pickles
Member of DD Central
Posts: 94
Likes: 57
|
Post by pickles on Jan 20, 2018 8:28:18 GMT
The way the FC selling engine chooses loans to sell is completely opaque, and having just sold out a five-figure portfolio I am none the wiser.
Back when the switch happened, I was 100% in property with 6+ months remaining, and heavy in some borrowers. The plan was to sell out before any hit the penultimate payment. Maximum exposure was around 4.0%. Yesterday was the first selling date. Quite a few property loans had paid early and been re-invested along with interest, so property was down to around 70% of the portfolio.
The autobid engine will only buy loans up to the 0.5% exposure limit. So surely the selling engine should work to increase diversification, or at least maintain it? Or, in the worst case it would be random? So by selling bit by bit, perhaps matching sale quantities to specific holdings, I should be able to sell up to the point that the shortest dated loans were gone and stop.
Not at all. Only the recently-acquired loans were sold at first, regardless of the amount chosen to sell. At one point my max exposure was at nearly 20%. I had to sell the whole portfolio to get the last near-dated loan sold.
The algorithm, whatever it is, wouldn't work on a normally-diversified autobid portfolio either. Trying to sell small amounts resulted in both parts of each loan being sold, instead of one part of each of a larger number of loans.
|
|
markr
Member of DD Central
Posts: 766
Likes: 426
|
Post by markr on Jan 20, 2018 15:33:36 GMT
From FC's point of view, I think this was a deliberate choice, or at least a happy accident.
Remember, as FC has grown, their intention has always been to level the playing field, to remove the advantages that time-rich, interested investors (the sort that take the time to read and contribute to forums about P2P!) have over the autobidding majority. Being able to choose which loans to sell (either directly or by being able to predict and manipulate the selling algorithm) would be the last way to game the system, so having an opaque selling system that potentially disadvantages the seller is better for FC.
|
|
benaj
Member of DD Central
N/A
Posts: 5,610
Likes: 1,739
|
Post by benaj on Jan 20, 2018 16:23:48 GMT
I don't no idea what the algorithm is exactly supposed to do for FC loan sales, but I can tell my experience here.
The selling process is very robust. It usually sells just more than the amount you need, at the same time limiting the each loan does not exceed 1% of the portfolio except when your portfolio is below £2000.
When you add more money and buy loans from PM and SM, it will increase to level of diversification to 0.5%
|
|
pickles
Member of DD Central
Posts: 94
Likes: 57
|
Post by pickles on Jan 21, 2018 8:32:23 GMT
I don't no idea what the algorithm is exactly supposed to do for FC loan sales, but I can tell my experience here. The selling process is very robust. It usually sells just more than the amount you need, at the same time limiting the each loan does not exceed 1% of the portfolio except when your portfolio is below £2000. In my case it didn't - it left the biggest loans so that my exposure went steadily upwards.
|
|
benaj
Member of DD Central
N/A
Posts: 5,610
Likes: 1,739
|
Post by benaj on Jan 23, 2018 12:20:34 GMT
In my case it didn't - it left the biggest loans so that my exposure went steadily upwards. FC selling algorithm does not sell loans with 1 repayment remaining nor other flagged loans cannot be sold on SM.
|
|
markr
Member of DD Central
Posts: 766
Likes: 426
|
Post by markr on Jan 23, 2018 17:02:29 GMT
The best approximation to the algorithm I have deduced is that it will sell the least number of loan parts necessary to get as close as possible to, but not less than, your target amount.
So, I can reliably make it sell one of my £100 parts (bought post-September on the PM), by requesting to sell £99. I can't make it choose which loan part, although if I do it, cancel the sale and do it again, it will always choose the same part. If I own two £100 parts in the same loan, then sell one of them, if I request a £99 sale again it will sell the second part of that loan, even though I now own less of it than several other loans.
If I request to sell £101, it will ignore all my £100 parts and make a total slightly larger than £101 using several pieces of pre-September shrapnel parts (£20 parts that have made repayments). This is presumably because £100+£smallest shrapnel is bigger than the amount it can make by some combination of shrapnel parts.
It seems that the algorithm takes no account of the total amounts of any one loan you hold, only the sizes of parts considered independently. So for example, suppose you had 10 parts of £100 in different loans, and 100 parts of £20 all in the same loan. If you asked to sell £999, it would likely sell the 10 £100 parts even though this is the worst case for diversification.
|
|