Smart Alpha Page Views Decreasing

Seems like some time ago, when Smart Alpha was R2G, page views were around 300 - 400, if I remember. A few days ago it dropped below 100 and it seems to be headed to 50. Any explanation for this?

We noticed. Lots of models were launched that had high turnover and small number of stocks and a significant monthly cost. These models will inevitably experience volatility. This volatility combined with the workload from high turnover, and the monthly cost, is A LOT to ask from an investor.

We’re working to fix one of those things: the workload. By automating the rebalance user will just be able to set it and forget it for a while. We’re also learning along to see what worked and what didn’t and will make future enhancements.

If you have used SmartAlpha let us know your feedback. Thank You

Hi Marco,

I noticed the same as MisterChang.

Do you really think that high-turnover models are to blame? High turnover will cause less subscribers to a specific model, but why is the overall interest in looking at Smart Alpha models decreasing? Why don’t you guys set up a poll and find out why users are less interested?

Possible reasons are:

  • Most smart alpha models have too short history to judge if they will outperform the market;
  • i can’t see the simulation to give me an idea of potential returns;
  • the strategy/idea behind most models is poorly documented and hence not very convincing;
  • there is a high degree of survivorship bias and few models have actually succeeded;
  • model designers are not having enough guidance to avoid curve-fitting and/or don’t spend enough time towards this topic when designing a strategy;
  • other investment platforms have a wider footprint in social media or on the www and thus attract more attention.

Cheers,
Florian

PS: besides Smart Alpha page views, the overall activity in the forum has decreased as well. This might be a purely seasonal factor, but still raises some concerns on overall interest.

I’m one of the people that is not visiting the page much anymore.

While I never had any intent to sign up, I used to like to see how the models are doing and perhaps get some ideas.

I just don’t enjoy navigating through it. I do not have any specific recommendations but I went there again today, didn’t change any of the settings and just left. For whatever reason I do not want to spend any time there–maybe just jealousy–but maybe the layout.

Anything I would say on this would just be meant to possibly help. Also, I have no complaints at all about how much P123 is helping my personal investing: actually praises.

Regards,

Jim

I think the reason are fairly obvious. The prime reason is the restructuring from R2G to Smart Alpha which meant de-emphasis of simulations. This was the proper way to go. The problem lies in that P123 didn’t (or hasn’t) improved the ability for designers to market their models to compensate for loss of views. This means HTML-embeddable performance graphs of SA models with affiliate links, straight to model signup with no detours, etc.

The second reason is the poor performance of value factors for the last couple of years. There is not much we can do about that but at least we can now offer strategies other than value without attempting to compete with ridiculous simulations.

The third reason is seasonal. Subscribership, at least for my models, has always dried up during the spring/summer. It tends to pick up in the fall.

Fourth, and this will be contentious. P123 needs to stop monkeying with the factors and needs to freeze the database. In the beginning (of R2G) I thought the factors had been vetted quite well. And when there was a change, P123 would introduce a new factor and deprecate the old. Well everything was great for about a year then there seemed to be a continuous stream of changes to factors and also the database. I could not get the same sim results from one week to the next. In my opinion, this is a very bad situation for developers and subscribers alike. If I had known that the as-designed models would not be respected, I likely wouldn’t have issued any models to begin with.

Steve

P123 is pushing high-turnover models down the default SA screener list by marking them as high “Trading Commitment” models. What’s high commitment? One of my favorite SA models has a 20% daily liquidity threshold of $5M and 10 trade/year and is rated a 4. Next to the most difficult to trade. I would rate it a 2 or maybe 3.

I think some good models are being buried by the screener, making them hard to find.

Walter

I think Florian’s idea of doing a survey of all current subscribers is good. the first question could be if the sub is (mostly) a developer of their own-use models or do they mostly use SAs. And then go from there.

I think all of the developers have their own opinions but it would be great to get feedback from those who mostly use SAs.

I stopped using other’s SA models primarily due to a lack of faith in the model itself (after using for 1-2 years). So I spent all of my time after that doing my own. I don’t have any comments about the traffic, I never really noticed it before.

I also agree with Walter about finding good models. Another suggestion is to play up in the Browse the total performance since inception and its consistency in beating its benchmark. It is hard to see that from the 3mo, 1yr and 2 yr numbers columns. You have to look at the pages individually to see it. If you just look at the 3mo, 1yr and 2 yr numbers, you can easily dismiss virtually all of them and just stop browsing.

As a subscriber, I have less interest in SmartAlpha because it is clear that many (most?) are over-optimized and I do not have confidence in them. There are few models with significant OOS data that have done well. I only look at models with 2+ years OOS data or those by a trusted author (Gerstein,etc.). Anyone can create an over-optimized model with great simulated performance (myself included) – but they should not be allowed in SmartAlpha. I have been focusing my time on creating my own models. I know they are not over-optimized and I know how they work. I would have greater interest in SmartAlpha models if I knew they were not over-optimized.

