Am I alone in thinking sales take too long nowadays?

I care about such things! Unfortunately this seems to have become the norm, I’ve never actually timed it but based on recent experience I would say that it takes longer than 2 hours.

I just went through major sell exercise and must say I'm impressed. I don't care about 3 or 4 hours. All loans sold withing 1 day.  

I would like a much faster turn around.   I still remember the day when I put an extra zero into the sell field by mistake and clocked okay - my hard earned investment was gone before I could do anything about it.

I wonder if the delay is not intentional (it's about 2am now and not much is happening, including my buy/sell orders).   There's a great book called Flash Boys where they deliberately introduced a trading delay to make things fair.   There are API users now.

I think it's partly do to The Algorithm being so riddled with bugs that the only way they can stop the system completely melting down is to throttle it to only be allowed to consider each loan once every few hours.

In the past, we've seen examples of faults in The Algorithm such as:
- using only 98.01% of available cash when buying a loan at a 1% discount (then the next time the loan is processed, using 98.01% of what's left, and so on, until the remaining 1.99% is so small as to be below the minimum representable currency unit).
- repeatedly buying and selling the same loan via investment accounts.

By slowing it down so much, the effects of such bugs are largely masked - several hours later the conditions will probably have changed sufficiently that the next transaction in the anomolous series is no longer possible and/or desireable.

In principle, The Algorithm should be fairly straightforward to verify - given a current database state and a loanId to process, it should either report (correctly) that no trades are possible for that loan Id, or generate a complete list of transactions that fully exhausts all possible trades (so if that transaction list is processed, and no other actions are performed that change the database state as it pertains to that loan, the resulting database state would have no trades possible in that loan Id).


Some of the visible side-effects of current faults in The Algorithm include:
- several loans at any given moment with "<£0.00" available. A single loan could occur by co-incidence on rare occasions, as the last buyer left less than a penny of the amount offered for sale. Multiple loans suggests instead that it is The Algorithm at fault.
- significant numbers of transactions of tiny fractions of a penny. Again this might occasionally happen in "genuine" circumstances, but more often indicates a problem where The Algorithm left behind some shrapnel that it has to clean up the next time it processes that loan.
- an investment accounts (including MLIA) being left with a tiny fractional amount uninvested more than very occasionally [unless this occurs by design with a later redesign of The Algorithm - such as to increase the implicit 'share price' for loans so that it is above rather than below the smallest representable currency unit].

Loan 106. I had principal credited at 16:56 on 22nd. As at 08:49 on 23rd my loan book has still not caught up. Something ain't working.

Is it purely my imagination, or was the original proposal for the shrapnelator to have a minimum unit size? A pound I seem to remember? With just one unit of each sale being less than a pound to clear the end bit of any sale. So if say £100 of a loan was offered and there were 150 buyers interested, 100 lucky buyers would be picked at random.