AI company CrowdSmart is looking to build a better mousetrap by seeking to quantify industry consensus surrounding investments. The crowd isn’t always so smart though.
What we call artificial intelligence certainly has its place, and the sheer processing speed and capacity of today’s computers can accomplish feats that would be far from possible otherwise. The human mind is far more sophisticated than these computers, as computers do extremely well with managing information that they are given, but don’t do so well trying to think on their own, outside the models that they are given.
Humans provide the qualitative element in this, and for example, computers can process the ideas that we give them, but the quality of their output will depend on the quality of their inputs, where, as they say, if we have garbage going into it, we will have garbage coming out of it.
Investment analysis is first and foremost a qualitative task, even though few realize how much, those who think that qualitative approaches to this are fairly standardized, the way we typically look at investment potential and risk, and have settled for a mundane quality or worse without even realizing it.
Almost all of our potential to improve the average results that we get does not come from doing the same things faster, it instead comes from doing better things first and foremost, approaches that better capture and understand what moves asset prices and the implications of variance that we call risk management.
While artificial intelligence can certainly assist us with this project, we need the better approaches first before we can leverage the power of computing to make these tasks more productive. Any revolutions therefore will not come by asking more of the same people what they think and then look to quantify that, as CrowdSmart is looking to do, it will need to come from our leveraging the power of our own minds first and then looking to accelerate the better ideas that emerge.
If you don’t know a whole lot about what is going on, and assume that those considered to be experts do know what’s going on, the idea of quantifying a consensus of these experts can seem a worthwhile task indeed. CrowdSmart is excited about such a tool, and many others probably will be as well, but we need to take a step back to examine the real limitations of this idea.
The problem with this isn’t the idea itself, which would be a good one if it can be assumed that the inputs to this system are good ones, but that’s where this particular idea breaks down, and it goes to the inferior approach and understanding that their affirmed experts are burdened with.
We would even go as far as to say that all this serves to do is to try to put lipstick on a pig but you have to know it’s a pig first and not something that is actually as attractive as a lot of people think. This approach could indeed be used to put makeup on much better-looking models, the ones that are already pretty attractive and could be made to look even better, but we need to use this make-up more discriminately than this.
The industry itself doesn’t recognize this as a pig because they live on the same farm comprised of pigs of various qualities and don’t even know what good looking livestock look like. There are other farms out there, better ones, but they haven’t visited them much and don’t really understand them, or why their farm is so inferior, so a consensus of pigs won’t count for much.
This idea of using artificial intelligence to quantify consensus therefore really isn’t all that terrible, although you can’t really do this that well considering you have ideas of various qualities and you just end up conflating them, dumbing down the better ones in favor of the mean. We are prone to do this under the best of implementations, and this would only convey a benefit if we have no idea how to decide between these ideas and choose to average them out instead. This is not a particularly good approach to add quality to our decisions though, which is what we need.
The way we do it now at least seeks to distinguish among this advice, where we can use quantitative data to measure it, looking at the performance records of the various actors and using that to guide our decisions. Even if we confine our attention to the pig farm, we’ll at least be trying to seek the cleanest looking among them rather than the average amount of mud they are covered with.
We do need to distinguish between these pig farms and other types, and the defining condition is that pig farms focus on a limited and incomplete view of the behavior of stocks, which by way of its very nature will serve to misguide, no matter how good of a pig farmer you may be. Once we understand this well enough, we will readily see how inferior this AI approach actually is, although CrowdSmart won’t, because they think the crowd is a lot smarter than it actually is.
This sort of idea stimulates excitement in the financial media overall, because they don’t know what is broken on pig farms, and without getting that, this can seem like a fine idea indeed. This discussion can also serve to be of benefit to readers who may not be that familiar with the mistakes that are made on these farms where they at least can gain a better understanding of how we should be approaching our own farming.
It’s hard not to be unkind to approaches that make such blatant mistakes and do not have any idea that they are doing so, and especially when we try to use this garbage as in input to anything. Mistakes are still mistakes when they occur by consensus, which can only enlighten us to the degree that they are made, not serve as something useful.
We can sum up what is wrong with this approach pretty simply, by just sharing what CrowdSmart CEO Kim Polese sees the ultimate goal of this being, which is being able to predict a company’s potential success might be. Wait a minute though, are we talking about companies or their stocks? Mixing these up is the big mistake here.
It is not that we do not want to do any sort of analyses of these companies, what is called fundamental analysis, because these things do matter as far as what we may expect the companies to be facing. We’re not buying the companies though like we would if we were buying private equity, we’re buying public stock, and buying public stock is a whole lot different game, they just don’t realize it.
Even though this fundamental analysis that they limit themselves to only paints part of the picture, even this tends to be messed up a lot, and we’ll start by how this happens and then move on to the other things that these experts should be looking at but do not.
