Last week we announced our CHARM product with a press release titled, Swap Margin Check in 10 milliseconds.
In this article I will provide the detail behind our thinking.
Mandatory Clearing of Swaps in the US, means that CCPs have to perform a Clearing Acceptance Check, which requires first checking that the executed trade submitted to clearing meets the product eligibility criteria e.g. sub-type, currency, maturity.
Next practice varies amongst CCPs, so either
- The trade is risk-checked in isolation against limits and accepted,
- Or the trade is queued for the next intra-day margin run, at which time the new margin required is established,
- Or the trade is risk-checked incrementally with all existing trades in the same account and the new margin determined.
In the past CFTC Regulations gave CCPs 60 seconds to complete their acceptance process. Recent Data received by the CFTC, showed that CCPs are accepting 93% of trades within 3 seconds and 99% within 10 seconds.
Subsequent to this CFTC established that for trades executed on a SEF, the time of acceptance or rejection at a CCP should be within 10 seconds.
Participants in a SEF, whether it offers an RFQ or Order Book model, need to demonstrate Clearing Certainty, either by Self-Clearing (Dealers) or via their FCM (Clients).
This requirement for near instantaneous acceptance or rejection of an order means that Clients are relying on their FCM. While for Futures trading FCMs have provided this certainty without stringent intra-day risk checks, the nature of the Swaps market (low frequency, large size) means that it would not be sensible to do the same for Swaps.
This means that SEFs need a very fast limit check at an order level, based on information set by the FCMs.
However each SEF does not want to connect directly to each FCM and each FCM does not want to connect to each SEF. In addition Clients do not want their FCM to decide how much risk limit to allocate for them to each SEF.
So in step the Hubs.
The major ones being Traiana Creditlink and MARKITSERV Credit Centre.
These connect the SEFs and the FCMs in order to provide Clearing Certainty.
And allow the management of Limits and these limits to be carved out to the SEFs.
And am sure much more, for details of which please consult their websites.
For SEFs using RFQ, the preferred approach is for the SEF to ping (ask, send message to, …) the FCM, via the Hub, to determine if the client is good for the proposed trade. The FCM then needs to reply to this request with a yes or no, and if accepted this token is then used at the Clearing Acceptance stage to signify that the trade has already been accepted by the FCM.
The preferred risk measure in this model, is an Initial Margin (IM) Limit.
As IM is not additive, if we need to offer margin benefit for risk reducing trades, then we need to calculate the new margin requirement of a portfolio with the proposed trade included. This is a computationally expensive calculation.
For SEFs using Order Books, the preferred approach is for these SEFs to locally monitor limits and to do so based on limits that have been pushed (sent) to them from FCMs, via a Hub. Such monitoring is performed on risk measures such as DV01 or Delta by tenor, both of which are additive.
The problem for FCMs, is that there is no direct way to get from an Initial Margin Limit (the preferred risk measure) to a DV01 or Delta. In-fact it requires an iterative solver as well as a number of assumptions, which may well be found wanting in times of market stress.
So in practice, FCMs can create separate DV01 limits and use just these for intra-day risk monitoring and acceptance. The drawback being that such limits may not effectively utilise client collateral held, creating situations where a trade would be rejected even though in IM terms it would have been accepted.
Which brings us to Charm.
FCMs need to respond very quickly to an ping originating from an RFQ and to do so by calculating the impact on the whole portfolio Initial Margin, checking there is sufficient Collateral Held in the Client Account and returning a yes or no.
FCMs need to respond very very quickly to a ping request originating from an Order Book, with the same calculation.
And that is the essence of what CHARM does.
By doing this calculation in 10 milliseconds on commodity hardware, we make possible a ping response to an RFQ or an Order.
And there is more.
That tricky problem of IM and Delta by tenor? Well CHARM also has a solution for that. However that is a blog for another day.
Which leaves just the product name.
Given that our current products are SDRView and SDRFix.
Well we could not decide whether we should use SEF or CCP in the title or for that matter FCM or DCM or GCM?
And what action verb to add to it, Check, Calc, View?
None of the possible permutations sounded right.
So in the end we decided to take a different tack entirely and one of my colleagues came up with CHARM.
There are a number of possible interpretations of C-H-A-R-M.
My personal favourite is Clarus Hyper Analytics Risk Management.
However all that matters is that it is memorable, pronounceable and distinct.
Try this test.
Close eyes, repeat 10 times, Charm, Charm, …
Open eyes, repeat 10 times, Charm, Charm, …
You cannot forget it then.
Well I certainly cannot.
Thanks you for reading to the end of Push, Ping and Charm.