Monte Carlo Simulation of Losses

What exactly do you suggest?

Gulp. Do I have to? I voted with my feet when it came to AC, without bothering to create a model... A proper scrutiny of the "LTV"s was enough for me.

I'm not sure I understand how you are deriving these numbers. These loss estimates are far lower than I derive from my models.

In the case where the default probability D = 20%/annum and the loss given default L = 40% (so a recovery of 60%), I don't see how you obtain the result that the loss on the portfolio is so small. Unless L isn't the loss given default ... or is L how much you are tweaking the LTV (i.e on a 70% LTV loan, a 40% L, leads to only a 10% loss)?

Sorry I'm old fashioned and just tend to think in terms of PV(recovery).

If you are wanting to consider worst-case scenarios, then your loss on default numbers could be on the optimistic side. For US sub-investment grade corporates, senior secured bank loans (sub 70% LTV) recovered at around 80% over the past 20 years. Senior secured loans and bonds (sub-70% LTV) recovered at around 65% during the same period. The standard deviation for the secured bonds/loans is around 22% but that hides a fairly large skew. Anything that is mezz or junior is obviously below those numbers. Clearly, these are US, rather than UK, numbers and for corporates of a much higher quality than the typical SME (so they have a lower default probability but not necessarily a higher recovery rate).

In terms of default probabilities, the worse case scenarios are obviously found during credit crunches. During the last crisis, European single B corporates (typical high-yield bond) hit default rates of around 13-14%. I tend to think SMEs, with much weaker balance sheets and poorly diversified revenue streams, would be higher than that but it's pretty speculative how much. Something like 20-25% hardly seems unreasonable for a stress loss scenario.
                

For the purposes of stress testing:
D = 25%
L = 20%

I believe the biggest risks are market systematic ones, eg interest rate rises, sharp economic downturn etc.  Bridging loans are particularly sensitive due to refinancing risk, hence banks and the general market charging typically 2-3 times the rate on a capital repayment mortgage as the likelihood of default is much higher.  

I have tried think about a typical stress scenario, eg:

-small/medium shock results in banks and other lenders tightening credit terms on mortgages
-borrowers are unable to refinance due tightening of credit market and/or slowdown in house market prevents sale of property in required timeline leading to default
-expect some haircut to realised value on security due slowing housing market and urgent marketing - typical time from declaration of default to sale of security - 6-12 months maybe more if there are possession issues.

On reflection, D should probably be higher than 20% for a stress test scenario.

Not a statistician but is there any argument for including an input(s) for the fact that P(B)2B platforms are not 'too big to fail' and therefore unlikely to be bailed-out in a 2008 scenario repeat.

I'm particularly interested in the effect this would have on time-to-recovery.