Sharing my portfolio returns is something that I intended to do for quite some time, but I wanted to wait to season it for a good while. Now that I'm over the "hump" of 10 months on average, I thought it time to share.
SetupThe following statistics are based a portfolio that I created in January 2012 exclusively using the P2P-Picks Profit Maximizer. The following schedule describes my deposits:
8 Jan 2012 - $5,000
22 Mar 2012 - $5,000
1 Jun 2012 - $5,000
13 Jun 2012 - $20,000
This portfolio is not traded on the secondary market. It attempts to be the optimal buy and hold strategy from newly issued LendingClub loans. All payments have been reinvested into the portfolio.
SeasoningThe attached graphic summarizes the age of Issued, Current, and Late notes in my portfolio. The unweighted average age of the loans in the portfolio is 295 days. The average age weighted by remaining principal is 278 days. The months with the most notes are now well past months 6-10, which have the highest percentage of charge-offs.
PerformanceCurrently LendingClub reports my NAR at 12.0%.
An XIRR calculation on my current balance and the cash deposit dates is 11.1%. This, however, also measures my somewhat slow deployment of that $20,000 deposit, and is biased below the true performance of the strategy. Spreading out that $20,000 deposit into 4 x $5,000 deposits each a couple weeks apart bumps up the XIRR to 11.5%.
Looking toward the future, I expect the NAR to bounce between 11.7% and 12.3% as the portfolio continues to push over the 10-month hump and experiences relatively fewer charge offs (compared to the last 3 months). The XIRR should also push closer and closer to the NAR as (a) the effects of cash drag decrease through automated investing and (b) the $20,000 deposit cash drag effect is diluted over time. This is precisely where I expected to be given my backtesting. You can see the backtesting in Peter's article (look where the curve meets the imaginary vertical line at the Top 10% policy).
http://www.lendacademy.com/p2p-picks-hedge-fund-style-p2p-investing/My active portfolio has 959 Issued and Current notes, 4 Late 16-30s, and 12 Late 31-120s, and 0 In Default. There are also 101 Fully Paid loans and 25 Charged Off loans that are no longer part of the active portfolio. I feel this is a solid distressed note performance given the portfolio size.
[attachment deleted by admin]
Thanks for the info. Very impressive performance. It does beg the question of the precise definition of the Bryce Mason buy and hold strategy. All are PMAX, so how many were Top 1's, 5's and 10's? Were all notes the same size, or did you vary note size by Top (if you used varied Tops)?
Finally, if I read correctly, your portfolio was auto-invested beginning at some point in time. If so, do you have a way to measure the improvement in the quality and availability of loans purchased, if any, due to this perceived advantage over slower manual browser based purchasing? If you do, what did you find? If not would you care to speculate? As you know, and for the record, I personally do not think auto-invest is currently significant so long as LC continues to release at pre-announced times. Otherwise it is obviously mandatory.
Cheers,
Rob
Hi Rob, thanks for your attention and feedback, as always.
Everything was $50 per note. I just ran down the list from the top loan to the bottom (until cash available was exhausted). Sometimes I'd reject a loan for personal reasons, but not too often even in the beginning and certain more rarely today. I'd say this is a good reflection of the Top 10% strategy (buying a good mix of 1%, 5%, and 10%). These days I only infrequently make it out of the Top 5% because I'm only buying 4-5 notes at a time, so maybe a few years from now my portfolio will look more like a Top 5%. There were some weeks in July-August 2012 where I bought 120+ notes at a time. There was huge--HUGE availability while LC was helping a major fund ramp up.
Auto-invest has not yet been turned on. I've always done it manually. But, it's coming! Frankly, I do not expect the performance to differ once it's turned on. Nobody has been able to tell me a factor yet that creates any evidence of differential performance across picks within the product. I used to look at every loan I bought. These days I just "Go" and "Place Order." The lust is gone and it's just another chore. If I don't just trust the model, why bother at all?
Hi Bryce,
Thanks for the clarifications. Glad to know your portfolio contains notes from all the PMAX Top rankings. I think that's an important point, and a question that surely would have been asked by others had I not. I'm surprised you" looked" at any of the loans early on since that was contrary to the fundamental premise of your work. However, I suppose there was a period of reasonableness cross checking just to be sure nothing screwy was going on, and hey real money was involved.
My reinvestment strategy is looking like yours. I'm going wiith PMAX 1% and 5% only but waffling between a constant $50 per note, or $75, $50. Since I have more than 300 of each now I can view them as a separate investments with sufficient diversification within themselves. I think I remember 400 notes to be the point at which further diversification doesn't mean much.
I see we agree about auto-invest and its lack of any meaningful difference (other than convenience) in the past and for now. If continuous release is in LC's future though it's a game changer, but for my investments over the past month I'm confident I haven't missed a thing.
So, the building of the machine was more fun than turning the crank. Makes sense, but the sausage is tasty.
Cheers!
Rob
Hi Bruce,
Thanks for sharing your data. It's good long term data. However, how do you see going forward when so many investors competing so few quality loan. At 10:05am PT today, your max. profit model only had 1 top 10% note available. By 10:09 am, there is none. I assume all those top ones are taken. So investors will have to take lower and lower quality loans and lower and lower returns. Don't you have to readjust your model and lower return rates?
