Trial and Epic Fail

Ventures of an ex indie game developer

Good bot, bad bot

 My automated trading system has two bots running. One with most of the dough running a low-risk, long term strat. The other one is handling less money, but running high risk settings.

Both did poorly over the summer, but at least the long term one seem to be close to break even. At least now they have a chance as I fixed a couple of serious bugs that I had eliminated in my previous bot, but apparently my substandard cut'n'paste skills didn't cut it...

Also, the above can also be partially explained by that Bitcoin normally goes into trading range during August and September, here is the average monthly movement since 2014:


I have a couple of trading bots with slight variation in their algorithms, which I believe can be applied to any asset: Twerk and Poem. Their annual profit is on par, but Twerk wants lower leverage and a larger part of the portfolio, while Poem is more choppy with a higher leverage.

If Bitcoin continues to disappoint for the next three-four months, I'll move to some other asset. But I do think that as more players move in, the manipulation is going to dwindle.

This should conclude my 2.5 year long exposé into trading in general and cryptocurrencies in particular. This is what I've learnt:

  • Technical analysis works.
  • It's hard to get an AI on par with static analysis.
  • Futures are much more profitable than spot trading, as you can bet on both going up, and going down.
  • Cryptocurrencies are extremely manipulated (starting within one month of Binance introducing their futures).
  • High leverage trading require large amounts of unmanipulated historical data for backtesting bots.
  • Cross margin is bad. Use higher leverage and get a substantially higher ROI using isolated margin instead. Cross margin and high risk in a manipulated market gets you rekt eventually.
  • All exchanges have their quirks, especially under large volume.
  • The more seldom a bot trades, the longer it's going to take to find it's critical bugs.
  • The best trading instruments are relatively stable. Small, unknown assets might bring in some bucks, but that's nothing compared to stable earnings.
  • Exploration before refinement.

Now it's on to new adventures of me. I'll maintain finplot, but apart from that try to avoid thinking about trading.

About the author

Mitt foto
Gothenburg, Sweden