Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Candle Patterns - research strategy

On behalf of the author the attached research strategy (1 of 2) is offered.

In the Februrary 2015 issue of "Stocks & Commodities" magazine (www.traders.com) an article, "Candlesticks, Condensed", is presented outlining a novel means of candlestick identification. Quantopian provided an open and capable venue for which to share a python version of the calculation required to construct the code outlined in the article. This code is provided here to assist any of those interested in the testing and examination of the technique.

The author wishes to thank Quantopian for providing such a useful platform and hopes to offer future articles and technical market explorations on their platform. The attached strategies leverage code and techniques from the open community and the author wishes to thank all members of the Quantopian quantosphere for their willingness to share their insight and hard work.

This strategy is used to identify candlestick patterns and present the most favorable patterns at the end of the test run.
A secondary strategy is used (presented below) to trade these patterns.

As the February 2015 issue is, as of now (Jan, 4th), undergoing distribution, requests for comments related specifically to article content will have to be deferred until the issue has had a chance to be delivered.

3 responses

On behalf of the author the attached trading strategy (2 of 2) is offered.

This secondary strategy is used to trade the pattern identified in the above research strategy.

As the February 2015 issue is, as of now (Jan, 4th), undergoing distribution, requests for comments related specifically to article content will have to be deferred until the issue has had a chance to be delivered.

Hi, I have only two questions.... for now. What is the the assumed validity in terms of time of the patterns? So, if I identify a pattern that works for a specific security, how long can we trade on it before we have to run the discovery anew?

Wouldnt it be better to use schedule-functions to integrate the 2 strategies together and create a running discovery after an initial discovery on history data, where the trading will be using the discovered patterns? Then one can age the patterns in time and anticipate new market conditions....

looks impressive though...

PB

@Peter B., If we only had a sneak-o-scope to peer into the future eh? The viability of any pattern discovered today will last how long? Who knows. The idea of combining these two halves of this strategy into one is an excellent idea. Continuous optimization is always a bonus when you can perform it. Yes, perhaps using the schedule function to periodically take a large chunk of history and remake or append-to the candle returns collection, and then to use the collection's sorted extents (top N for longs, bottom N for shorts) as the candle patterns on which to trigger trades would be a worthy refactoring.