Price prediction using historical trajectories

Was curious what it would look like if I simply scanned through all of history, look for price movements within the same 60-day window that are similar to current movements, and then see what happened to them 30 days later. The results are above.

Each line is a 90-day price trajectory. I used the first 60 days to fit the pattern to current price movements, and selected the top 14 where the patterns matched the closest. The decimal in the parenthesis is the range of the movement, denominated by the starting price (on day 0). The patterns are filtered such that only ranges of >15% are selected.

Some current caveats: is not a quant study - nothing very rigorous about this nor will I place any bets on it. Another caveat: I simply normalized the frame by min and max values - and simply filtered them to be > 15%, judge for yourselves how representative each move is with respect to the current one (35%).


Leave a Reply