|
|
Post by chris on Feb 25, 2018 11:50:11 GMT
happy / vaelin - further changes made, more rebalancing happening. It wasn't correctly calculating where lenders had zero holding in a given loan which was throwing everything off when other "optimisations" then kicked reducing the number of accounts being rebalanced and skewing the totals. That said in the GBBA 1 there are three loans with more than 6% average holdings (#227 at 14%, #544 at 13%, and #495 at 10%). GBBA 2 is has just under 10% in #441 and 4.6% in #550. PSA has 8% in #532 and #414 with a few others above 5%. So all lenders, within the other constraints like having at least £100 in the account, should be trending toward those levels at least until the account as a whole diversifies more between new stock coming in and the next improvement to the algorithm in 8 - 10 weeks.
|
|
|
|
Post by vaelin on Feb 25, 2018 11:55:57 GMT
How is it that some of us have 10% or more in #441 whereas others have 1.8% in the same loan? I thought the algorithm aimed to move us all towards the account average within a tolerance of 0.5% above and 0.1% below? It is my understanding that the algorithm works by directly exchanging loan parts between 2 accounts and therefore loan parts have to be able to move in both directions for the exchange to take place. So to take #441 as an example where you have 10% and I have 2.3%. If we assume that you are overweight and I am underweight in this loan, then for the exchange to take place there has to be a loan where I am overweight and you are underweight. My biggest loan is suspended so that doesn't help. My next biggest loan is #544 where I have 9.6%. However, if you are also overweight in this loan, this also doesn't help. I could list my other large loans, but hopefully you can see that it is not impossible that there are no loans where I am overweight and you are underweight where the difference is greater than £5, and therefore the exchange of #441 can't happen. Where I am uncertain is whether the algorithm would allow the exchange to happen if I was more overweight than you in #544, but I am going to assume that this is not the case to keep the algorithm simple. It seems unlikely that we would reach an impasse like that with the accounts currently so far from equal distribution. In our N=2 example where you selected only two loans in the account, I can see that we could have immediately swapped my #441 for your #544, because I am underweight #544 at just 0.66%. Either there is still a lot of opportunity left for exchange, or the single example you picked is only a tiny handful of opportunities left. Obviously the former is more likely. Edit: Cross posted with Chris above, which explains the unexpected behaviour. Will review when the algo has had a chance to work its magic. The quoted averages for the top two loans in the GBBA2 does seem to be in line with my current holdings though - perhaps my account is already as diversified as it can get for the time being.
|
|
happy
Member of DD Central
Posts: 397 
Likes: 497
|
Post by happy on Feb 25, 2018 12:21:39 GMT
happy / vaelin - further changes made, more rebalancing happening. It wasn't correctly calculating where lenders had zero holding in a given loan which was throwing everything off when other "optimisations" then kicked reducing the number of accounts being rebalanced and skewing the totals. That said in the GBBA 1 there are three loans with more than 6% average holdings (#227 at 14%, #544 at 13%, and #495 at 10%). GBBA 2 is has just under 10% in #441 and 4.6% in #550. PSA has 8% in #532 and #414 with a few others above 5%. So all lenders, within the other constraints like having at least £100 in the account, should be trending toward those levels at least until the account as a whole diversifies more between new stock coming in and the next improvement to the algorithm in 8 - 10 weeks. Thanks chris, that makes more sense and hopefully your algo tweaks will move things further along the path. It seems the lumpiness of the GBBA1 and GEA will more difficult nuts to crack as no new loans are likely to be available any time soon (never for the GBBA1) and we will have to see if the phase2 changes provide more improvements. I think the real results will however be seen more in the newer GBBA2/PSA accounts with a larger loan pool and supply of new loans. We should all perhaps be a bit cautious of expecting diversification perfection in GBBA1/GEA as IMHO it is unlikely to be able to achieve this.
