bugs4me
Member of DD Central
Posts: 1,845
Likes: 1,478
|
Post by bugs4me on Jun 3, 2014 18:22:17 GMT
chris, I take exception when you say users who "blindly invest in every loan" should be regarded as "not fair on large swathes of our lender base". This kind of investment behaviour is a personal decision like any other. You've said previously that loans will be categorised so that the AI2 can be told to (blindly) only buy certain types of loan so this is already something that, I think, you want to provide for. It's different if they are attempting to gain an unfair advantage by opening multiple accounts but then doing that might allow them to practice all sorts of other undesirable stuff not just this perfectly rational and understandable investment behaviour. The elephant in the room is quite visible I know. Do you think the thousands of zopa and FC members are careful with each and every loan decision they make? There is basically virtually no loan decisions to make on Z*** any more. The only decision is whether to turn on auto invest or not. Plus of course whether to add money or withdraw. That's effectively all there is and lending speed, interest returns (apart from the Rate Promise promotion) is down to TGLA. No longer familiar with FC as I got out with a few pennies profit after the PG/DG fiasco so cannot comment on that one. Think ATM I'll leave it with chris and see how it develops. So far so good with them although the flow of loans is of course outside of their control.
|
|
|
Post by westcountryfunder on Jun 3, 2014 18:26:32 GMT
chris what's radically wrong with the current AI set-up? Most people who have commented seem happy enough and I don't remember you detailing it faults. I could have missed that though to be fair. Mike As far as I am concerned what is wrong with AI is that it doesn't work! On 1 June 17:17 and 2 June 13:15 AI claimed to have bought for me loan parts in Kent Bridging Loan. The money was there but it just didn' t happen. Not on my statement, and definitely no units acquired. Sorry to say, AI is rubbish at the moment, but I keep hoping.
|
|
|
Post by chris on Jun 3, 2014 18:28:24 GMT
Someone with 5 accounts would be able to deploy their cash 5 times faster than you no matter what the cap. You also fail to deal with all the other issues I've raised some of which are significant. Chris, I'm slightly at a loss here. I fully comprehend that someone who has 5 family accounts will be able to deploy 5 times the amount to invest than say me with my sole account. But if they are only allowed to buy on AI on say a loan on AM, is it really going to make a huge difference whether they managed to buy £20 or £100 worth? If you were talking allowances/caps of say £1k each, then I would fully agree there. That presumes a single rotation. If all orders are fulfilled and it goes round the loop a second time they get 5 hits to your 1. So if you're both trying to deploy £1,000 in a single loan even if we rotate round in £20 chunks they'd get there with 5 fewer loops through the investor base than you.
|
|
|
Post by chris on Jun 3, 2014 18:33:04 GMT
chris, I take exception when you say users who "blindly invest in every loan" should be regarded as "not fair on large swathes of our lender base". This kind of investment behaviour is a personal decision like any other. You've said previously that loans will be categorised so that the AI2 can be told to (blindly) only buy certain types of loan so this is already something that, I think, you want to provide for. It's different if they are attempting to gain an unfair advantage by opening multiple accounts but then doing that might allow them to practice all sorts of other undesirable stuff not just this perfectly rational and understandable investment behaviour. The elephant in the room is quite visible I know. Do you think the thousands of zopa and FC members are careful with each and every loan decision they make? There is basically virtually no loan decisions to make on Z*** any more. The only decision is whether to turn on auto invest or not. Plus of course whether to add money or withdraw. That's effectively all there is and lending speed, interest returns (apart from the Rate Promise promotion) is down to TGLA. No longer familiar with FC as I got out with a few pennies profit after the PG/DG fiasco so cannot comment on that one. Think ATM I'll leave it with chris and see how it develops. So far so good with them although the flow of loans is of course outside of their control. I can categorically say that within the bounds of physics (an automated decision will always be quicker than a manual decision, but we'll mitigate where we can and AI is one of those tools where you can manually decide but automatically buy loan units) we want to give you as much or as little control as is right for you personally. Fully automatic works for some, look at how RateSetter has grown for example, fully manual works for others. As far as we can we'll try and cater to all strategies but I'm sure a few compromises will need to be made.
|
|
|
Post by chris on Jun 3, 2014 18:34:46 GMT
chris what's radically wrong with the current AI set-up? Most people who have commented seem happy enough and I don't remember you detailing it faults. I could have missed that though to be fair. Mike As far as I am concerned what is wrong with AI is that it doesn't work! I agree, it's algorithm is too primitive and too much happens in parallel which is causing clashes and race conditions. This new release aims to build upon AI to improve it and fix those issues for the few who do appear to fall through the cracks with the current setup.
|
|
j
Member of DD Central
Penguins are very misunderstood!