-Debbie

Definitely a big factor. And that’s reality. Newly launched mutual funds have the same problem. Newly established newsletters suffer from it. Ditto with newly established relationships with human investment advisors. And finally, the same holds true of one’s ability to evaluate one’;s own efforts. There’s no way around it. Some thing take time.

If we thought the simulation could give you and idea of potential returns, we’d still have it in Smart Alpha. The problem is that the appropriate use of simulation data was widely misunderstood and was leading subscribers toward the models with the least probability of satisfactory real-money outcomes.

Yes, yes, yes and yes some more.

I have often urged designers to do a better job of explaining their ideas (without resorting to boasts about simulations). I have also urged designers to speak in the forums in ways that make actual and potential model subscribers comfortable with who they are as humans and as investors (generally, apart from any specific models). This is critical. So far though, designers have not been doing this.

It is true that the success rate has not been good. That said . . . .

This is absolutely positively not true . . . at least not any more. You probably have seen occasional forum references to an on-line virtual strategy design seminar I’ve been doing (which is now completed in the main portion and I’m working on post-graduate material). The group (a private group housed within the p123 community) has 169 members and remains open to anyone.

The good news is that any designer who has gone through all the material has the capability of designing strategies with absolutely zero data mining and which have legitimately good possibilities of beating the market.

The bad news is that this is a lot of content and it may be quite new to many and that it takes time to read, master, and incorporate into strategies. Since it is a virtual seminar, I don;t know how much each of the 169 members has mastered yet and how much time it will take for those who are in progress (most I assume) to be ready to create new strategies. Perhaps now that the content creation phase is near an end for me, I can shit attention to figuring out how to get going with live conference calls (an idea I floated in the past but which fell by the wayside as my first implementation experiment bombed and as I pushed further with creation).

Tell me about it! :frowning: :frowning: :frowning: :frowning:

Actually, it’s more complex than that. Pretty much all financial media and related social media is built around the notion that a person has specific stocks in mind and wants to read, hear or gossip about them. That’s not really what p123 is about. P123 is about getting the right stocks to your attention whether for further research on your part (the classical approach) or for full-list investing. That is a major chasm. There is absolutely nothing out in the established or social media that is compatible with stock finding except if one carries forward a brand already established pre internet. So that leaves us with a problem. I have a great public platform (4 web site one of which is Forbes) and related social-media pushing and can pretty much put anything I want out there. But none of that guarantees an audience because the venues are not designed to channel viewers to content that is not stock specific. In other words, i can and do publish a sh**load but have to struggle to pull traffic. So I’ve bene doing a balancing act between publishing media-friendly content (single-stock features on stocks that come up in my models and which may be of interest to broad audiences) and general pulls to popular topics, such as a recent URI feature that has a "sharing economy: angle (which I exploited through headlines, hashtags, key words etc.). Strategic headlines, keywords etc. on purely p123-niche type topics is another thing. As is the case with new models, it takes time to build new brands in new areas. Fortunately, the guy who runs Forbes understands the significance of what we do and why reasons for looking at stocks aside form current events and portfolios are of greater interest than the hot stock, and we work together well. But he can’t snap his fingers and re-architct Forbes in the strategy direction because his job is to bring in the revenue.

I noticed that too. Seasonality is likely relevant. Then, too, there’s the market. A lot of forum discussion in the past has been about testing and strategy development methods and accomplishments and for a long time it correlated with successful real money results. Unfortunately, a lot of it was data mining and dealt with strategies focused on the lowest quality portion of the market . . . the portion which zoomed as excess liquidity ran out of better places to go. That game is over. Best case scenario is that interest rates-liquidity remain as is, and the market has become more willing to discount an eventual change toward tightening, which means the market is going more “risk off.” That’s not necessarily bad news for the many p123 subs who like small and micro cap stocks. Over the past year, my best personal performance has continued to come from my low-priced stock model (development of which was finalized in mid-2010 with everything since then being “out of sample”). But the way I go after thenamno-caps is different form what had been discussed in forum, so much so that had I laid bare my model-building techniques as I built it, I’d have been ripped pretty badly since everything I did ran afoul of just about everything that was being preached in the forum. But my approach, the approach being presented in the aforementioned virtual seminar, can still allow people to pursue the better end of the teeny-tiny niche.

Ultimately, I think we’re in a transition period along with the market. Topics that used to be big have lost their zing. And people haven’t yet had the time to get comfortable with and immersed in the what I think will become the flavor of the discussions in the next decade or so. But I do see the discussion moving in that direction, slowly so far. But it’s moving. And I really do believe more and better is coming as our platform can do far more than many have realized to date (lately, I’ve been edging away from general-size 4-week rebalancing models to larger 3-month models).