|
|
loadsahope
Member of DD Central
Posts: 84
Likes: 45
|
Post by loadsahope on Feb 25, 2018 13:08:10 GMT
happy / vaelin - further changes made, more rebalancing happening. It wasn't correctly calculating where lenders had zero holding in a given loan which was throwing everything off when other "optimisations" then kicked reducing the number of accounts being rebalanced and skewing the totals. That said in the GBBA 1 there are three loans with more than 6% average holdings (#227 at 14%, #544 at 13%, and #495 at 10%). GBBA 2 is has just under 10% in #441 and 4.6% in #550. PSA has 8% in #532 and #414 with a few others above 5%. So all lenders, within the other constraints like having at least £100 in the account, should be trending toward those levels at least until the account as a whole diversifies more between new stock coming in and the next improvement to the algorithm in 8 - 10 weeks. Here are my numbers for GBBA1 544 16.03% 495 12.77% 388 6.59% 336 5.99% 295 5.76% 542 5.53% 389 5.13% 304 4.95% 381 4.62% ... 227 0.22% (42nd largest holding) Still a way to go...
|
|
mikes1531
Member of DD Central
Posts: 6,453
Likes: 2,320
|
Post by mikes1531 on Feb 25, 2018 13:29:53 GMT
I expect that a lot of investors will have been avoiding these accounts in the past because of the perception of inadequate diversification. (I know I was.) Once people realise what the implementation of this feature has accomplished, they probably will be a lot more willing to use them, and that should mean more money added to the accounts available for investment, and that probably ought to mean more opportunities for further diversification for the accounts overall and for individual investors. holdings.
I suspect that AC's most difficult task right now is ensuring investors know the feature is now working and what it means for those hesitant investors. This forum would serve that purpose, but the proportion of AC investors who follow it is bound to be tiny so AC need to work out how they can reach all the others.
|
|
amphoria
Member of DD Central
Posts: 156
Likes: 124
|
Post by amphoria on Feb 25, 2018 14:16:57 GMT
So much for my carefully contrived example above, it turns that I am underweight in #544! Of my 4 largest loans, 2 were overweight and 2 were underweight prior to the re-balancing. One of the overweight loans #227 is suspended and #388 has since exchanged loan parts so that it is either at or closer to the target weight. #544 and #495 are still underweight against the targets that Chris gave above.
I see that there are still some smaller exchanges going through after Chris' latest tweak to the algorithm, but none going to #544 and #495 so far.
|
|
cb25
Posts: 3,528
Likes: 2,668
|
Post by cb25 on Feb 25, 2018 14:21:36 GMT
I expect that a lot of investors will have been avoiding these accounts in the past because of the perception of inadequate diversification. (I know I was.) Once people realise what the implementation of this feature has accomplished, they probably will be a lot more willing to use them, and that should mean more money added to the accounts available for investment, and that probably ought to mean more opportunities for further diversification for the accounts overall and for individual investors. holdings. I suspect that AC's most difficult task right now is ensuring investors know the feature is now working and what it means for those hesitant investors. This forum would serve that purpose, but the proportion of AC investors who follow it is bound to be tiny so AC need to work out how they can reach all the others. I'll wait until I see a majority of people on this forum being happy with the results, considering i) allocation percentages seem to have gone in the wrong direction for some ii) routines don't seem to have been fully tested before implementation, requiring amendment soon afterwards iii) more changes to come in 8-10 weeks (if these are needed/desirable, suggests there's still something wrong with the current algorithm)
|
|
ceejay
Posts: 975
Likes: 1,149
|
Post by ceejay on Feb 25, 2018 14:47:51 GMT
I'd been staying away from GBBA etc, one major reason being the diversification issues much discussed here. However, with this announcement I thought I'd chuck a bit in to GBBA2 to see what happened, and I'm happy with what I've seen, with my cash spread over more than 30 loans so far. Yes, the biggest chunks are a bit bigger than I'd ideally like, but I'm happy enough.
That said, I don't think I will be making any wholesale changes to my strategy for AC, which is to focus on MLA lending as it means I can aim for higher returns and also avoid some categories of loan which I'd much rather not be in. Still, as I can't find enough loans that pass all my tests, it's good to have the option of parking a modest chunk in GBBA2 as a higher-return variant on using QAA/30DAA.