Let’s Look at How This Idea Could Be Made to Work
What we’re going to do here is pretend that CrowdSmart asked us how they could actually use their idea of using quantification to improve stock predictions in a way that actually would be useful. We’d be happy to do this, since investing is not a zero-sum game, and it’s actually better to see other participants improve their investing because this makes movements more sustained and predictable and even traders would benefit from such a thing.
The first thing we need is to ensure that the fundamental data that we are putting into our AI system is relevant and on point enough. We don’t write about fundamental analysis all that much on here, mostly because it is so badly used, but this does not mean that good fundamental analysis does not have its place and we rely on it as well when we look at the future outlook of stocks for the benefit of those who want to invest longer-term.
The first thing that we need to make clear is that the outlook needs to be clearly defined, and this presents a challenge since investors have different time frames and expectations. The further out we go, the more difficult the future will be to predict with reasonable accuracy, and we don’t want to look so far out that we’re only guessing here.
There’s also the practical matter of being able to decide this along the way, and while investors aren’t going to be checking their positions daily, nor should they, we need to not only set a course which we expect a stock to follow, we also need to provide road marks where the accuracy of our predictions can be measured.
In other words, we don’t want to tell you that Apple will be at $5000 a share 10 years from now, as 10 years is too long and we don’t want people holding it for years when it has failed to hit its marks, when a company called Orange takes over the market and the apples fall on the ground and rot, and people are still expecting our predictions to come true. It is not just a nicety to provide intermediate term predictions alongside the longer one, as we need to make sure that the foreseen path is being followed or not.
A lot of these predictions are based upon time frames far too short to be of much use to investors, who just don’t want to know where Apple will be 6 months or even a year from now, they want to be able to see the future better. Trading on a 6 month to 1-year time frame doesn’t require any of this, as we can just trade the charts, for instance seeing Apple continue to go up and just ride it until it really shows that it is tired. Perhaps we need to jump off here and there but fundamentals only matter to the road far beyond the horizon, and is an inferior way to navigate on the road that this time frame seeks to do.
It is even more important that the fundamentals focus on the future exclusively, and this is the biggest mistake by far that these analysts make. Focusing too much on the present and the near term, and often exclusively, will in itself render their analysis useless and actually quite harmful, and our analysts need to understand this, although this would require some serious retraining.
If we’re spending our time poring over current and upcoming results that we already have, we are not only wasting our time, we are using this data to blur our vision. None of this data is relevant, as all of this has already been accounted for in the stock price, what the company has done, what they are doing now, and what they will be doing in the near future.
What happens when we ignore this is that we look to go off the highway that the stock is on and head off in the woods, where we see stocks on a different path than they are on by looking an incomplete picture of why we are here, and not even accounting for what we are trying to do, figure out where we are going from here, not where things are now, which we can just get from today’s prices.
An example of this would be an analyst going deep into the company data to discover what they think a company should be valued at now, come up with a completely different figure than what the market values it at based upon everything we know, and then try to substitute their views with what is actually happening. Stocks cannot be worth more than their market price right now, and that idea is actually a stupid one because stocks are worth what people are willing to pay for them, not a cent more or less, and certainly not what an observer supposes they should be paying.
Anyone who doubts this needs to try to sell their stock for what they personally think they are worth, although if they undervalue them, they will have no trouble. They misvalue stocks by definition, insist they are right, and only end up right later by accident, if the market chooses to go their way which often require an abrupt change of circumstances as they start out on the wrong side.
The future value of a stock is another matter though, but just like we need to completely delete from the hard drive of our brains the idea that present value could be accurately portrayed any differently than by market price, we also need to realize that this value as represented by a stock’s price encompasses a stock’s present perceived future value as well.
This turns predicting the future into an entirely different task than fundamentals lend themselves to, but fundamental analysis can still add to this, because this future value does not get added in by the market in its entirety, but instead is a process where confidence increases over time. We expect Amazon to continue to grow a lot over the coming years, which has us valuing the stock at 4 times the multiple of current earnings of the average stock, but as they grow and this belief gets better confirmed, as the confidence in this builds, the price keeps rising even though they may only be hitting their marks.
When we try to predict where Amazon’s stock may be in 5 years, for instance, we cannot allow ourselves to be confused about this being about where Amazon’s business will be in 5 years, it’s about where the outlook for the 5 years after that for the stock will be then. This is far from the same thing and something we will never be able to do if we just think that we are analyzing companies and not stocks, which are definitely not the same thing.
The fundamental part of this will still involve us looking at the future expectations of the company, its markets, and its expected market share. These are not areas that it is even possible for us to estimate with any great accuracy as there are too many variables, but a good idea sure beats one not so good and that is the task here.
