What’s the Deal With High/Low Forecasts?

Since finding an algorithm to predict high and low security prices, we assembled statisticians, programmers, mathematicians, data analysts and tech executives to develop and explore this relatively new “kid on the block”  Our background is primarily in information technology (IT), so the first thing we did was create applications in the cloud. Today, we offer more than 200 forecasts of high and low stock, exchange traded funds (etf’s), indexes, and futures over horizons of one to 30 days. We believe we are the only service to provide this scope of high and low price forecasts. We are focused on improving the algorithms and providing reliable to our customers.

For the record, predictions of high and low stock, currency and commodity prices are consistent with standard finance hypothesis such as “rational expectations.” Forecasts of daily high and low prices for the S&P 500 stock index can exhibit lower error by widely used metrics – such as the mean absolute percent error (MAPE) or mean square error (MSE) – than the “no change” forecast – a benchmark for random walk series (See IPD White Paper #1.5).

So you can forecast the stock market – at least features related to volatility or the swings in prices over time.

BIG QUESTION – can these forecasts provide forward guidance for stock or other asset prices? Here is where high and low (H/L) price forecasts, seldom discussed in the finance literature, mess with some of the sacred cows. The answer is “yes, under some circumstances, H/L stock price forecasts can indicate forward direction of a stock price series.”

There are various ways of showing this (see for example the charts in BIG WINNERS and BIG LOSERS pages), but, recently, we found H/L forecasts can connect directly to rates of return, calculated as percent changes in closing prices. Soon you will be able to download a short White Paper here on this, but the basic point relates to the magnitude of predicted growth in high or low prices. When the predicted 5-trading day growth of the S&P 500 high price, for example, is “large in magnitude,” the directional accuracy improves. So if the predicted 5-day growth of the S&P 500 high price is 3% or greater, there is a high probability the actual forward growth of the S& P 500 5-day high will be positive.

This sort of result is mathematically consistent with random walk theory, as is the basic ability to predict H/L prices over various periods of a series.

We have found an additional effect, however, which seems to link with “momentum” in stock prices in fairly strict analogy with that term in physics. This momentum almost surely involves trader behavior layered on top of the basic underlying mathematical facts.

Finally, we find that external events do introduce jumps in stock prices not strictly accounted for by previous price movements, and we are developing “sentiment analysis” to address these types of events. Again, more to come on that.

In the meanwhile, feel free to email us at service@infodynamicsinc.com for more direct communication.