Hi Marc,

Thanks for your comments. You are certainly doing a great job in educating the users on how to improve stock selection. I was part of the strategy design class right from the start and enjoyed every chapter!

Cheers,
Florian

“Another suggestion is to play up in the Browse the total performance since inception and its consistency in beating its benchmark.”

David - that was the point behind SA… to take away the performance since inception (start of simulation).

Now there are things that should be corrected, like using benchmarks that are chosen by the designer (I think?). Benchmark should be chosen based on what the system is trading i.e LC / MC /SC / AllCap / Sector or industry / Market neutral. Then the performance tables should only be comparing systems that use the same benchmark. Otherwise we are comparing apples to oranges.

Marco’s comment on models with few stocks / high volatility -

I think it would be a mistake to de-emphasize models with a small number of stocks. If we do then every model will consist of at least 20 stocks and the subscribers will probably only commit to one model and success will ride on the performance of that model. SA will then die over time due to lack of subscribers.

A better way is to emphasize models with few holdings and substantially improve the book functionality. Thus the subscriber can subscribe to multiple models holding a small number of stocks, each targeting a unique segment of the market, or some unique strategy. The performance of an investor’s account is much more important than performance of any individual model. The issue of high price has to do with number of subscribers that can be supported and for that reason smallcap / microcap models should be discouraged.

Steve

Steve, I meant since the inception of out of sample, not simulation. You can only see that by looking at the Performance tab for ‘Annualized Return Since Launch’. It would be good to see this on the Browser page as a separate column of data.

Hi Steve,

I agree with your comments about using relevant benchmarks. Unfortunately, that’s a hard problem to solve. Many SA models are built in a way that relevant benchmarks are hard to find. What would I benchmark a equally-weighted, defensive sector restricted, small/medium/large cap, total return SA model against?

Walter

One way to do this is to use the performance of the custom universe as THE benchmark. Then SA models would be grouped such that the benchmarks are somewhat correlated to each other. The first part is easy. I’m not sure how the second part would be done but I’m sure a stats genius could figure it out.

Steve

Some of my universe are dynamic. So I could benchmark one of my models against it’s universe. But it wouldn’t help bench models from different developers unless, like you said, P123 can group somewhat similar models. I just end up looking at total return and ignoring things like alpha.

Walter

For benchmark, my guess is that most subscribers, as a first rough cut, would compare to either SPY or IWM, depending on the typical market cap in the P123 generated metrics and the description from the Developer. From a subscriber perspective, I would keep it simple. They can contact the Developer if they want to compare to other benchmarks.

Just a thought, P123 should roll out a separate web site for Smart Alpha. P123 should stay the same minus the Smart alpha portfolios. A separate Smart alpha web site can focus on advertising, give more control to designers with more detailed reports, let designers have their own page to help sell their product, give designers an option to build a newsletter type service, etc. If designers want to push their models give them a good platform that can utilize advertising, social media, and control in a well designed controlled environment.
The most successful designers on P123 all seem to have their own web site already to promote their models. All designers and P123 could benefit from a separate new Smart Alpha site.

Derek.

The problem is SA models shouldn’t be used alone and are best instantiated within books with other models.

I’m running a 10 model book now. Some of the models are SAs I subscribe to, others are my own SA models and then there are some private live ports. I don’t want P123 to stay the same minus the Smart alpha portfolios.

Best,
Walter

Burd4 / Walter - I believe the original intention was for R2G to remain internal to the existing P123 community and Smart Alpha was to be a different website geared for the mass market, high liquidity, low priced models, different forums. I’m not sure what happened to that, or if they are still planning on it.

Walter / David - For comparison purposes, several standard baselines could be chosen such as the S&P 500, Russell3000, value, growth, and sectors. Then the performance of the custom Universe for each model could be calculated and used as the model benchmark. The model would be grouped to the standard baseline based on highest benchmark-baseline correlation. The models grouped to a baseline would give users a way of viewing similar systems. The excess performance would be model performance minus model benchmark (custom universe performance).

An interesting book feature would be to provide combined performance for the underlying model benchmarks (i.e. custom universe performance). This gives the user a feeling as to how the combined performance of the models would be if the models’ performance matched their benchmarks in the absense of any kind of optimization. Unique custom universes (or dynamic custom universes for that matter) could flourish in this (book) environment as similar model benchmarks would not be desirable from a diversification perspective.

Steve

Who is the intended audience/user for SA? Is it retail or professional? My input has really been more from a retail user standpoint. I don’t have any experience with professionals. I could see where pros requirements could be different for benchmarking, books, etc. though.