14 Dec 2012
Can you hide in the dark?
Short Article(2 pages)
Can you hide in the Dark?
MTF dark pools are playing an increasing role in European algorithmic execution strategies.
In recent years the job of effectively executing large orders and minimizing impact costs
has become ever more complex. Lit markets have fragmented, trade sizes decreased and a
perception has grown that market impacts are being exacerbated by the strategies of heavily
‘quant’ HFT participants.
In an ideal world, from the Buy-Side’s point of view, their larger orders would be matched
with other Buy-Sides in natural crosses via an agency matching mechanism. This minimizes
information leakage and so decreases the possibility of being ‘gamed’ by other market participants.
In reality, creating venues that can achieve this type of match ‘immediately’ has proven difficult,
and is usually arrived at through a negotiation process. So if a natural block match can’t be found
quickly, larger Buy-Side orders are usually sliced, scheduled and then fed in chunks to smart order
routers that will hunt for liquidity while attempting to mask the true size of the order.
Until very recently, the algorithms used to optimize trading schedules were based on the assumption
that the sliced orders would be sent aggressively to lit markets and would impact the market price
in a predictable way. Picking the best trading schedule becomes a job of balancing the market impact
of trading too fast in lit markets versus the market risk of waiting too long.
One of the side effects of MiFID, apart from lit market fragmentation, has been the formation of a
number of new ‘MTF’ mid-point matching Dark Pools. These MTF Pools are open to both larger resting
orders and smaller aggressive ‘IOC’ flow and are therefore a potential additional source of liquidity.
Use of Dark Pools in Execution Strategies
How might the presence of these MTF Dark Pools affect the choice of optimal execution strategy?
In theory, due to the mid-point matching nature of the dark pools and the fact they are pre-trade
opaque, orders sent to - and trades executed in - dark pools should cause less (or even no) market
impact compared to orders sent to lit venues. So, any part of a trading schedule that can be
executed on a dark rather than lit venue would seem to be ‘cost free’ in the sense that the
volume traded in the dark won’t contribute to market impact. Some researchers [e.g. Schöneborn, Kratz]
have recently updated the ‘standard’ pre-trade cost / scheduling models to include the possibility
of executing on Dark Pools.
However, this is all based on the assumption that trading on a Dark Pool really does have less market
impact than trading on lit. It seems intuitive, but is it actually true? And are all Dark Pools the same?
We present a result here that tries to broadly answer that question based on analysing price movements
following executions of a similar size on lit and dark venues where we know the side of the aggressor.
Figure 1: Adverse price movement in BPS (resting orders)
Figure 2: Adverse Price Movement for different dark pools
Figure 1 shows the adverse price movement against resting orders (measured using EBBO mid-price moves)
following an execution on different lit and dark venues. Price impact for the aggressor is the inverse, of course.
Looking at the results it’s clear that EBBO price movements in the direction of the aggressor
following a trade on dark venue are considerably smaller than on lit. There is evidence of mean
reversion in both types of venues, though reversion for the dark trades is greater leaving a much
smaller lasting price impact. The impact characteristics of different dark pools varies considerably
(see Figure 2).
MTF Dark pools appear to offer an opportunity to trade with less market impact (or rest
passively with less adverse selection) than lit venues. For this reason, they should be part
of any optimal execution strategy seeking to lower impact cost. However, impact in dark pools
is not zero and different pools appear to offer quite different behaviours. As always, it’s
vital that any use of these pools is based on an analysis of what actually happens when trades
are sent to them as opposed to what theoretically should be the case. Also, of course, executions
on a dark pool isn’t guaranteed so any modelling of their use must take into account the probability
of execution.
So, whilst you can’t hide completely in the dark you can at least lower your profile.
Also published in Best Execution magazine Q4 2012 edition