10 a.m. this morning was filled exceptionally quickly. Between the 6 a.m. and 10 a.m. loads, however, I was able to purchase 7 loans today. I don't have any expectations for change due to loan availability.
Auto-invest will help alleviate this, too, when it's implemented.
Auto-invest will help alleviate this, too, when it's implemented.
+1
Bryce, have you ever compared how the BCDEF notes picked by LM compare with the Top 1, 5, 10 on PM?
Maybe I'm getting slow on the trigger, but it was really difficult to buy PMAX loans yesterday, and even the 6am upload this morning was blitzed. I wonder if this is a brief abnormality or a paradigm shift. I've never seen such heavy buying at the top of the morning; 147 new loans listed, 0 fully funded in the first two minutes, 16 in the first 3(actually first 2 minutes and 5 seconds), still 16 in the first 4 and finally 19 in the first 5 minutes. All but 2 were 36 month loans.
Probably should have posted this in the unashamed griping thread.
August 2013 Update
BalancesAccount: $39,996.50
Impairments*: $981.00 (2.5% of Balance)
TransfersDeposits: None
Withdrawals: None
Active Portfolio DescriptionLoans: 1007
Average Age (unweighted): 310 days
Average Age (weighted by principal remaining): 290 days
ReturnsLendingClub Net Annualized Return: 12.1%
Excel XIRR (inc. cash drag): 11.3%
Strategy- Buy and hold LendingClub securities - no secondary market activity
- Idle cash invested using all available P2P-Picks notes, from Top 1% to Top 10%
* - Impaired loans are all Late and In Default loans.
[attachment deleted by admin]
Is that only PMAX notes?
Yes, per the original post.
Thanks for the update. We should all hope to do so well! Are you surprised at your "Fully Paid" amount or has it been as expected? So far I've had 13 of my 1700+ loans paid in full in their first month, and only 3 that are 15-30 late. That "Fully Paid" number seems odd to me but I have no data to back it up. Those borrowers are getting killed by origination fees.
I haven't done such a tally, Rawraw. I suppose I could do such a thing, but it might be difficult because very early on in 2012 I was not keeping the scores around. My feeling is that I am heavily weighted in 10% notes, but that will shift to 5% over the next year.
Zigrtd
Point taken. In future updates I'll try to provide the portfolio mix characteristics.
September 2013 Update
Balances
Account: $40,310
Cash/In Funding: $524
Impairments*: $1222 (3.0% of Balance)
Transfers
Deposits: None
Withdrawals: None
Active Portfolio Description
Loans: 1030
Average Age (unweighted): 326 days
Average Age (weighted by principal remaining): 303 days
Returns
LendingClub Net Annualized Return: 12.0%
Excel XIRR (inc. cash drag): 11.2%
Strategy
Buy and hold LendingClub securities - no secondary market activity
Initial deployment Top 10% graded notes. Reinvestments Top 5%.
* - Impaired loans are all Late and In Default loans.
[attachment deleted by admin]
October 2013 Update
Note that LendingClub made a change to its reporting and that accrued interest is no longer a part of the account balance. This may be a fair exchange, as previously I wasn't writing down the impaired portion of my portfolio. This has the effect of dropping the XIRR measure of return on investment.
Balances
Account: $40,335
Cash/In Funding: $788
Defaults
Impairments*: $1343 (3.3% of Account)
PY Defaults: $1564 (3.9% of Account)
Transfers
Deposits: None
Withdrawals: None
Active Portfolio Description
Loans: 1070
Average Interest Rate: 16.8%
Average Age (unweighted): 354 days
Average Age (weighted by principal remaining): 318 days
Returns
LendingClub Net Annualized Return: 11.5%
Excel XIRR (inc. cash drag): 10.0%
Strategy
Buy and hold 36-month LendingClub securities - no secondary market activity
Initial deployment Top 10% graded notes. Reinvestments Top 5%.
* - Impaired loans are all Late and In Default loans.
[attachment deleted by admin]
November 2013
In November, LendingClub again made changes to the way it presents returns, and thus P2P-Picks had to reconsider how best to present its performance. We mostly mirrored their method, but have one fundamental difference in the account balance that influences the alternative returns (XIRR) calculation. In particular, we took a write-down for impaired loans (according to the default LendingClub parameters), but we also included accrued interest in our estimated account value. We felt that most accountants would consider accrued interest payable within a month to be a current asset (we discounted accruals by distress level just like the principal amounts).
Another fantastic change is that LendingClub now allows users to compare their returns against all buy-and-hold strategies. P2P-Picks Profit Maximizer appears to be achieving its goal of being in the top decile of performance compared to all buy-and-hold accounts (while at the same time recommending ~10% of the population of loans). You can see this visualization in the attachment.