Posts: 2,188
Likes: 540
|
Post by j on Jun 3, 2014 18:36:32 GMT
Chris, I'm slightly at a loss here. I fully comprehend that someone who has 5 family accounts will be able to deploy 5 times the amount to invest than say me with my sole account. But if they are only allowed to buy on AI on say a loan on AM, is it really going to make a huge difference whether they managed to buy £20 or £100 worth? If you were talking allowances/caps of say £1k each, then I would fully agree there. That presumes a single rotation. If all orders are fulfilled and it goes round the loop a second time they get 5 hits to your 1. So if you're both trying to deploy £1,000 in a single loan even if we rotate round in £20 chunks they'd get there with 5 fewer loops through the investor base than you. OK, makes more sense now. How many members, at a guess as I know you cannot probably divulge actual ones, do AC think have multiple accounts with say more than 2 family members to make a significant dent & affect other singular members substantially? I am not familiar with Zopa, nor do I want to be, I like AC the most & deploy what is effectively my whole p2p investment with them. AI has not been good to me (but I am not knocking the system down, it seems to work for many others, just not me ) but, if we wnd up going down a similar route of what others have been describing on Zopa, then I would seriously rethink my position & most likely begrudgingly leave AC. Will certainly wait & see the developments, it might even improve my view of the current AI arrangements - specifically that related to AM not in general.
|
|
|
Post by batchoy on Jun 3, 2014 18:37:39 GMT
The only fair system is a queue. First come, first served. Decide the basic quantum of capital (say £100) and then rotate people in the queue. New investors should not get any priority over old investors. If this is true and can be demonstrated then I'll build it that way. But a basic queue can be gamed and does nothing to resolve the order in which lenders orders are fulfilled when thousands of them are placed electronically and instantaneously. As I can't fully describe what we're building take the example of FC's autobid system. Presume that hundreds of lenders sign up and the system automatically invests in new loans as they become available, what is the fair order for resolving that? If it's in the order they signed up to autobid then you build in permanent advantage for some lenders. Do you want auto investors to always hold advantage over manual investors, or the other way round? What about if people sign up for multiple accounts using multiple family identities in order to get multiple places in the queue? Should someone with five accounts be able to bid out five times quicker than someone who only uses one account? Should someone who blindly invests in every loan be able to deploy their cash much faster than someone who manually focuses on a couple of loans? I don't think the queue you describe is at all fair but I'm willing to be shown otherwise. The premise of your argument against the system I'm planning is that you personally will lose out, if you can show me how then I'll make sure that you don't. I have no intention of deliberately favouring any single group of users but it is a balancing act and it will need fine tuning over time. In response to your edit: Unfortunately for those who do have the time to invest in micromanaging their portfolio (not sure that's fair on those that have money to invest but not the time but that's another argument) automation tools are coming to every major platform, and I'd be utterly shocked if TC were not included. So gaining advantage from manually watching screens and having a faster finger than others isn't going to be possible on any platform of decent scale in the coming months. All I can promise is that all our lenders will have access to the same tools and that I'm here trying to work with you to make sure that you are not disadvantaged. Life is not fair, and no system however you look at it is going to be fair to everyone and the more complex and the less understandable you make it the less fair it will be perceived to be and there are plenty of examples on other platforms where in an attempt to be fairer to one section of lenders they have discriminated against another or made the queue system so complex that it is hard to fathom what is actually happening and people walk away. I have been through is with Zopa where they weighted their queues to new lenders with new money and thus discriminated against long term investors who were reinvesting their income, and I am suffering from it now with Bondora where I have gone from having my income automatically reinvested within hours to having to manually bid in a riskier market in order to prevent a cash mountain being built up because the automatic system is barely picking up a loan a week. The simplest most transparent system and thus the least unfair has to be a basic queue that allocates loan parts on a first come first served basis to the limit of of a lender's requirement within the limit of their available funds, with a people joining the queue in the order that they expressed a preference for the loan with the person staying at the front of the queue until their requirement is filled or they run out of cash. Yes there will be perceived unfairnesses if people are have large sums available to invest or have multiple accounts to invest through but manipulating the queue to combat this turns a perceived unfairness into actual discrimination. Rotating through the queue based on the lowest quantum of loan part is patently ridiculous. Transpose it to a filling station, the lowest quantum the filling station sell is a litre, I drive up to the pump with the intention of filling my tank with 40 litres and I have the cash to do so, however because someone has driven in behind me the pump restricts me to a 1 litre which I have to pay for then drive out and back round to the end of the queue and start all over again and each time I do this there are potentially more new people sat in the queue in front of me so I never know when I am going to actually get my 40 litres as I have idea of how big the effective queue will be. When it comes to automated platform bidding i.e. a platform wide AI based on global variables, then the initial queues for each loan need to be generated completely randomly from those lenders whse AI settings match the loan and with no reference to any times or monetary values. Subsequent to the initial generation of the queue lenders are added in the order that they show an interest in the loan.
|
|
|
Post by chris on Jun 3, 2014 18:39:10 GMT
OK, makes more sense now. How many members, at a guess as I know you cannot probably divulge actual ones, do AC think have multiple accounts with say more than 2 family members to make a significant dent & affect other singular members substantially? I am not familiar with Zopa, nor do I want to be, I like AC the most & deploy what is effectively my whole p2p investment with them. AI has not been good to me (but I am not knocking the system down, it seems to work for many others, just not me ) but, if we wnd up going down a similar route of what others have been describing on Zopa, then I would seriously rethink my position & most likely begrudgingly leave AC. Will certainly wait & see the developments, it might even improve my view of the current AI arrangements - specifically that related to AM not in general. There's only a couple at the moment but when people figure out it's advantageous it would balloon. Given we're pretty open on here about how things work it would be a case of when not if. I want AI to be good to you like it has been to others. It's much simpler at the moment but it means there are ways in which it can be unfair. I'd like to address those. We're not going the same route as Zopa.
|
|
|
Post by batchoy on Jun 3, 2014 18:50:06 GMT
OK, makes more sense now. How many members, at a guess as I know you cannot probably divulge actual ones, do AC think have multiple accounts with say more than 2 family members to make a significant dent & affect other singular members substantially? I am not familiar with Zopa, nor do I want to be, I like AC the most & deploy what is effectively my whole p2p investment with them. AI has not been good to me (but I am not knocking the system down, it seems to work for many others, just not me ) but, if we wnd up going down a similar route of what others have been describing on Zopa, then I would seriously rethink my position & most likely begrudgingly leave AC. Will certainly wait & see the developments, it might even improve my view of the current AI arrangements - specifically that related to AM not in general. There's only a couple at the moment but when people figure out it's advantageous it would balloon. Given we're pretty open on here about how things work it would be a case of when not if. I want AI to be good to you like it has been to others. It's much simpler at the moment but it means there are ways in which it can be unfair. I'd like to address those. We're not going the same route as Zopa. The issue of multiple accounts is a red herring, as each account must be dealt with individually regardless of who is managing it and even if you do know in detail why a person is managing multiple account and can thus discriminate between someone who is managing accounts for multiple individuals or multiple accounts just for themselves, otherwise you end up discriminating again.
|
|
bugs4me
Member of DD Central
Posts: 1,845
Likes: 1,478
|
Post by bugs4me on Jun 3, 2014 18:58:13 GMT
That presumes a single rotation. If all orders are fulfilled and it goes round the loop a second time they get 5 hits to your 1. So if you're both trying to deploy £1,000 in a single loan even if we rotate round in £20 chunks they'd get there with 5 fewer loops through the investor base than you. OK, makes more sense now. How many members, at a guess as I know you cannot probably divulge actual ones, do AC think have multiple accounts with say more than 2 family members to make a significant dent & affect other singular members substantially? I am not familiar with Zopa, nor do I want to be, I like AC the most & deploy what is effectively my whole p2p investment with them. AI has not been good to me (but I am not knocking the system down, it seems to work for many others, just not me ) but, if we wnd up going down a similar route of what others have been describing on Zopa, then I would seriously rethink my position & most likely begrudgingly leave AC. Will certainly wait & see the developments, it might even improve my view of the current AI arrangements - specifically that related to AM not in general. IMO irrespective as to the good, bad & ugly with AI (which I do not have turned on personally), at least with AC they freely engage with their lenders. Sadly Z*** seem to be more and more reluctant to do. As far as I'm aware, no one understands how TGLA works and there is zero input from Z*** to try and rectify the many theories as to how it does work - assuming of course that Z*** know themselves. A big turn off in my book. Provided AC do not follow the same route then I cannot see a problem. It's all about communication whether it's good news or not so good.
|
|
|
Post by chris on Jun 3, 2014 19:00:43 GMT
Life is not fair, and no system however you look at it is going to be fair to everyone and the more complex and the less understandable you make it the less fair it will be perceived to be and there are plenty of examples on other platforms where in an attempt to be fairer to one section of lenders they have discriminated against another or made the queue system so complex that it is hard to fathom what is actually happening and people walk away. I have been through is with Zopa where they weighted their queues to new lenders with new money and thus discriminated against long term investors who were reinvesting their income, and I am suffering from it now with Bondora where I have gone from having my income automatically reinvested within hours to having to manually bid in a riskier market in order to prevent a cash mountain being built up because the automatic system is barely picking up a loan a week. The simplest most transparent system and thus the least unfair has to be a basic queue that allocates loan parts on a first come first served basis to the limit of of a lender's requirement within the limit of their available funds, with a people joining the queue in the order that they expressed a preference for the loan with the person staying at the front of the queue until their requirement is filled or they run out of cash. Yes there will be perceived unfairnesses if people are have large sums available to invest or have multiple accounts to invest through but manipulating the queue to combat this turns a perceived unfairness into actual discrimination. Rotating through the queue based on the lowest quantum of loan part is patently ridiculous. Transpose it to a filling station, the lowest quantum the filling station sell is a litre, I drive up to the pump with the intention of filling my tank with 40 litres and I have the cash to do so, however because someone has driven in behind me the pump restricts me to a 1 litre which I have to pay for then drive out and back round to the end of the queue and start all over again and each time I do this there are potentially more new people sat in the queue in front of me so I never know when I am going to actually get my 40 litres as I have idea of how big the effective queue will be. When it comes to automated platform bidding i.e. a platform wide AI based on global variables, then the initial queues for each loan need to be generated completely randomly from those lenders whse AI settings match the loan and with no reference to any times or monetary values. Subsequent to the initial generation of the queue lenders are added in the order that they show an interest in the loan. Life may not be fair but that doesn't mean I should accept a system that is open to abuse. The simplest and least open to abuse system would be random lottery of those with funds available at point of sale on the aftermarket, and in many ways that would be fairer than any queue based system. In your own example you start deviating from the "simple" queue when the loan is first listed, but that's just one scenario where inequality and abuse will set in.
|
|
|
Post by chris on Jun 3, 2014 19:02:45 GMT
There's only a couple at the moment but when people figure out it's advantageous it would balloon. Given we're pretty open on here about how things work it would be a case of when not if. I want AI to be good to you like it has been to others. It's much simpler at the moment but it means there are ways in which it can be unfair. I'd like to address those. We're not going the same route as Zopa. The issue of multiple accounts is a red herring, as each account must be dealt with individually regardless of who is managing it and even if you do know in detail why a person is managing multiple account and can thus discriminate between someone who is managing accounts for multiple individuals or multiple accounts just for themselves, otherwise you end up discriminating again. Not if weight of capital or weighted lottery is used (not that I'm proposing it is). The same amount of capital on a single account or across multiple accounts wouldn't matter. But in a simple queue even just two accounts would mean you deploy cash twice as fast as your neighbour.
|
|
|
Post by chris on Jun 3, 2014 19:11:49 GMT
Chris, Apologies I seem to started a small brush-fire ... or at least re-lit it! The difference for me is more conceptual. I suppose I am used to a market which is regulated (rules that define how investment can/cannot occur) and then there is order-management system (often with an automated investment engine). In your model it seem the investment engine IS the market (or will become it eventually), applies all the regulation, and then feeds down to an order management system. The problem for me is that in your model the investment engine will enforce a bunch of regulations on the way the market works. However, these regulations are hidden inside what you have implied might be some stochastic (or pseudo-stochastic) engine that investors won't be able to predict the functioning of. The bottom-line is that if I can't predict the functioning of the engine, how can I have any confidence in the market and whether it is fair or not? If the principles AC eschew are things like "new lenders are favored over old lenders", "small lenders are favored over big lenders", "automatic lenders can't outbid manual lenders" and "those with time on their hands can't outperform those who don't who don't have time on their hands" then fine. I don't agree with any of those ideas but at least be open enough to state these top-down principles rather than bury them in a black box. It seems we may not even know qualitatively what the principles are when we should be know both qualitatively and quantitatively what they are. Exactly what is a big vs small lender? Exactly how much cash balance do I need to buy 1k of a loan vs £100 of a loan. What exactly is a "diversified portfolio" and how will it impact what loans I buy on the AM. My fear is that either AC don't want to tell investors what the regulations/principles of the AM are because it might alienate them (a la ZOPA) or the AC can't actually state those principles in a simple deterministic set of rules (in which case hold off until you can). I've ranted on again way too much... apologies again. I do somewhat regret starting this debate AC doesn't support any of those principles in particular nor are we saying we'd heavily weight the markets. If someone came on to the platform with £100 and all their cash had to be deployed before a HNW with £1m already invested in our platform could get a loan unit on the aftermarket then that wouldn't be fair either. I can understand the giving that new £100 lender a small helping hand, say a 10% greater chance on the aftermarket, to get their first £100 deployed or if one lender chooses to focus on a handful of loans by only investing in them then giving them a small helping hand over someone just spreading their funds across as many loans as possible. But it would only be small tweaks to the natural order not anything like the Zopa system as it's been described. Or perhaps lenders we want to give a helping hand to can invest in £110 chunks whereas others invest in £100 chunks, whatever it is. Given we can split loan units arbitrarily that becomes practical. I can't tell you what the rules will be today but I can by the time the system launches and I will happily divulge the way the system works and the way in which we try to balance things. A queue is just as arbitrary a solution as any other and I can't see the benefits it brings *except* that it's simple to understand and explain conceptually. That is a big positive but not if it breaks the workings of the system and alienates users as they can't get their funds deployed due to people gaming the system.
|
|
|
Post by batchoy on Jun 3, 2014 19:13:13 GMT
The issue of multiple accounts is a red herring, as each account must be dealt with individually regardless of who is managing it and even if you do know in detail why a person is managing multiple account and can thus discriminate between someone who is managing accounts for multiple individuals or multiple accounts just for themselves, otherwise you end up discriminating again. Not if weight of capital or weighted lottery is used (not that I'm proposing it is). The same amount of capital on a single account or across multiple accounts wouldn't matter. But in a simple queue even just two accounts would mean you deploy cash twice as fast as your neighbour. But only if the money in all the accounts was the property of one lender. In addition to my money I manage the funds for two family trust funds which if the trustees all agreed could be invested in AC, thus giving me three accounts at which point under your rules the accounts would be discriminated against because they were under my management even though only one of them actually contained my money. At this point because I am discriminated against and getting a lesser return on each individual account due to the system treating them as an aggregate account I walk away and take all my business elsewhere.
|
|
|
Post by chris on Jun 3, 2014 19:17:31 GMT
But only if the money in all the accounts was the property of one lender. That's the scenario a queue opens itself to. We already have users with multiple accounts because they manage their own funds and those of their business, there are others who have several family members lending and we have no way to tell if that's one person or several. In addition to my money I manage the funds for two family trust funds which if the trustees all agreed could be invested in AC, thus giving me three accounts at which point under your rules the accounts would be discriminated against because they were under my management even though only one of them actually contained my money. At this point because I am discriminated against and getting a lesser return on each individual account due to the system treating them as an aggregate account I walk away and take all my business elsewhere. Under the queue system you'd gain an unfair advantage. Under a capital weighted lottery you would have the same chances whether you had one account or three, but if you have £1m to deploy as cash on the system you have greater chance than if you have £100. I don't think that's fair either as it biases things too far towards the large investors (they'll always be someone with deeper pockets than you). So some kind of limit or cap on the weighting may be preferable.
|
|