![](https://static.wixstatic.com/media/dfb9ac_4c7bbe9632e544c3a2f757909d894068~mv2.png/v1/fill/w_980,h_980,al_c,q_90,usm_0.66_1.00_0.01,enc_auto/dfb9ac_4c7bbe9632e544c3a2f757909d894068~mv2.png)
Back in December, JPMorgan predicted that US stocks would gain 5% this year, economists were expecting the 10-year US bond yield to stick around 2%, and Goldman Sachs raised the prospect that bitcoin would hit $100,000. But six months later, US stocks are down 20%, the 10-year yield is at 3%, and bitcoin more than halved to around $21,000. The truth is, the pros on Wall Street have a terrible forecasting record, and all you need to do better is to follow five simple steps inspired by Philip Tetlock’s book Superforecasting.
Step 1: Define what you’re trying to forecast
Imagine you came across a headline that says: “Markets are significantly overvalued and are about to crash”.
First things first, you need to understand what “markets are about to crash” actually means. Which markets are we talking about? What percentage loss counts as a “crash”, and over what period of time? To assess the accuracy of this statement, we first need to define what exactly we’re predicting. So in this case, you might decide to rephrase the problem as: “What’s the probability that the S&P 500 will be down more than 10% in one year’s time?”
Step 2: Break the problem down
To work out the probability of this problem, you need to recognize that it’s made up of a couple of parts. By simplifying the problem into smaller, more manageable bits, we can pretty quickly formulate a surprisingly accurate forecast.
There are two parts to the original statement: “markets are significantly overvalued and are about to crash”. So there are actually two elements you could try and analyze: the probability that stocks are “about to crash”, and the probability that stocks crash when you know they’re overvalued.
Step 3: Strike the right balance between inside and outside views
Tetlock discovered that superforecasters see things in two ways, which are known as “inside” views and “outside” views.
They start by posing an outside view question: that is, they aim to remove emotions and look at cold hard data. They want to gauge how often outcomes of this sort happen in situations of this sort by looking at the facts. Going back to our market example, we could look at how often (in percentage terms) the S&P 500 has lost more than 10% over a one-year period: only 15% of the time since 1996.
The trouble is, this case ignores important information that makes this situation somehow unique. A unique factor, in this instance, might be how high valuations are right now. If you take those valuations into account by looking at the relationship between forward P/E ratios and subsequent one-year returns (left-hand side below), the probability of a 10% loss increases to 30%.
Forward P/E ratios and subsequent returns of the S&P 500 Index. Source: JPMorgan
It does bear pointing out, however, that the relationship is quite weak and there’s an almost equal amount of times stocks ended up much higher. So an estimated range for the “outside view” probability is actually around 15% to 30%.
Now, of course, no situation can be fully summarized by a number. Looking at the potential impact on the economy if Russia pulls the plug on European gas, or if the US Federal Reserve changes its tone, requires judgment. This part of the analysis – using your judgment to assess the specifics of the particular case – is what Tetlock calls the “inside view”.
But there are two problems with the inside view. First, it’s prone to biases, since it depends on the main market narrative and is driven by factors like the dominant news stories. Second, our brain is naturally biased to the inside view: it’s often filled with engaging details and easier to make a good story out of.
To combat those biases, Tetlock explains that it’s important to use your “outside view” as your main anchor and use your “inside view” to adjust it. In our example of how likely it is that stocks will crash, the 15% to 30% probability is our outside view, so even if our inside view assessment is quite bearish, it’s unlikely we’re going to give an emotionally driven overestimate. The probability, then, could be 30% or 40%, but it’s unlikely to be as high as 80% or 90%, as the original news headline suggested.
One of the reasons market commentators have predicted so many crashes that haven’t materialized is that they’ve overweighted their emotionally-driven inside view over the cold, hard outside view. Using the outside view as an anchor would’ve fixed that.
Step 4: Update your forecast frequently
Tetlock found that superforecasters update their forecasts much more frequently than regular forecasters. In other words, when the facts change, so should your forecast.
For example, whatever your original probability was regarding the likelihood of a recession across Europe, it should be adjusted now that we know that Russia could pull the plug on European gas and trigger a downturn in the region.
The best forecasters aren‘t married to their view: they’re willing to significantly change their forecasts even when it means contradicting their own previous thesis. I think this is what makes investors like George Soros and Ray Dalio so successful: they constantly attempt to understand where they could be wrong and they focus more on making money than being right.
Step 5: Learn from your mistakes
If you had to guess the one quality that Tetlock says makes a superforecaster, what would it be? Intelligence? Nope. He found one quality roughly three times as powerful as that: a relentless commitment to updating your beliefs and focusing on self-improvement. That means spending time analyzing your failures and successes alike, focusing on what you got right (or wrong), and what you could’ve done differently.
So if you’re keeping score of all your trades and constantly trying to figure out what you could have done better, congratulations – you’re already on the right track.
Comments