Systematic Internalisation (SI) Summary & FAQ
Systematic Internalisation (SI) is a risk methodology for retail portfolios. It has comparable returns to B-Books with reduced risk (VaR).
Whilst we see over short time horizons retail flow herds, in particular in response to news events. Over a larger time horizon we see a levelling off in the amount of directional risk we see, if the underlying trades originate from an exclusive retail portfolio (where we receive all the trades).
In a typical retail portfolio 90% of it's XAUUSD trade volume is closed within 30 minutes
B-Book Risk Accrual
B-Books are making a behavioural bet that the client loses over a long term. Like a casino, the odds are in the broker's favour via spread, margin requirements.
Here is an example of risk accrual of XAUUSD at the day level granularity
On the 22nd, there is a -50M XAUUSD short position from 2,210M of XAUUSD traded. i.e. 2.3% of the trade volume wasn't internalised.
On the 21st, there is a -35M XAUUSD short position i.e. 1.75% of the trade volume wasn't internalised.
On the 20th, there is a -20M XAUUSD short position i.e. 0.9% of the trade volume wasn't internalised.
it's the accrual over time of the non-internalised element of the portfolio that results in the large B-Book positions
Let clients reduce the risk, where it is sensible, don't selectively broker flow on a trade by trade basis.
A clients position over time should revert to zero i.e. it naturally closes out. if you broker part of the flow, you fork from clients natural tendency to close out i.e. go to zero. by brokering selectively we fork our positions in a way that has no guarantee to ever return to zero, without hedging.
When we receive an open trade. We know at some future time we're going to get a close. This is information. It's not just receiving a random time series of buy/sells. Bear in mind, not all portfolios are constructed this way.
SI Summary
Systematic Internalisation (SI) uses a time based internalisation window where flow is internalised with the balance transferred to a Compass Managed book which is actively hedged via passive hedging/opportunistic/signal driven hedging.
Benefits
- Limits the NOP at LPs as much as possible
- Produces great percentage of spread PnL retention.
- Limits the amount of portfolio risk you take as a broker
- Avoids over-hedging as the internalisation window is tuned to internalise > 90% of the flow, without the month to month risk build up that can during black swan events bankrupt a broker
- Execution, LP and vendor commission costs are largely eliminated relative to STPing flow.
FAQ
Q: Am I sacrificing profitability for consistency of revenue?
Here we demonstrate outperformance of a retail B-Book strategy with a significant reduction in overall VaR of the portfolio.
So it's not necessarily the case you will sacrifice profitability as part of the strategy.
You may want to partition your overall book by counterparty/region/margin ratio etc to apply different risk management strategies according to your local requirements.
Q: What is the drawback or benefit of doing this compared to say a VaR or risk limit bounded B-Book?
VaR bounding the portfolio. Hedging when open risk hits an arbitrary limit results in a position that forks the house's position from the underlying client position, which can result in the risk being run diverging from the client's positions and result in over-hedging. The position is hedged, the clients then close the position some time later, resulting in risk ringing through zero and being double hedged.
Q: How does it compare to an STP risk model?
An STP model involves passing all risk on to other LPs. At most we see these brokers make 40% (the markup) on the Spread PnL charged clients. We are seeing SI return a PnL in the range of 80-400% of Spread PnL.
Q: How does it compare to immediately trying to clear the risk?
Overactive hedging can result in double hedging risk. In the given interval it is most likely that the client will close their risk. Hedging it early, will require the risk to be double hedged, the client "open" and client "close" get hedged separately. The window of internalisation is tuned to result in most of the risk closure being completed via clients i.e. earning spread.
Q: How do we pay rebates and rev shares with SI?
In the process of changing your risk management strategy you may also want to review the nature of how you conduct ongoing rebate and rev share deals.
For agreements where you have to issue a rebate for X% of B-Booking their risk, you can split off risk that matches that deal into it's separate rev share / rebate portfolio adding X% to the rev-share portfolio and the rest to an SI based portfolio.
Q: Why would we sit all that money in a hedge account to make like for like returns?
You probably need to sit more in an account if you do pure naked b-book, as the swings and runs of negative PnL can be bigger and longer.
From IG's public report: