I did. Kalman filter on price data, plus constant analysis of lookback periods. They also measured the strength of the trend through cross-sectional analysis, volatility in trend, and volume profile analysis.
Can a slope which exceeds a value be considered a pattern?
- Thread starter kmiklas
- Start date
What were the results of their efforts?
Initially not that great. Took them about 3 years of research to finally launch a profitable strategy. I believe they have moved into parts of the HFT business though the bulk of their trading is diversified momentum.
Yes it was their primary project. 7 people. About $100MM to fund it. Trading profitably is a lot harder than most people think.
They started by proving the existence of momentum and then had to prove that they could create a system that could harvest it. Once they developed a strategy that worked, they spent 6 months on improving its risk characteristics before applying live money. They trade every liquid stock now (and added currencies & commodities).
As part of their research they ended up disproving 99% of strategies most people think work (they proved that they actually don’t work). I was actually surprised by how many of the “smart” strategies don’t work (or no longer do). When a profitable trading strategy becomes publicly known, its returns go down to zero.
Indeed. It's reversion to the mean. Here is the frequency plot for daily close price change %. They just hover around 0 with occasional shift to the left and right.
Certainly slope can provide a signal. Increasing slope suggests increasing strength in a continuing move or trend - decreasing slope suggests the opposite. It would have to be properly interpreted within the context of the price action, of course, and the trick would be to determine the most effective threshold for the signal. Then like any other signal you'd need a plan to take advantage of the opportunity while guarding against excessive loss.