Optimal execution of illiquid securities

Question

I am using an API to direct orders based on some proprietary buy/sell signal. I am trying to frame a thought process which outlines the slippage/impact risks versus execution risks given the option to utilize different order and algo types. The intent is to identify a (quasi-)optimal execution method, either by individual security or a blanket approach for the entire portfolio. Implicitly, my ultimate goal is maximizing my portfolio’s logarithmic growth rate, so I am willing to trade off some execution price for slippage risk, and vice versa. I have modeled the expected execution costs, but this only tells me that I should be willing to forego a fill if the trading costs exceed the expected return.

General Information about the strategy:

• Buys/sells illiquid securities (e.g., small and micro-cap stocks; some OTC)
• Holding periods tend to be in the months or longer

The API I am using allows for the following basic order types (there is additionally a little bit of customizability for algos):

• Limit
• Market / Market on Close / Market on Open
• max % of volume
• Relative to NBBO
• Relative to NBBO + limit
• VWAP

So far, I have ruled out market and simple limits orders. VWAPs (best efforts) seem sensible, but I am worried about their susceptibility to gaming since this is what probably what liquidity seeking algos expect. Relative orders are interesting, but a similar problem is that I would think anything at the top of the book is subject to game playing. % of volume orders seem highly susceptible to fill risk, especially for illiquid securities.

How should I begin to think about optimal execution given a choice of execution methods? What simplifying assumptions or heuristic frameworks could be useful in identifying quasi-optimal execution strategies? Is it worth investing serious time and energy into investigating algorithmic order types which seek hidden liquidity on illiquid securities?

Note: Given my trading frequency, I am not particularly interested in doing better than the midpoint of the NBBO. I am just trying to figure out how to execute at the highest rate possible without becoming scalper bait. Strict optimality conditions and dynamic stochastic control are not required.

Answer 1

Please note that my answer is primarily opinion/experience based. If it is not appropriate I will take it down or edit accordingly.

How should I begin to think about optimal execution given a choice of execution methods? What simplifying assumptions or heuristic frameworks could be useful in identifying quasi-optimal execution strategies?

I think optimal execution for you is achieving the same (or close to it) fill as your model. Though without extra details, this is hard to say. For illiquid securities, I have had lots of success over the years executing a high percentage of average daily volume by using a combination of orders. For example, you might start a time slicing algorithm (TWAP) early in the day and just set it to buy 100 shares (or some small arbitrary amount) every X minutes. You could simultaneously have a VWAP algorithm running that has a limit price associated with it. You will find the most liquidity at the closing auction--use it! Also, I wouldn't throw away a traditional limit order just because it is simple. If you can get filled with a limit at your price or better, you should use it. With a limit order, you should also be able to dictate the order size shown to market. In this way, you will not show the world how large of an order size you have. The order may not get filled, but nothing is going to go wrong by leaving a limit on the book.

With orders that are the size you mentioned in your comments above, a split between TWAP, VWAP, and MOC should be more than sufficient. I have executed orders of more than 30% of average daily volume within the last hour of trading in this way and not had any major slippage issues. Orders larger than that I tend to use a similar approach just starting the execution earlier. IB also has a newer algorithm called Adaptive as well that you can set a limit price with as well as the "urgency" of the order. I have used this to execute during the day quite often over the past few months and I really like it.

Is it worth investing serious time and energy into investigating algorithmic order types which seek hidden liquidity on illiquid securities?

I think it is worth investigating. It can't hurt. Leaving an order routed to a dark pool isn't going to cost you anything. With illiquid securities, it may not be all that fulfilling unless the pool you are routing to is the one getting the flow. You will find this out when you start executing. If you have access to several dark pools, you should be able to find a cross somewhere though it may not be worth the trouble unless you have a large order.

Furthermore, using IB, you have access to Jeffries and Credit Suisse algorithms. They are available through the API and if you are seeking hidden liquidity you might take a look into them as well.

Link to IB's algo guide: http://interactivebrokers.github.io/tws-api/algos.html#gsc.tab=0