273: TZ Discussion – Limitless

Justin and Jason discuss why Disco hasn’t worked out and what Justin has learned from the experience, balancing product development with sales and marketing, how to select and validate a business idea, the basics of algorithmic trading, how Jason increased his vertical jump by 8 inches, high and low adaptors, Justin’s desire for a walking desk, how to raise your IQ,Β the research in support of it,Β and Jason’s effort to keep the math team together.

  1. Hey guys, great episode. I’m interning for John Somnez and his podcast Get Up and Code. It’s a programming meets fitness podcast. Now seeing how Jason has been working on “Project Superhero” and his vertical jump, and that Justin is seeking a walking desk, I think it would be totally beneficial for you guys to appear as guests on his show!

    For example, John has an entire blog post on how to get set up with a treadmill desk! Here’s the link: http://simpleprogrammer.com/2014/10/07/treadmill-desk-walking/

    I think it would be great exposure for both podcasts and I think everyone will benefit from the discussion. Please let me know if you are interested and I will schedule something with John. I look forward to hearing from you!

    Keep up the great work guys!

  2. Jason says:

    @Doug Coleman – Sounds good! I already replied to your email. But just a heads up … Justin doesn’t know jack sh*t about fitness. πŸ˜‰

  3. Justin says:

    @Jason Thanks… πŸ˜‰

  4. Mohammed Firdaus says:

    Just me or did Justin sound like he was close to tears while talking about Disco…

  5. Jason says:

    @Mohammed Firdaus – Justin is an emotional guy for sure, but honestly I don’t remember noticing him tear up. πŸ˜‰ Besides, … there’s no crying in baseball!


  6. @justin, check out Quantopian. They offer a platform that allows you to write the trading logic without having to mess with the infrastructure.

  7. Jason says:

    @Micah Alcorn – Following is an email response to another listener who made the same suggestion:

    Yeah, I’ve mentioned Quantopian to Justin in the past, but as far as I know he’s never really looked into it. But you’re right that it’s really well done and a good place to start.

    One major advantage of trading Bitcoin on a site like BTC-e (https://btc-e.com) or Cryptsy (https://www.cryptsy.com) is that he can just make REST calls using PHP or Javascript, which are the languages he already knows. Another, of course, is that you can trade real money intraday without needing $30K in the account.

  8. Justin says:

    And here’s the answer I sent…

    I have been looking at it. I think it could be a good opportunity to learn a new language… Although I do suffer from the issue that I like to know all the code inside out and build from the ground up as an educational exercise. I also like the fact that with bitcoin you can use real money from the get go.

  9. @Jason, yeah I realized after posting that comment that I probably first heard about Quantopian on techzing. The show is my only source for all things worth sharing. πŸ™‚

  10. Alfie says:

    Justin: don’t forget collective2.com – you interviewed Matthew Klein a couple of years ago on the show πŸ™‚

  11. nethy says:

    This was a really great episode.

    Pretty big contrast between Justin’s idea of spending 6 months exclusively on the stuff he doesn’t like and the idea of a rebound project that includes none of the stuff he doesn’t like. He sounded so sad about the first one and you guys both lit up talking about at the rebound project (great name).

    Unsolicted suggestion

    I wonder if this trading stuff could is a potential entry point for learning to program. It doesn’t require a UI, maintainable software, many tools, frameworks or other other hair. All the focus is on “fun” stuff, for some definition of fun. You might be able to get to some really interesting AI-ish stuff pretty quickly.

    If it’s possible to go from 0 to implementing a script that buys a Satoshi a day sell it the next based on some very simple rule in 10 hours and from there to genetic algorithms, implementing trading strategies, cool stuff. Going straight to algorithmic crypto currency trading sounds fun and dangerous.

    The cool part is that if you put a small but sufficient amount of money into it, it becomes immediately interesting, a built in motivation device to keep programming and learning. I think there are a lot of people who want to learn programming, but don’t want to do it professionally. This could be a fun approach for some people, like building games.

    “Algorithmic Ccrypto Currency Trading for Amateurs” has a nice irreverent ring to it too.

    It could also work for kickstarter, which could be fun. Luck surface area-ish with all the people interested in bitcoin. Everything except a good chance at actually making much money, and who wants money in a rebound project anyway.

  12. Alfie says:

    Justin: Although Jason’s right in saying that Technical Analysis is more like astrology than astronomy, the following books should at least give you some insight to what techniques people are using:

    – “Technical Analysis” by Kirkpatrick
    – “Technical Analysis from A to Z” by Achelis

    These should give you a good feel for how to look at your data. Then, take a look at Jason once recommended to me to actually test your trading strategies:

    – “The Evaluation and Optimization of Trading Strategies” by Pardo

    After those, you’ll probably want to use books using real science (e.g. quant books for pricing models etc). However, those three should be enough to give you a landscape of what it’s all about and to test the waters with your trading strategies.

    Good luck!

  13. Thom says:

    I did this course a while ago and found it to be a really good introduction to trading: https://www.coursera.org/course/compinvesting1

    Re: wasting time on MVPs, one thing that I’ve always thought would be an interesting experiment (but never tried) is to outsource it via a freelancing site – paying a few thousand dollars and getting someone else to do all the grunt work, which you can then take ownership of or throw away and rebuild if it turns out to be successful.

  14. Sandy says:

    I loved to hear about Disco, just because I’m going through the same – worked really hard for several months just for the project to die on the sales side. Devs take all the risk when we start a project like this.

    I too, and not going to start a project unless (1) the validation is rock solid (2) the sales cofounder takes some of the risk (I’m thinking maybe paying me for the dev time, and then the business pays him back when it makes money).

  15. Ahoy guys,

    Great episode. I always enjoy them when you get talking about side projects and businesses. Re trying to validate before building I went through a similar revelation last year. I spent a bunch of time doing customer interviews and research before I wrote a line of code on a new project. I then got the madness for a different idea and burnt an entire Sunday building it (http://hulimail.com). I pulled back from that and went back to doing some research after telling myself for rushing off and building before validating again. Anyway, that site is sort of ticking over OK now, just launched recently and I’m now getting better at validating before building features.

    However, it is also true that validating an idea is still hard without building it. People will give you an email address or say it’s a great idea but when you give them the opportunity to buy they don’t put their money where their mouth is.

    One good book I read on doing customer interviews is “The Mom test” by Rob Fitz, the idea of the book being to structure conversations in a way which means people can’t lie about how they would behave. Worth a read before validating your next idea. I’ve also just bought Dan Norris’ “The 7 day startup”, although I’ve not read it yet the central tenet of the book seems to be “You don’t learn until you launch”. He seems to be more about launching, then validating from there or switching to a different idea. This reminds me of advice I’ve heard YC used to give to founders to launch something as soon as possible and then work with the audience they got to discover what they needed to build.

    So, a couple different ways of looking at it but neither proposing spending a long time building before getting feedback which is I think the key thing to change.


  16. Juha says:

    It was interesting to hear you guys talk about the n-back game. Lately I have been coding more frontend JS
    and felt I am doing less hard problem solving so decided I needed something to fix that. I started doing sudoku’s to train my patience, concentration and problem-solving, but then stumbled onto the n-back game. After few weeks I highly recommend that, it kind of feels like going to gym for the brain and helps to concentrate on task for longer time.

    @Justin About the exercise: I think you are missing the best part of exercising if you just do short or low-intensity stuff. Your body starts to release “feel good” hormones after several minutes of proper exercise, this is partially the stuff that helps you to ignore the pain and relieve stress. I would recommend starting to go to gym to do 15min of semi-hard cardio minimum, and most of the time, you’ll feel good enough to actually push it to 30min (same as when Jason said he just wants to keep pushing when he is exercising). After making it a habit your mind learns to anticipate the good feeling already when you are packing to go to the gym! I hope you’ll succeed in your push for getting more fit πŸ™‚

  17. Jason says:

    @Other Robin – I REALLY like the 7-Day startup concept and I think that for someone like Justin it could really hit the sweet spot between validating the business and just building something.

    Here’s a link to an interview with Dan Norris about his experience:


    And here’s a link to the book:


    Thanks for the suggestion!

  18. Jason says:

    @Juha – Interesting … I have a few questions for you in regards to n-back if you don’t mind sharing.

    1. What software are you using?

    2. Is it single or dual n-back?

    3. How many times per week do you use it and how long are your training sessions?

    4. How long have you been training?

    5. Have you experimented with varying your frequency or volume of training and if so, what have been the differences in the results? According to the research it appears that the results tend to be dose dependent and at least on the scale of training for 4 or 5 weeks, the relationship has been largely linear.

    6. How much have you improved your working memory overall?

    7. Have you taken any time off from the training and if so have you detected any decay in your working memory? My guess is that it’s similar to the decay that occurs after taking time off from weight-training in that basically it takes about 1/3 the amount of work to maintain than it does to build.

    8. Did you pre and post-test your IQ to detect any change? If not and you’re interested, here’s the Raven’s Progressive Matrices test, which I believe is the most heavily g-loaded of the IQ tests – http://www.raventest.net. There might be better options out there, but this one’s free. πŸ˜‰

  19. Justin says:


    – Ok. So, I might have a little bit of madness about the trading thing, so that’s a good sign πŸ˜‰
    – Amy Hoy put a little thing together in 24 hours called Just Fucking Ship – https://unicornfree.com/just-fucking-ship
    – I bought 7 day startup and did think it was very good. The main point I noticed is that it seems to work very well for semi-consulting gigs (like wp curve) but not so good for anything that actually takes any time to code.

    @Juha – I know, I know. But let me be deluded won’t ya πŸ˜‰

  20. I’m doing that crypto trading stuff and it’s a lot of fun, as long as you don’t stress out about loosing some money πŸ™‚ I started in April as a total noob in trading, wrote a very basic trading bot that didn’t look at anything except price movements. Lost a bunch of money until I figured out the whole “shorting” thing πŸ™‚

    I was playing with my bot and improving it little by little ever since. It’s not as sophisticated as what Justin wants to do, no neural nets or genetic algorithms or anything like that. It just looks at several parameters like overall price trend, volume, price velocity etc and decides how much to sell/buy. Trades bitcoin and few dozen other coins on different exchanges. It’s a fun hobby and I’ve been spending too much time playing with it instead of doing something productive πŸ™‚

    I would suggest starting with bitcoin only and use an exchange that allows “shorting” like Bitfinex or OKCoin. Their APIs are really easy to work with. You kinda need a high-frequency trading bot (or at least what passes for high-frequency in crypto world) because bitcoin price can jump 10% in a matter of hours. Other coins can loose/gain 500% in a day, so trading once a day won’t really work. If you have any questions please feel free to send me an email. But I’m a complete noob and haven’t read even a single book on the subject. It’s too much fun figuring it out myself πŸ™‚

  21. Jason says:

    @Michael Rakita – Very cool! If we decide to get this little competition going, you’re definitely going to have to participate.

    So, tell us a little more if you don’t mind?

    1. How many different trading strategies do you have currently deployed?

    2. How many trades do they generate per day?

    3. Have you found anything that works, and if so, what are it’s performance metrics?

    4. What’s your research protocol? In other words, once you have an idea, do you backtest it, do walk-forward tests, do Monte Carlo simulations, paper trade?

    5. What kind of data are you using for research – bar data, tick data, order book data and how far back do you go for testing – 1 years, 3 months, 1 month?

  22. @Jason – As I said, I’m trying to figure out everything myself so I’m not sure about the terminology.

    1. I usually have one strategy per currency pair that does real trades and one or two that I test with simulated trades. Right now I have 2 different algorithms doing real trades on bitcoin/USD and one that I’m testing. I almost gave up on all altcoins except litecoin. Unfortunately there is no way to short any coin except bitcoin and litecoin, and most altcoins just go down with very rare upward spikes.

    2. Each algorithm is allowed to make one trade per minute per currency. Sometimes it doesn’t make any trades for days when volatility is low. And sometimes it trades every minute for hours. On average my bot makes about 100 trades per day across all currencies.

    3. Trend following definitely works for bitcoin and litecoin. One of the first algorithms I tried just follows a trend and tries to calculate probability of trend reversal based on price velocity and volume. Even when it sells or buys too early or too late it still makes money in the long term because price trends/movements are huge and easily recognizable.

    Right now the algorithm thinks that bitcoin is in a downtrend but the probability of trend reversal is increasing, so it started to reduce its bitcoin short position.

    For altcoins I’m trying a lot of different things but so far all my algorithms were losing money. An algorithm might make 500% on 2/3 altcoins but lose almost everything on other 10 altcoins.

    4. To test an algorithm I do simulated trades using live market data. Is it paper trading? I’m not sure. I don’t do any back-testing. I probably should but it’s a lot of code to write πŸ™‚

    5. Every minute my bot gets order book data and recent trades from different exchanges, then calculates about a dozen parameters like mean price, variance, price velocity, volume etc. for different time intervals. Those parameters are fed to the trading algorithms.

    I will probably start saving all raw data that my bot gets. I don’t think you can get this type of historic data anywhere. If you guys decide to do something like a bot competition the data will help. If I have time I will start implementing back-testing using raw data.

  23. Jason says:

    @Michael Rakita – Very, very interesting. It sounds like you’ve learned a lot already and are making good progress.

    The next big step would be to start doing some backtesting on collected market data, followed by walk-forward testing and Monte Carlo simulation. The only problem is that you really need a minimum of 3-months of book data to accurately test an intraday trading strategy, but I suppose it might be possible to use transaction data and just build in an estimate for slippage.

    Anyway, here’s a good introductory book on the subject if you’re interested:

    * Building Algorithmic Trading Systems, + Website: A Trader’s Journey From Data Mining to Monte Carlo Simulation to Live Trading