Importantly, we believe that LendingClub may not be calculating portfolio characteristics correctly--erroneously including fully-paid and charged-off loans that should be scuttled from the portfolio--as the portfolio age reported by LC is 13.1 months (whereas even the most generous age we calculate on active notes is just under 12 months--and the average weighted by principal remaining is 11 months). An average weighted interest rate of 16.67% doesn't jibe with our calculations, either. In any case, such miscalculations do not detract from the comparison value of the chart, as such biases would be shared equally among all accounts.
BalancesLC Reported Account Value | $40,685
Have you been able to analyze your late notes to identify any common contributing factor? 4% of the total number of notes, higher if you look at dollars, seems high imho.
I'm surprised that the weighted average interest rate is twice the NAR
(1) I'm in the bottom-most part of the default curve--I suspect the 12-month trailing defaults will smooth out as the portfolio ages.
(2) Recall that the purpose of P2P-Picks is to make the best portfolio possible with the expected top 10% of all notes. One can certainly design a very tight filter that can outperform, but good luck investing tens of millions with it. I'm trying to design a solid-performing product for what I hope to be thousands of users.
Q1 2014 It's time to update the public on our performance. It's been a bit over a quarter since our last update, and I thought that quarterly updates are probably a better use of my time while still providing decent information. This data is as of April 4th, 2014. Our methodology for reporting hasn't changed since the November post. Basically we're calculating Estimated Account Value to be the total Principal + Interest outstanding for all loans still in the portfolio, less LC's discounted amount for impaired loans. We feel accrued interest payable within a month is a current asset. The key takeaways for this review are (1) our average interest rate has been creeping up as our older B loans (identified by P-Max v1) are repaid and the principal reinvested into D & E loans, (2) as seen in our lower impaired loan counts we have emerged from the most distressing part of the charge off curve for the huge $20,000 deposit that we made in June 2012, and (3) NAR and XIRR have creeped up to the mid 9% range. Barring external events, we anticipate this level of performance should continue. We are particularly proud of where we show on the "Understanding Your Returns" chart, given that the strategy selects such a large volume of 36 month notes (~5% of the platform). BalancesLC Reported Account Value | $41,569
With respect to the "Understanding Your Returns" chart, I still find it funny that LC appears to be taking into consideration all repaid & charged off loans in their calculation of age & average interest rate. Once loans are repaid or charged off, they are no longer part of the active portfolio.
I tried to overcome this by making a separate portfolio for charged off and repaid loans but you can't ask LC to do one of these charts for a specific portfolio. They do it for the whole account. LC should call it the "active portfolio" and remove those types of loans (possibly even "in review").
Thankfully my whole account is dedicated to P-Max so the NAR is accurate, but the reader should trust my average portfolio characteristics rather than looking at LC's average interest rates, age, and number of loans.
Very nice job Bryce. Thanks for the update.
My guess is they discount the interest in the same way as the principal. That's what I did,
Per https://www.lendingclub.com/public/lendersPerformanceHelpPop.action: Here is the formula for NAR: And Adjusted NAR: No Accrued Interest in either one. The Interesti is for paid interests from month 1 to now (month N).
Hey Bryce, really enjoy seeing your updates as a majority of my account is filled with your pmax strategy. Can you give any insight to why you are reinvesting into D & E and not C loans?
Q2 2014 The key takeaways for this review are (1) our average interest rate has been creeping up as our older B loans (identified by P-Max v1) are repaid and the principal reinvested into D & E loans (from 17.7% in Q1 to 18.1% today) and (2) NAR and XIRR continue to creep upwards. Barring external events, we anticipate this level of performance should continue. We are particularly proud of where we show on the "Understanding Your Returns" chart, given that the strategy selects such a large volume of 36 month notes (~5% of the platform). LendingClub's snafu with loan application data caused an outage in our strategy for almost three weeks, thus the stymied originations in "Month 0" in the age chart. We expect originations to be similarly low in months 3 and 4 in next quarter's update. This may have a modest negative impact on our returns due to increased cash drag, but probably nothing major. BalancesLC Reported Account Value | $42,905
Are you interested in user returns? I have a few separate P2P-Picks portfolios that I'm tracking the returns, but I do engage in Folio Sales of risk indications (such as crashing FICO)
It might be fun to have a separate thread of user returns, but I would need to make some guidelines for participation such as knowing whether there were sales, targeting of specific grades, etc. Then we might get a confidence interval of the strategy. However, I suspect my personal portfolio is close to the mean, as it's a random subset of the picks.
Define management of holdings.
Please refer to the first post on this thread, which details the timing of my cash outlay. It's been reinvested since those dates.
After three and a half years of maximizing returns, I've decided to reduce the risk of the portfolio and invest in A/B grade paper. This is mostly due to my account being taxable. Each year I wind up roughly with enough capital losses to max out my $3k deduction. I do not want to be in a situation where the economy stutters, defaults double, I have to pay taxes on all the interest, I'm stuck with a tax asset I might not use for some time, and the after-tax growth on the portfolio is 0 or negative. The A/B portfolio should have roughly 1% defaults per year in steady state, allowing me to add capital to the account. It will be a long haul to turn over the whole portfolio!
Here's the final resting place of the returns: 9.7%. The blue dot's in a good place. Pretty satisfied with that for a buy-and-hold strategy.
Not really, as I've changed strategies.
|
|
|