We could look at Ford and GM compared to Tesla and wonder who will be able to better manage to keep their growth outlook with the way this market will be changing in the coming years, and the decision has thus far come down well against these old fashioned car makers and well in the favor of more innovative car makers. If you don’t understand why it actually makes sense for Tesla to be the biggest car maker in the world now by market cap, you might know a something about car companies but you are completely mixed up about stocks.
This Might be a Fantasy For Them, But it Doesn’t Have to Be One for Us
We already are at a point where we’re going to require that we completely reprogram the minds of fundamental analysts, and this is a task that enters the world of fantasy. While the understanding of this industry does evolve over time, at the pace we’re on now, it could take centuries for us to be able to look at the consensus of opinion out there and expect it to be valid enough to be of use. We’re far from finished yet though with what we need to change.
Stock prices don’t just move by way of future company expectations, as there are other forces involved and some pretty significant ones at that, macroeconomic ones. This involves more than just looking at the current horizon, as what we know now which is to a large degree priced in, like everything we know about anything affecting stocks.
Just like the market discounts the future when it comes to the potential for growth in a company, where it adds or subtracts future value but at the same time discounts it, as these future expectations manifest, they will discount them less and less, we can do the same with macroeconomic expectations as well. If the market thinks that the sun will shine on the economy next year, they will add to stock prices accordingly, but when they see the sun, they will add in even more because probabilities have improved, from likely to it’s now happening, as well as more likely continue to happen.
This is the real reason why it’s so effective and profitable to follow trends with stock prices, because the trend is a manifestation of a journey that is seen as more and more positive as it actually unfolds, and we can invest far better than the blind manner that fundamental analysts use just by following trends and nothing else.
However, our task here is to get fundamentals to chip in, to help clarify the future by not examining the present but actually seeking insight at a time out of view, and this includes doing this with macroeconomics as well. This goes beyond just economic forecasts or even geopolitical expectations, as we need to account for expected changes in investor behavior as well, as the more money that people have in the stock market, the higher stock prices will be.
This part is all at the macro level, and the macro level can be extremely influential to stocks, and to the extent that it will dunk the heads of even the best stocks out there when it gets angry. It will also boost the price of stocks in general, and we often refer to macro market forces as the level of the ocean, where some boats may tower over others but they all rise and fall with the ocean.
A stock’s price accounts for the present view of both micro and macro factors, where the fundamentalist just looks at the micro and ignores the macro. This renders their analysis incomplete at best and a horribly incomplete when you factor everything else that they miss.
Once we get all of that figured out, our job is far from finished, as we need to account for the area between the company and the stock, which we could call the degree of market favor. If we had to rely on only one thing, this would be the one, as this is the engine of our boat.
Boats can go faster or slower depending on the size of its engine room, and this is also the one that will keep us honest, the marks that will define how well we have predicted the future based upon these future fundamentals of both the market and the economy, the horse’s mouth itself actually.
A good way to look at this is to see fundamentals as the gas and investor behavior as the engine, and we want to see how far our boat will travel on a similar amount of gas. If we look at two companies and we want to use earnings as the gas, we want to see how far each will go on a 5% increase in earnings, and this reveals the power of this engine.
While quantifying analyst outlooks of the future is a daunting task that may only be able to be accomplished through artificial intelligence, quantifying the effect of the mood of the market towards things can be quantified with just natural intelligence, if we have enough of it to recognize this that is.
We weren’t joking during all the times we have pointed out that you could turn the idea of valuation using price to earnings completely on its head where high multiples are bullish and low ones are bearish, as this ratio measures the power of the engine behind stock prices beautifully.
We don’t just want to take snapshots of these though, as this is also all about trends, as a stock may have an above average P/E but it also may be trending down. The stock may be able to stand some of this and still outperform the market, but ideally, P/E should be both strong and trending upward or at least be stable enough. Like with prices, upward trends are just better.
This concept would appear to be bizarre to fundamental analysts, but that’s because they do not understand that stocks are actually a popularity contest. P/E multiples not only reveal how popular stocks are, they also show us how bright or dim the future of the stock is seen, where brighter is just better and dimmer is nowhere you want to be.
P/E is a very course indicator though and we could do a lot more to better quantify and elucidate this effect, and this is what we would be doing with the final step in our AI stock predicting project, the final piece on the puzzle but also the central one which always has the final say. This is the part that we’re looking to get help with when we look to down the road visions of fundamentals, looking to see where this mood will probably look like then, but this step is powerful enough in itself to make this the burger and the bun, with the fundamentals at best being the garnish.
We’ve completely lost ourselves in the world of science fiction now, where this is so far out there that we may never see anything like this happen ever, but science fiction can be pretty entertaining as well as educational. Just having a better idea of what the world really looks like when you take off your kaleidoscope glasses is pretty educational in itself, although we expect fundamental analysts will keep them on tightly, and this is no crowd that we want to go with a consensus. We’re allowed to look for ourselves though.