “benchmark” does it use and how does it evaluate venues or methods that are new and innovative?  If it is based on a benchmark such as VWAP, for example, the machine will “learn” that creating market impact (to the detriment of their client) is the best way to beat that benchmark).   If, however, the benchmark is customized to each client and strategy, the issue becomes one of understanding what each trade is trying to achieve, which depends upon information that brokers may well be unaware of.  (As a side note, in many cases, the buy side trader often does not even know the purpose behind each trade.)  That said, I have little doubt that using machine learning techniques to parameterize algorithms based on JP Morgan’s vast amount of trading data will uncover better techniques for making routing and strategy decisions to minimize trading costs.  The proof, however, will be if their algorithm performance, compared to other algorithms, is better when a reasonable pre-trade model is used to assess the degree of difficulty of individual trades.  In addition, proper performance measurement must include metrics to measure the opportunity costs incurred when orders are not filled. (if not, the machine will learn to “game” such analysis by always trading passively, since those orders always capture spread, when they are executed.)  If such an analysis showed that the AI or machine learning approach outperforms the “traditional” statistical analysis employed by their competition, then expect most of the market to move in that direction.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s