The main thing that will hold me back from jumping deeper into the GBBA2 pool is the number of longer-term loans (>3 years) which - in the current climate - is not something I want to be too heavily involved in, although a broad diversification certainly lessens the risk.
|
|
|
|
Post by stuartassetzcapital on Feb 25, 2018 14:53:02 GMT
I'd been staying away from GBBA etc, one major reason being the diversification issues much discussed here. However, with this announcement I thought I'd chuck a bit in to GBBA2 to see what happened, and I'm happy with what I've seen, with my cash spread over more than 30 loans so far. Yes, the biggest chunks are a bit bigger than I'd ideally like, but I'm happy enough. That said, I don't think I will be making any wholesale changes to my strategy for AC, which is to focus on MLA lending as it means I can aim for higher returns and also avoid some categories of loan which I'd much rather not be in. Still, as I can't find enough loans that pass all my tests, it's good to have the option of parking a modest chunk in GBBA2 as a higher-return variant on using QAA/30DAA. The main thing that will hold me back from jumping deeper into the GBBA2 pool is the number of longer-term loans (>3 years) which - in the current climate - is not something I want to be too heavily involved in, although a broad diversification certainly lessens the risk. Shorter term loans are one of two things - loan type of bridging and development OR an account designed to seek to create that type of exposure in other ways but still likely to be quite focussed on those type of short term loans on the whole. So a one year Access Account OR a developmentment and bridging investment account could suit ?
|
|
|
|
Post by chris on Feb 25, 2018 14:55:49 GMT
I expect that a lot of investors will have been avoiding these accounts in the past because of the perception of inadequate diversification. (I know I was.) Once people realise what the implementation of this feature has accomplished, they probably will be a lot more willing to use them, and that should mean more money added to the accounts available for investment, and that probably ought to mean more opportunities for further diversification for the accounts overall and for individual investors. holdings. I suspect that AC's most difficult task right now is ensuring investors know the feature is now working and what it means for those hesitant investors. This forum would serve that purpose, but the proportion of AC investors who follow it is bound to be tiny so AC need to work out how they can reach all the others. I'll wait until I see a majority of people on this forum being happy with the results, considering i) allocation percentages seem to have gone in the wrong direction for some ii) routines don't seem to have been fully tested before implementation, requiring amendment soon afterwards iii) more changes to come in 8-10 weeks (if these are needed/desirable, suggests there's still something wrong with the current algorithm) i) They're heading towards the average allocation for all, which necessarily means some lenders are going to see their allocations increasing. ii) There will always be edge case scenarios in any complex system. No other platform has, as far as I'm aware, implemented a similar diversification algorithm. Once you're invested those are your loan units whereas we're now actively rediversifying your account all the time. I'm also not aware of other platforms offering direct access to their CTO like this so where they've tweaked and altered systems they've done so silently without lenders noticing. Can anyone explain precisely how the loan matching works on RS or Z for example? I've made one main bug fix change this morning and tweaked a calculation a day or so ago if memory serves, all other changes detailed here have been changes to the parameters of the algorithm, such as reducing the minimum adjustment from £10 down to £5 and now £1. As we understand real world usage, which isn't something you can reliably test against, we'll make further tweaks and adjustments. iii) As I've detailed there are two algorithms at play. One to deploy new funds (and withdraw funds when requested); and one to diversify existing holdings. That second algorithm is the one that has been subject of this major upgrade, the former is the one being upgraded in 8 - 10 weeks. Upgrades are part and parcel of all products. You don't write off the existing iPhone as having something wrong with it just because it's going to be upgraded further down the road. I don't consider any part of the system 100% finished, there will always be new things we want to do, new ideas to incorporate, adjustments to the way systems work as the platform matures and the demands placed upon it change over time. The whole platform will be continually evolving and the upgrade in a few weeks time is just part of that process.
|
|
loadsahope
Member of DD Central
Posts: 84
Likes: 45
|
Post by loadsahope on Feb 25, 2018 15:29:57 GMT
I expect that a lot of investors will have been avoiding these accounts in the past because of the perception of inadequate diversification. (I know I was.) Once people realise what the implementation of this feature has accomplished, they probably will be a lot more willing to use them, and that should mean more money added to the accounts available for investment, and that probably ought to mean more opportunities for further diversification for the accounts overall and for individual investors. holdings. I suspect that AC's most difficult task right now is ensuring investors know the feature is now working and what it means for those hesitant investors. This forum would serve that purpose, but the proportion of AC investors who follow it is bound to be tiny so AC need to work out how they can reach all the others. I'm definitely more interested now. When I initially staretd I think I got lucky with GBBA1, and didn't pick up too large a percentage of anything (and def not anything I was too worried about). The property linked percentages were dreadful. I immediately sold out of property, stayed in the GBBA but stopped putting any more in, and mostly focussed on manual. But now I'm definitely thinking of putting some into GBBA2.
