Skip to content


  1. Jason
    June 12, 2020 @ 8:11 am

    I am so glad I stumbled across this site while researching how to properly do backtests for my options trading strategies. Thank you!


      June 12, 2020 @ 3:56 pm

      Hi Jason – happy to hear it’s helpful.


  2. Erick C
    June 14, 2020 @ 3:47 pm

    I like the reduced risk by limiting the trading time to 8 hours/week. Good job


  3. Uri
    August 14, 2020 @ 4:29 am


    1) with these Strategies, where you close and open a new position intermediately, what will be the difference if you just buy and hold, without selling?

    2) in regulated brokers it’s not possible to buy bitcoin (only bitcoin futures, where I think that results will be similar to this strategy) and GBTC

    3) When you close a position, in this backtest, it can be a profit or loss, correct?

    4) it’s possible to back-test a longer holding period, like 1 week or 1 month? and GBTC?

    thank you very much and keep with your great research!


      August 17, 2020 @ 12:31 am

      Hey Uri —

      1) I’m not sure I understand the question. Perhaps restate differently?

      2) This is true, and agree the results should be similar.

      3) Correct.

      4) It is. Is the idea to stratify results by week of month or month of year?

      You’re welcome! Appreciate the kind words.


      • Uri
        August 17, 2020 @ 5:07 am

        Thanks for your answer!

        In question 1, I mean, that this strategy will be similar to big and never sell?

        what is the advantage of close a position and buy again intermediately?

        5) I think that this back-test show a good performance, because it include a big bull market in bitcoin,
        if you make this back-test from 2018 to 2020, I think that you will get very different results…


          August 18, 2020 @ 12:06 am

          Correct, should be similar to buy and never sell.

          There is no advantage if the intent is to buy/hold. It’s strictly an administrative activity to measure daily returns. For example, if someone wanted to “buy and hold” only on Tuesdays (granted, this isn’t an actual buy and hold), exiting and reopening the following day does nothing more than allow measuring of the previous day’s performance.

          5) It’s quite possible. The same thing happened on the AAPL equity study:

          The outperforming strategy started underperforming in 2016 onward.


  4. Michael
    October 25, 2020 @ 12:35 pm

    I do not believe you are able to trade BTCUSD commission free anywhere (with the exception of derivatives which obviously complicates everything) so do you think it is likely the backtest results are not very representative when assuming 0% commissions? It would be interesting to see the results assuming the fee structure of a major BTC fiat exchange like Coinbase Pro.



      October 25, 2020 @ 11:17 pm

      That’s a good point.

      According to Coinbase Pro the trade fee is a % of the notional traded – 0.5% for amounts <= $10k.

      That's a 1% drag every week! I didn't realize commissions are that high. I guess it's apparent that I don't trade BTC so my assumption is a poor one in this case 🙂

      The relative performance of the different strategies stands but implementation is impractical due to costs. I'll look into reproducing this study with the trade fees intact and will be sure to include them in other BTC studies.

      Good call!


  5. Uri
    November 6, 2020 @ 6:52 am

    Hi, how will be the results if instead of backtest, BTC, Long Day Trade, you do backtest, BTC, Long MONTH Trade, that means open a position the first of the month, and close it at the end of the month, and then open a new long position…


      November 6, 2020 @ 8:35 am

      Great question.

      Not sure on performance. However, it would be quite susceptible to timing luck. For example, we could open a position on the 1st and close on the 30th, or we could open on 15th and close on 14th. There are essentially 30 variations of this (ignoring Feb), one for each day of the month. Each will have a unique result.

      Since BTC only has about 6 years of trading history, that’s only 6 “January” trades. The observation count would be too low to derive defensible stats.


Leave a Reply

Your email address will not be published. Required fields are marked *