Ghost models

Hi all,

just a quick post to get your feedback (both from users and P123 staff) on this question: Why are so few people subscribing to designer models?

As example, I attach the top 10 designer models ranked by 1 year performance: there are only 4 subsribers in total.

Hast the P123 community become a pool of designers only? Are we not trusting other models due to the failure of our own algorithms? Or are the current designer models ranking criteria not adequate (i.e. models scoring above 90 but underperforming their benchmark over the last 1-1.5 years)?

If we are not succeeding with individual models, should we think about a “P123 index” - a kind of book which groups the top 10 to 20 models together? Or what else do you think would attract more interest to subscription to both models and the P123 website?

Br,
Florian


The top ten models w/r/t 2yr or 3yr performance have more subscriber. In both cases, it’s around 30. So maybe potential subscribers want to see 2+ years of out-of-sample data?

Or maybe there are too many DMs and it’s too hard to find an attractive model? Whatever criterion P123 is using for it ‘Top Designer Models’ listing, it doesn’t catch the most popular DMs (that is popular DMs that cost $$$; free models have lots of subs). Maybe some users have the same problem.

Also, there’s no good way for Designers to directly promote their own models. I would love to have a landing page on P123 that listed just my models. I could then use that link in emails, business cards, etc.

Or maybe most P123 members just prefer to build their own models. It can be amazingly fun. And build-your-own is the original P123 philosophy.

Walter

Hi. I think the display and calculation using the Multi-Sort criteria is the problem. These suggested “top designer models” are not really attractive to many users.

Why not have different criteria (such as top 1-year return, top 2-year return, top 1-year Sharpe ratio, top 2-year Sharpe ratio, showing only defensive models, showing only aggressive models, etc.)? I would suggest presenting a user each time a different ranking the user visits the designer model page, and not always the same Muli-Sort ranking.

I am not a designer (for others) at this time, so just trying here to provide a bit of info from the perspective of a potential DM customer who is cost conscious while obviously wanting to “beat the market.” Attached is a plot of the number of subscribers versus monthly subscription cost for open and full models, ignoring the free ones of which there are 37 models with nearly 800 subscriptions. One point to the data is that the number of subscriptions begins to drop for subscription costs above $20/month. So that, combined with the large number of subscriptions for free, seems to show significant resistance at higher costs.

Of the top ten 1 year return models, only one costs $20 or less per month. Four of the top ten 2 year excess return models (one of which is full of subscribers) costs $20 or less per month. In addition to subscription costs, a subscriber actively trading a DM will also absorb transaction costs so the turnover rate can be very important.

Currently a potential DM customer has to consider the amount of capital they might apply to a model, then estimate the transaction plus subscription costs over a period of time versus a hoped-for return to compare models. Perhaps Portfolio123 might consider providing a very basic calculator to assist comparing models after applying assumed transaction fees? Or just adding a metric per model that uses an assumed set of inputs? That type of analysis is necessary, in my opinion, to confirm and compare the value of the models. Obviously all kinds of warnings about the assumptions used would be necessary.


DMSubsVsCost.PNG

My biggest issue with most models is the # of monthly transactions. I think the outperformance in practice is a lot lower once spillage takes affect even with large cap models. I wouldn’t want to spend money, do a dozen+ transactions per month only to have a good probability of still underperforming.

I’d be way more open to subscribing to models with lower trading requirements.

THE MORE TRANSACTIONS THE BETTER for me.

I can clearly beat the P123 calculations for slippage. More transactions means more incremental improvement over the P123 models’ calculations.

This does assume that the outperformance is real and will continue. But the average model does outperform its benchmark.

So, for the average model, more transactions might improve the performance. It might be worth checking your slippage and comparing brokers.

I have to at least ask myself: In a pretty efficient market, how long does the information-gain from P123 last? When does the rest of the world—including institutions—catch up.

If you can trade infrequently and make the same returns—you should continue to do it. It is easier.

-Jim

Who are the targeted users for DMs?

I do wonder these days how many people really feel comfortable owning individual shares, as compared to ETFs, for instance.
People have been trained over the years to believe that either you cannot really beat the market with conisistency over time or that the market is ‘rigged’ so only a few make money. So the thinking goes: just own an SP500 index fund and a bond fund and hope.

If the DM targeted market are advisors of some sort who trade on their clients behalf, maybe the P123 marketing efforts should be focused more on them. I am not in that industry so don’t know how big that base could be. My sense is it is shrinking too.

I just trade my own models. At least I know what is inside of them and on average, they do pretty much what similar models (in terms of liquidity of the trades and turnover) provide from P123 DM providers. I use mostly Russel 1000 stocks with TO of less than 150%. Frankly, all I really care about is having a better return with somewhat less drawdown, for my overall portfolio, than SPY. It is a low hurdle but I still lost against SPY in 2018. So still a challenge…

So who are the targeted users?

Sure, the ones that you manage to buy/sell. But how often does the price move away from you before the market even opens. P123 will report “buying the stock” but you never stood a chance because it opened way outside what you were willing to buy.

Transaction day prices are backfilled in live portfolios. Any jumps in price will affect the (H+L+2*C)/4 recorded price and the port’s equity curve will reflect that effect.

Different answers for different people and different types of ports. But to answer the question for my case: I ALWAYS buy (sell) the stock.

Even if you use the price of the previous close and use this “signal price” as the definition of slippage, I still beat the sim’s performance because it always buys the stock too (with even greater slippage). So, I think it is just a fact that I beat the sim (or say, a port on automatic).

Does infrequent trading mitigate this problem for you? I so, how? Do you still let some go when the price does not come back?

