Backtesting with talib causes incorrect candle dates

[southerton81]

I have encountered a problem when backtesting, it manifests when I add talib indicator to mock strategy. Tested on 1m and 15m candles and it produces gaps in the log. 1 hour candles work ok.

var strat = {};

strat.init = function() {

  var settingsSMA = {
    optInTimePeriod: 20
  }

  this.addTalibIndicator('mysma', 'sma', settingsSMA);
}

strat.update = function(candle) {}

strat.log = function() {}
strat.check = function(candle) {
  console.log(candle.start.format());
}

module.exports = strat;

2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T20:54:00+03:00
2017-04-27T21:02:00+03:00
2017-04-27T21:02:00+03:00

Here's a related closed issue, however, the case seems to be different. #681

[askmike]

Nice find! Investigating..

[askmike]

I'm pretty sure I fixed this, I was relying on this.candle which in case of async indicators (such as talib) might be more recent than the candle this indicator was run over.

Can you download the develop branch and see if this fixed it for you?

[southerton81]

Switched to develop branch and the problem is gone, thank you.
I assume the problem was only with backtesting, not real market watcher.

[askmike]

yes indeed: with backtesting the strat gets fed a lot of candles very quickly. This was the result of a race condition (also the reason why you only saw it with 15min and not with 1h candles), which won't happen in realtime mode since candles only come per minute.