Big difference in live papertrader & live tradingbot
#1
Hello!

I just started using gekko and deployed my first live trading bot yesterday. At the same time i started a live paper trader with the same strategy.

When i checked the results today i saw that the paper trader had made twice the amount of trades and the profits are insane.
I'm running the exact same strategy so why am i getting such a big difference?  I use the bot on Binance, and set the paper traders trading fee to 0.1,  is that the right value? Binance fee is 0.1%?  Also paying fees with BNB giving a discount of 25%.

The trades being done by the two are similar when it comes to profit target and duration, but i do see that the trading bot is closing more orders at the same price it bought compared to the paper trader.


exchange  currency  asset         status duration                          strategy      PnL      type trades
binance      BTC VET                  running 19 hours, 57 minutes     -----          -0.044      tradebot 82
binance      BTC VET                  running 19 hours, 52 minutes     -----           45.374    papertrader 206



Hope someone can help me with this, i have no idea what causes the difference.

Thanks!  Big Grin
  Reply
#2
I just tried my first live bot and I had to turn it off as it’s clear that live bot takes about 5m to see/act on the current candles on exchange and issues orders too late. Anyone knows a bot that can trade in real time?
  Reply
#3
That's a huge difference.

Do you have logs that indicate what the differences are? For example the buy/sell strategies should be 100% the same since they take the same market data. Also are you sure you are using the same candleSize and the exact same start time (if you don't Gekko is calculating completely different candles).
  Reply
#4
So, I was digging deeper into this, trying to figure out what’s wrong. I’ve experimented with the command line trader, adjusted the ticker to 1s along with 1m candles. Debugging/status messages did start to poll Binance every second but candles were still not flowing into my strategy code in real time, but only after triggered by some activity on exchange. What appears to be happening is if there was no activity on the exchange, the candles were not coming into the code, only after someone sold or bought something on the exchange. So if there was 10m of no activity on the exchange, no candles would show up in the code and only after someone sold or bought something, a batch of 10 candles would show up at the same time.

Unfortunately I can’t continue experimenting as I’m getting “[sticky order] FATAL ERROR’ “Error - 1105: Parameter ‘order Id’ was empty.” today. It was not happening yesterday, no clue what changed. Paper trader works great. I wish I could run my strategy live.
  Reply
#5
(12-06-2018, 04:59 AM)vtms Wrote: So, I was digging deeper into this, trying to figure out what’s wrong. I’ve experimented with the command line trader, adjusted the ticker to 1s along with 1m candles. Debugging/status messages did start to poll Binance every second but candles were still not flowing into my strategy code in real time, but only after triggered by some activity on exchange. What appears to be happening is if there  was no activity on the exchange, the candles were not coming into the code, only after someone sold or bought something on the exchange. So if there was 10m of no activity on the exchange, no candles would show up in the code and only after someone sold or bought something, a batch of 10 candles would show up at the same time.

Unfortunately I can’t continue experimenting  as I’m getting “[sticky order] FATAL ERROR’ “Error - 1105: Parameter ‘order Id’ was empty.” today. It was not happening yesterday, no clue what changed. Paper trader works great. I wish I could run my strategy live.

Take care with a 1s tick rate, you quickly hit exchange api rate limits. So your prolem will be even worse.

If you run paper and and live trader side by side you also double the traffic. In this case you might better use one paper trader
and one trader configured in leecher mode.

And keep in mind that not not every exchange trade will fire a candle in gekko, the lowest batch size is 1 minute.
  Reply


Forum Jump:


Users browsing this thread: