Research objective is to identify the pattern when consistent intraday growth/decline leads to subsequent positive/negative returns and to develop a trading algorithm based on research findings.

Currently, I am trying to identify the pattern (if there is one) when consistent intraday growth can lead to positive subsequent return. There are 3 different return periods for analysis: gap (day 0 close price, day 1 open price), open to close (day 1 open price, day 1 close price), and close to close (day 0 close price, day 1 close price). (day 0 is event day)

I am using Apple, date range is from January, 2002 until November, 2016.

Briefly, what I did so far:
1. Regressed daily returns to get b-coefficient and its standard error (volatility). (minutely prices were resampled to 15min bins)
2. Calculated corresponding gap, open to close, and close to close returns.
3. Generated scatter plots (did not tell much).
4. Generated heatmaps.

If you take a look at heatmaps:
In case of gap returns, the lower right part looks interesting. If we take beta larger than 50% decile and volatility lower that 30% decile, there are positive returns in 13 out of 15 observations (~ 86.7%).
In case of open to close returns, beta’s lowest decile look promising. Especially, if we take the lowest 10% decile, the open to close returns are positive across all volatility levels except two (80%).
Finally, in case of close to close returns, the higher left corner draws attention. Consider the section beta lower than 50% decile and volatility higher than 40% decile. Close to close returns are positive in 18 observations out of 24 (75%).

Next steps will be to analyse specific deciles (or ranges) and add filters.

Will be happy to hear any ideas! (how to interpret and visualise results better, what filters might be helpful, etc.)

P.S. It is my first project using Python and Quantopian. Hence, most probably my code is not super-efficient, so I am open to any suggestions.

Being updated.