-Jim

I just don’t think the total addressable market for DMs is that big. P123 could try to scalp from other platforms based on features and designs wins. However, that list of competitors is very short.

This is why I applaud P123 steering towards providing AMs with a more complete set of AM tools. The AM market on the other hand, while facing its own set of competitive pressures, is still very large.

I’d love to someday open up my own family office. I imagine it be much easier if I could run investment research and investment operations from a single platform, which could very well be P123.

The early days of R2G indicated that there is a reasonable market for DMs. The problem is that we aren’t being provided with the tools to properly market the models. As a marketer, I need direct signup to any model by nonmembers of P123, I need to be able to embed current model performance (graph) on an external site, and be able to use affiliate link for any page on P123, not just the front page.

I agree with all of that. I just dug up a screen capture of a 2014 DM (not one of mine) that had 70 subscribers. So at one time, the marketplace was pretty active.

Tools for Designers to directly market their models is critical. I like the idea of direct signup and/or a P123 provide Designer landing page and/or P123 providing just backend services. Models like those found on allocatesmartly could probably be built on P123, but if they were offered as DMs they would quickly get lost in the hundreds of current offerings.

Also, I would like to see affiliates get 10% of DM revenues for authors willing to give up 10% to have others market their models. I would love to market the best DM models but I won’t do it for free.

This is a big question and one that Marco and I have and will continue to be thinking about.

We learned at the start that there was a market for this sort of thing but as had been discussed heavily in the forum, version 1 was way too curve-fitting friendly, and as many models faltered the way such models eventually do, subscrbers vanished en masse.

One person in this thread mentioned cost. Yes, absolutely! Pricing by many designers is unreasonably high. A while ago I suggested that designers try to envision how much money a user might invest in designer models, imagine how many models, and how much they might attract. Then, multiply the monthly charge to infer an annual %. I’m not sure anyone did this because if they did, they’d be shocked into slashing prices.

I had also posted in the past about how designers present themselves and their models. Many had ridiculous names that not only failed to indicate what the model was about but were impossible to remember. And I’ve long believed that marketing that focuses on backtests, forgetting legal implications, is marketing suicide because everybody hots cold spells here and there and if you offer a client nothing more than an implied promise of great returns that match a great backtest, you have no chance to retain them unless they make the sort of money they envision from the testing. Backesting and simulation are p123 products — tools — for your own use as an aid to development of stratgies. That’ s what they are; tools for your private use. Results can not be used as products that can be sold to others unless you become super famous in which case they can be sold to collectors, like preliminary sketches by Rembrant.

We had hoped that when we sharply curtailed test presentations, designers would market themselves based on . . . themselves. After all, that is the “product” being sold. It means talking about the market, about how one approaches investing in general and stock selection in particular. It means conversation in the forums about substantive topics like that. It means blogging etc. in that manner on Seeking Alpha and hopefully in the future, on p123. Some folks have put themselves out there and have been developing public personas. More need to do likewise. Nobody subscribes to test results, but people will subscribe to individuals they see as authorities. A big reason why I wrote and posted the strategy design course was to help those whose professional backgrounds lie in other fields to learn how, not only to design professionally sound models but also to develop the idea inventory needed to project marketable authority.

This isn’t to say p123 was always in the right and has the perfect platform now. We weren’t. We aren’t. And we don’t. And we will be working to do better. We’ve also been thinking about newsetters and I’ve been wondering if maybe DMs might be superceded by subscription to some sort of designer (using the current label) content set that could established by the designer, which might include a newsletter, a systemic model along the lines of what we have now, or even ideas that are less systemic — the p123 platform can also be used to generate ad hoc ideas. Perhaps some combination, as the designer wishes (and we wlii need to replace the word “designer”j.

But I’m putting my shtick out there yet again because I know full well that there is nothing p123 can do to get $$$ for designers unless designers pitch in an do the things they need to do for themselves. Acronyms are popular so let’s try PPAPA . . . Proper Pricing And Projecting Authority. That doesn’t come over night, so I suggest designers start working along those lines now, while we do our thing. It would be great if we could all come together at more or less the same time.

Should I launch this model as a DM for $10 per month?

It should be useful to park one’s money if one is unsure where to invest. Higher annualized return than SPY, but with a max D/D= -3.3%. Shows gains for every calendar year and includes variable slippage…

Annualized Return (%) 5.95
Max Drawdown (%) -3.28
Standard Deviation (%) 2.75
Sharpe Ratio 1.56
Sortino Ratio 2.23
Correlation with Benchmark 0.09
R-Squared 0.01
Beta 0.02
Alpha (%) (annualized) 4.29


I will subscribe if liquidity is good, and slippage / turnover are low. Is the max drawdown in 2018-19?

Max D/D was in 2004.

Liquidity is no concern. The model has 75% of its funds in static investment of 7 fixed interest ETFs : MUB, VWOB, VMBS, SHV, VTIP, SHY, and IEI. The weighted average duration is 3.61 years, that is less risky than IEI which has a duration of 4.44 years. These 7 ETFs are only periodically rebalanced, never traded.

The other 2 ETFs which are switched according to market climate are VIG and UUP.

With nodays>45 the model’s return is 5.76% and max D/D= -3.29%, so very little effort is required to attend to this model. Looks like a money making machine to me.

Now that I provided all the info, anybody can replicate this model. I would like to see similar models from others.

Forgive me if I’m wrong, but from a quick look at this model, it appears to diminish drastically in percentage return every year. A straight line like this means that it’s not compounding at all. It looks like you’re getting 50% returns in the early years and 4% returns in the later years. I never thought it would be possible to construct a model whose graph looks like this, with logarithmically diminishing returns.

Here is the calendar year performance. Benchmark is IEI.
Also growth of $5,000.