|
|
loadsahope
Member of DD Central
Posts: 84
Likes: 45
|
Post by loadsahope on Feb 25, 2018 15:33:13 GMT
I'm also not aware of other platforms offering direct access to their CTO like this so where they've tweaked and altered systems they've done so silently without lenders noticing. And it's much appreciated here
|
|
|
|
Post by vaelin on Feb 25, 2018 16:26:05 GMT
I'd been staying away from GBBA etc, one major reason being the diversification issues much discussed here. However, with this announcement I thought I'd chuck a bit in to GBBA2 to see what happened, and I'm happy with what I've seen, with my cash spread over more than 30 loans so far. Yes, the biggest chunks are a bit bigger than I'd ideally like, but I'm happy enough. That said, I don't think I will be making any wholesale changes to my strategy for AC, which is to focus on MLA lending as it means I can aim for higher returns and also avoid some categories of loan which I'd much rather not be in. Still, as I can't find enough loans that pass all my tests, it's good to have the option of parking a modest chunk in GBBA2 as a higher-return variant on using QAA/30DAA. The main thing that will hold me back from jumping deeper into the GBBA2 pool is the number of longer-term loans (>3 years) which - in the current climate - is not something I want to be too heavily involved in, although a broad diversification certainly lessens the risk. Shorter term loans are one of two things - loan type of bridging and development OR an account designed to seek to create that type of exposure in other ways but still likely to be quite focussed on those type of short term loans on the whole. So a one year Access Account OR a development and bridging investment account could suit ? In my view, you run the risk of complicating things if you introduce too many different account types. You also then run into the same problem you had with the GEA, which is a shortfall in available loans due to changes in economy/business or whatever. I think you have the right idea with the planned automated MLA accounts. As long as you categorise each loan carefully and give the user a broad selection of criteria to choose from, then users can craft the exact type of account they want.
|
|
cb25
Posts: 3,528
Likes: 2,668
|
Post by cb25 on Feb 25, 2018 17:40:46 GMT
I'll wait until I see a majority of people on this forum being happy with the results, considering i) allocation percentages seem to have gone in the wrong direction for some ii) routines don't seem to have been fully tested before implementation, requiring amendment soon afterwards iii) more changes to come in 8-10 weeks (if these are needed/desirable, suggests there's still something wrong with the current algorithm) ii) ... I'm also not aware of other platforms offering direct access to their CTO like this so where they've tweaked and altered systems they've done so silently without lenders noticing. Can anyone explain precisely how the loan matching works on RS or Z for example? I don't know the details of other platforms loan matching, but maximum percentage of portfolio I have in any single loan is: Z - 0.2%, FC - 0.5%, RS - 7.4%, AC - 20%+ re CTO access: nice as that is, I'd readily swap it for a guarantee the PF would kick in within (say) a year of a loan defaulting AND maximum allocation of 1% of portfolio in any single loan (ideally any single borrower). One might argue that a PF makes percentage in any single loan irrelevant, but imo it does still matter - platforms can fail, so the greater diversity a lender has, the better.
|
|
|
|
Post by chris on Feb 25, 2018 17:40:47 GMT
Shorter term loans are one of two things - loan type of bridging and development OR an account designed to seek to create that type of exposure in other ways but still likely to be quite focussed on those type of short term loans on the whole. So a one year Access Account OR a development and bridging investment account could suit ? In my view, you run the risk of complicating things if you introduce too many different account types. You also then run into the same problem you had with the GEA, which is a shortfall in available loans due to changes in economy/business or whatever. I think you have the right idea with the planned automated MLA accounts. As long as you categorise each loan carefully and give the user a broad selection of criteria to choose from, then users can craft the exact type of account they want. Going forward the accounts will be implemented on top of the bespoke accounts, and there'll be overlap rather than loans tending to fall into only a single account. That will alleviate the pressures on lack of available loans. There are other planned changes to our origination strategy to address this as well. I suspect this is going to be a big year for us technologically with plenty of foundations and new ideas being brought online to stand us in good stead for the next few years.
|
|
