Disliked{quote} Hi knkie Each candle is composed from data during London session like this {image} Table from the prev post uses only this data, ignoring what happened in between London sessions. {quote} In some sense, testing only on one dataset - i.e. historical data - is already overfitting. I know 2 ways dealing with it: - randomly sampling entries and run MC simulation on equity curves (bc not every trade can be taken and so on) - generate as many data as you want and test 'strategy' or whatever on it {quote} +++Ignored
This is an interesting hypothesis.
Have to think on the synthetic data option ....

in wondering about yourself, you forgot what you came here to be part of
2