One of the indicators we use in Deanna is based around studying the past.
For that we use a pattern matching algorithm powered by AI. The algorithm compares a recent window of price and finds all the past windows of the same size that present similar behaviors.
To compare the different price windows, we use a type of artificial intelligence algorithm called deep learning and in particular convolutional neural networks. The goal of this algorithm is to compress the price time series into something less noisy. This can be seen as a filtering step. Instead of directly comparing the price windows, we compare the filtered versions instead.
Looking at the past windows we then look if, in average, the market went bullish or bearish in market situations similar to right now. This allows us to provide a value (the indicator) telling how much return we can expect in the next 10 hours if the past repeats itself.
The advantage, compared to other techniques based on price patterns, like Chartism, is that each prediction is based on statistics on real historical data and not on some ad hoc rule forwhich we have no evidence of their usefulness for cryptos!