Buy/Sell algorithms

For the buy algorithm - if there's less than £1 available per person wanting to invest then it's allocated randomly in £1 chunks.  If it's more than £1 per person then it's as you describe.

For the sale algorithm it sorts the sellers into random order weighted by the amount they're looking to sell, so underwriters are likely to be at the front of the list but it could still be anyone.  Each buyer is then matched in turn to the sellers in the list.  5 sellers and 5 buyers then all 5 sellers will sell something.  If there's an excess of buyers then the list of sellers is randomised again the process repeated.  There's also a chunking algorithm whereby if a buyer is buying a large amount then they can be split amongst multiple sellers.

This algorithm is mid-rewrite at the moment.  The buyer algorithm will remain as is, I'm actually pretty happy with how it's working.  For sellers they will be explicitly split so at a basic level 80% of the sale will go to underwriters and 20% to retail investors, with any excesses being allocated to the other pool.  Within the underwriter pool sales will then be proportionately distributed.  Within the retail pool we'll probably use the reverse of the buy algorithm but that's not yet been set in stone.

Whilst the sale algorithm isn't working too badly at the moment it's difficult for people to understand how it works in practice and it's fairly clumsy to rebalance the probabilities of different groups selling.  This new system should be easier to explain and help ensure underwriters sell down whilst guaranteeing some measure of liquidity for retail sellers.  We can always add other balances like varying the 80/20 split over time to make sure both groups are properly looked after.

Is this the reason why some times if I sell say £100 of a loan, it shows up as up to 100 £1 units all sold at the same time stamp in my Manual Investment Account.   I don't need to know this information.  The MIA becomes very cluttered and difficult to follow.
Perhaps this could be sorted out.

Ditto.

A simple suggestion, albeit perhaps not the most intuitive: on the Available Units tab, how about indicating a negative figure if there is demand to buy the loan?  Alternatively, put this information on a separate tab (which might be clearer), labelled "Demand".  For a given loan, the "Demand" is just the sum over all lenders of the difference between target amount and amount held (including any cases where a lender has the target lower than the amount held - although in practice such cases are only present fleetingly until Auto-whatsit kicks in and sells).

The combination of Available Units and Demand gives everyone a clear view of liquidity.

I'm not sure how meaningful that would be without some gauge of the amount of free cash available to buy parts if they were offered for sale.  If I set my targets £100 above my holdings for the 60+ loans with no units available at the moment, it would look like £6+k of 'demand' existed.  But if I have only £100 of cash available then that demand figure would be rather misleading.

Because I haven't a clue which loans might become available to buy, I tend to do what I've described above, and I buy whichever loan parts come along first..  I'd guess that many other lenders operate in a similar manner.

Points noted and understood - I had considered them.

chris, if are prepared to use historical data, then rather than second-guessing a "time to sell", I'd rather simply know the average volume traded over the last 7 days - that could just be updated once per day as part of the overnight batch processing on the DB.  Then I could make my own judgement about how quickly I can sell.  This would also prevent AC being the target of complain-nicks when they are unable to sell within the time-frame indicated.