Bitcoin Forum

Economy => Trading Discussion => Topic started by: TradesLikeAPotato on August 19, 2013, 07:39:12 PM



Title: How to Download Backtesting Candles For Any Exchange
Post by: TradesLikeAPotato on August 19, 2013, 07:39:12 PM
Here's another method I have found to download candles for the Gekko backtester. This lets you get information for ANY time period or exchange.

1. Pick your time period and interval on bitcoincharts, for hourly this tops off at 5 months
2. Hit "load raw data" at the bottom of the chart
3. Highlight the beginning of your selection
4. Scroll to the bottom and shift click the end of your selection
5. Open excel and click and drag your selection into excel
6. Repeat 1-5 as necessary for your desired time period
7. Delete the volume and weighted price columns
8. Use this formula to convert to unix timecode:=(A2-DATE(1970,1,1))*86400
9. Click the bottom right corner of the formula square and drag to the end of your data
10. Copy unix timecodes
11. Right click, paste values only into timestamp column
12. Rename column timestamp to date (not sure if necessary?)
13. Save in CSV format
14. Run the backtester using your shiny new candles

Strangely, these candles are slightly different than whynothing's and nitrous', and lead to totally different backtest results. I encourage others to try this method and to provide input. I am inclined to trust these candles more than the calculated ones, since bitcoincharts is widely used.

Please don't hesitate to ask any questions if my instructions are unclear!


Title: Re: How to Download Backtesting Candles For Any Exchange
Post by: nitrous on August 20, 2013, 11:50:49 AM
Hi, for anyone following this thread, I have posted some reasons for the data discrepancies here (https://bitcointalk.org/index.php?topic=239815.msg2970826#msg2970826) and here (https://bitcointalk.org/index.php?topic=239815.msg2970857#msg2970857). In short: don't worry about the differences, all this data should still be useful to anyone who needs it.

Also, well done and thanks to TradesLikeAPotato for finding a new method for easily obtaining trade data (especially for all exchanges)!