It’s been a rough last week anyways, but this morning, we had our first “out-of-sync” trade signal in a LONG time. And it was a very costly one. An out-of-sync trade signal is a signal that either a) is generated in backtesting that does not happen when trading live, or b) is generated while trading live but does not appear in backtesting.
While fill prices may vary a tick or two between live and backtesting, trade SIGNALS should NEVER vary.
This morning, every account we manage (including here at C2) opened a short position around 5:38AM. When I was online later in the morning and saw that open position, looking at the price chart, I was pretty sure that trade should not have happened (market was generally trending up at that point, and our short strat is pretty good about not opening a trade AGAINST the current market direction).
When I did a quick backtest, sure enough, that trade never happens.
The most probable explanation is also an explanation that we cannot verify. Periodically, our realtime market data provider that we receive trade and 1 minute candle trade data from, will “correct” a candle after the fact. For example, in real time, we may receive candle data that says the closing price of the 13:42PM candle was 5221.50. Then, when we receive the price data for the 13:43PM candle a minute later, our data provider could also send a “corrected” 13:42PM historical candle where the closing price has been changed to 5221.25.
What that means is in live trading, our strategy acts on a closing price of 5221.50 for that candle, but in subsequent backtesting, our strategy acts on a closing price of 5221.25 for that candle.
We will be doing two things: 1) keeping an eye on EVERY upcoming live trade, and making sure there is a corresponding trade in backtesting, and 2) auditing our last few weeks of live vs. backtested trade signals (more of a pain than it sounds!) and making sure there have not been any other occurrences.
