Since the inception of Modern Trader, EidoSearch has offered readers a foundation to make more actionable forecasts. The platform harnesses data to forecast financial outcomes and return probabilities. EidoSearch searches hundreds of millions of patterns to identify historical similarities. By capturing similar data patterns, the platform generates projections of likely future outcomes. This platform enables investors to calculate risk through probability assessments better. (EidoSearch used historical data and trend analysis to help us identify the 50% expectation of 21,406.)
“How do we quantify uncertainty rather than just giving it lip service?” asks David Kedmey, President and co-founder of EidoSearch. “We’re good at explaining uncertainty when it comes to games of chance, but that’s in a world where you have perfect information. The real world of partial information and changing relationships [is like] playing blackjack with the dealer changing the composition of the deck. At one time, [the deck] has four aces, kings and queens, then all of a sudden face cards are doubled. And that’s what it’s like in markets when you go from what people would call a ‘risk on’ environment to ‘risk off,’” he says.
Advancements in machine learning and artificial intelligence will help improve analysis and performance at a time when hedge funds grow more competitive. “Hedge funds are in an arm’s race accumulating all kinds of [non-traditional] new data sets,” Kedmey says. “That’s where we can provide the biggest impact, by letting people use our technology in a more intimate way, where we become part of the overall system.”
The company is also helping to determine whether a data set that may have been discarded in the past provides meaningful insight. “We sit next to the raw data and calculate forecast distributions. We look for conditions that have really interesting distributions about the future so that they can get a sense of whether there exists predictive content in each data set,” he says.
While advanced traders are making this leap, traditional retail investors are lagging. They are still locked in a world where mainstream media continues to market single-point predictions and statistically improbable click bait. Kedmey says it is a matter of human behavior.
“It’s a deep, psychological reason that comes up again and again in behavioral psychology. That can take on different forms. But we’re not good at dealing with uncertainty. For example, there’s belief in the law of small numbers where you have 20 samples. If you’re dealing with small numbers, it’s very easy to pick a non-representative sample.”
Kedmey explains that the law of small numbers was an original test that Amos Tversky and Daniel Kahneman conducted on statisticians. The subjects of Michael Lewis’ “The Undoing Project” found that the so-called experts screwed up by relying on small samples to make generalizations. “And if [the statisticians] are going to screw up, everybody’s going to misinterpret. We think about what many money managers are doing all day. They’re looking for stories based on limited information to come to a conclusion,” Kedmey says.
That small-sample bias is the reason why investors chase money managers or forecasters with a limited track record. If a manager beats the markets two years in a row or hits two predictions in a row, many people might believe those so-called experts are better than the average. However, that small sample size doesn’t provide much predictive power, even though the financial media and Wall Street are full of people who made one right call and market their knowledge on those events. It’s the same concept that if a coin is flipped heads three consecutive times, the odds are still not better than 50% that it will be “heads” again.
These cognitive biases exist everywhere, and the concept of probabilistic thinking still escapes many people, particularly in the wake of the election of Donald Trump. “When we look back on what people said about Trump leading up to the election, [we tend] to label people. They were either right or wrong,” he says. “But that’s not true. People said things with a certain amount of confidence; some, way too much confidence. Some were a little savvier by giving probabilities. What if you said there’s a 20% chance Trump will win. And then Trump wins, were you wrong? No, you gave a probability there’s someone out of five chance that he would win. But looking back, everybody was either right or wrong. And that’s, kind of, related to single-point predictions; it’s how we judge people.”