Ranking people by their recommendation performance is going to be very misleading for most people. There is this natural intuition to look at the top performers and try to learn from them. In reality, the rankings would look almost identical if everyone just picked everything randomly. Mutual funds and hedge funds exploit this all the time. They start lots of funds that all invest a little differently. They cull the losers and keep the ones that happened to do well. Then they sell prospective investors on their past performance of one particular fund when in reality they didn't need anything more than random chance and multiple tries to get it.
Thanks, this was my immediate thought. If Bogle's thesis is correct (which is perhaps equivalent to the efficient market hypothesis), then there should not be anything to be gained by looking at these folks' stock picks in general.
Retrospectively, of course, some of their picks should look like pure gold.
Agreed. I could register accounts for 1,000 five year old children and let them pick stocks, and one of them would have what appears to be great insight and recommendations. Does that mean we should follow the child's next move? I hope not.
I'd like to add that difficulty in measuring independence of recommendations makes the task of trying to estimate the likelihood of a particular result for a person or fund being a real out performance extremely difficult. If independence is underestimated (or not accounted for at all), the performance will look far more statistically significant than it actually is. There are other issues too, like non-gaussian return distributions, but I think estimating independence is the killer. Note that dependencies between securities are non-linear, eg increased co-movement of securities during times of high volatility. We're trying to model all this at http://backrecord.com We're very close to releasing something..
Notice that he wrote "features". That's plural, meaning more than one. And then notice that he wrote "such as validating phone numbers". The "such as" part indicates that phone number validation is just one example among several. I'm sure he's quite well aware that phone numbers alone are insufficient. That's why they would likely be combined with other approaches.
Curious to see how this will be different from Marketocracy http://en.wikipedia.org/wiki/Marketocracy which has been crowdsourcing public's opinion on stocks, then highlighting the best stock pickers, and then finally building a mutual fund based on the top stock pickers' recommendation.
Long-term Marketocracy funds are under-performing broad index funds FWIW.
I had an Econ professor who insisted individual stock picking is a fool's errand, and he had some pretty compelling evidence to back it up. If you have a thousand users picking stocks randomly, some will by chance do well. I wouldn't be comfortable putting money behind the lead users who are likely victors by luck. But then again, its very possible I'm missing something.
My finance professor suggests Buffett profits by being able to borrow cheaper. Sure, there's smart picks, but if you can widen the difference between the rate you're borrowing and your returns, your average returns increase. Here lies the unfair advantage - you can't compete with Buffett unless you can borrow at the rates Buffett can.
There's also the argument for diversification - investors holding large fractions of their portfolio in one security expose themselves to the operational risks of the company. By holding a broad variety of stocks, you'd expect their "lumps and bumps" to even out on average so you are only exposed to the overall market risk (i.e. hold airline stocks? Offset that with some oil stocks. And those with solar. etc.).
Nvest looks cool, but also seductively simple for the casual observer. It helps pick individual stocks, not necessarily build a balanced portfolio.
Buy/sells seem to be only time stamped to a day. Surely you need greater accuracy than that, e.g. the exact second? Otherwise, if I buy a stock on a day where it rises 10%, what price do I get?
Do you take dividends into account? If not, why not?
How do you rank people who make different numbers of picks? A naive method would be to add up all the returns, but this would then favour people who make more stock picks. For example, if I recommend two 'buys' that both give a 10% return, have I gained 20% in total or 10%? (As I would have had to split my bank to invest in both at once)
congrats on launching, yet another flashy site with an unoriginal idea. been done before (and better) elsewhere, yet you failed to tackle anything that your predecessors have failed at. Investing is not just about return; the other side of the coin is risk, and you fail to mention that anywhere.
investing is more than just a buy or sell. what about position sizing? what about position risk? portfolio risk? portfolio beta? how do you benchmark? what kind of drawdown do you incur? what's your sharpe ratio? why are you not adjusting returns for at least the market, and moreso common factor returns?
rank(total_return) != investing success
Your "metrics" should be educational and make people more aware of the financial decisions they are making. By boiling it down to buy/sell recommendations, you make investing into gambling with a 50% chance of being right.
your "transparent" ranking algorithm is not disclosed anywhere - do you have any documentation that your algorithm does more than just show who made the most "money" historically? (past performance is not indicative of future performance!) are your rankings stable? how do you identify persistence?
Hmmm... that would entail you going short when they go long. However, the potential payoffs from going short are asymmetric, with theoretical unlimited downside risk.
How would you deal with that, or would you try to execute through options?
Potentially Yes! You will be able to filter Nvestors and companies in a very powerful way. (Still in alpha development, but you can try it out after logging in: www.nvest.me/crossfilter)
Good luck with it guys. There was a site long ago called clearstation which was bought by Etrade and killed. It offered similar features (charting, profiles, following, tracking, discussion, etc.), they built a great community of traders which was educational and enjoyable to participate in. It's hard to build something good. Keep pushing and good luck.
Well done on the website. Navigation is smooth, and the layout is nice. I'm interested to see how this social network for stocks does once you get more users.
Are there really enough other places besides Seeking Alpha, Motley Fool's CAPS and StockTwits that publish people's recommendations in such volume that the number of recommendations would be large enough to be valuable outside of these already large established communities?
Or is this whole site's effort merely a feature that StockTwits and the other sites may add later on?
The platforms you mentioned don't track how well their users' stock recommendations perform. We want to change that and give you more insights about people's stock recommendations.
I'm not familiar with the others, but Motley Fool's CAPS certainly does... I.e. http://caps.fool.com/player/gibybo.aspx shows accuracy, score (this is % return after subtracting the index), average score (same thing, but per recommendation), plot of the score over time, and the performance of every recommendation.
I actually constructed some simple trading algorithms based on the CAPS data and back-tested them. I didn't find anything to exploit... didn't put that much work into it, but did enough to convince myself that following "nobodies" was probably not worth the effort. I moved on to trying to analyze professionals. Finding this to be mostly noise as well, however some of the outliers appear good enough that maybe their performance is not chance alone. It's difficult to analyze, see my other comment.
Just browse through the Motley Fool website, there are similarities for sure. Then I came across this, http://caps.fool.com/player/bbmaven.aspx. This person has a rating of 100, and for a period there, each of his recommendations have consistently made over 500%. I don't know how it is ranked, and I don't think it is correct either.
I mean there are many mathematical model to calculate risk. But, how do you objectively evaluate risk? We can definitely include multiple different models on the site.
Thanks for the feedback! We fixed the issue. You can now use hyphen or space in your name. Please try it again at www.nvest.me. If there's any other issue, please let me know! Thanks.
Will you pull existing recommendations from StockTwits and Seeking Alpha? Or will stock-pickers have to sign up on Nvest before you will track their recommendations?
Currently, they have to sign up before they can get tracked. There are challenges to track non-structured recommendations from other platforms. If you have suggestions, we would love to hear about them!
We have a system in place for quantifying quotes at http://backrecord.com . On the system that is live, all data has been entered by hand. The task is difficult enough that I've found even educated people who know about finance screw up the data entry more than I would like. So we have a sophisticated verification system in place to try and deal with that. We have also experimented with NLP algorithms with some success. If the data extraction is tailored to a particular source, it is quite doable in some cases.
Maybe import all of the posts, tickers and dates, and then get the user to manually go through and tag each one as buy/sell/hold, and enter the target price.
It's not perfect, but it might reduce the friction as you solve the chicken and egg problem.
Overall a fan and signed up. Some feedback: dated icon and layout has too much of a bootstrap feel. Also, single page application for more responsive UX.
