Home Community Support and Knowledge Base USDBTC Trade history profit but autotrade has lost money Reply To: USDBTC Trade history profit but autotrade has lost money


Hi everyone,

I’m definitely sorry to hear about these instances. Our personal accounts run on Auto Trade and we have had many customers on for a long time.

, would you mind please emailing an export of your trade history to us at [email protected] so we can review the details?

, when folks have shared such issues, we usually see the cause falling into one or more of these areas:

(a) They may start at a time where performance wasn’t great and they only stick around for a short while.

(b) For people trading smaller amounts, fees will eat up more than is tolerable. This is especially true on Coinbase Pro’s first tier, which is 0.5% per side. So a single round-trip trade costs 1%. That means a 3% loss becomes 4%. And a 2% gain only nets 1%. This is the most common difference we see, especially as our platform seeks smaller, more frequent and consistent gains rather than going after risky home runs.

(c) People turn auto trade off after a bad trade, miss the next good trade, then turn it back on. It’s a natural behavior but can nullify results.

(d) People may have limit orders on but don’t manage the trade when it cannot be filled. If you use limit orders, you have to actively manage scenarios when the order can’t be filled. Conversely, if you use market orders, you can expect varying fill prices–crypto moves fast and crypto exchanges don’t have the same controls as traditional exchanges.

(e) Building on (d), there was a major market dump in March that caused Coinbase Pro to go down for 45 minutes. This caused some much larger than expected losses for a number of people—on our system and otherwise. Anytime there is a major flash crash (which we’ve seen 30% drops in less than 2 minutes), you stand to take a significant loss–even if you have your stop loss on the exchange.

^^ To this point, we have reduced our entire execution pipeline time by 78% over the last few months. It is fast–but not instant–to ingest data, transform it, send it through the ML prediction, and then generate an alert and corresponding trade. We’ve upgraded infrastructure and refactored some of our code to bring this down to a minimal time frame.

You should continue to see execution speed improve. In normal conditions, these differences don’t matter. But in those rare flash moves, every little millisecond helps.

Additionally, exchanges are constantly moving to better handle orders during these crashes.

As for recommendations–

1. We see better fills from Binance.us and Binance.com during flash crashes.

2. For people trading smaller volume, consider using Binance to reduce fees.

3. If you’re using Limit orders, actively review and manage trades after an alert is issued.

4. If your results are below expectations, please share your trade history with us so that we can perform an analysis and find areas in which our platform can improve. We’re investing heavily in active development and want this to be an excellent end-to-end–your input and data helps move it forward.