开云体育

Paper Trading with MKT orders

kris k All Messages By This Member #46714 Hello Can anyone tell me how the IBKR paper trading via APIs is simulated ? This is how I am testing my system. I have a 3rd party tick data that I consume and build my predictions real-time (so I am not using any of the data feed from IBKR). When I place MKT order, is IB paper trading going to assume top of the bid/ask or is this last trade price or something else. I am not able to find any details on this. I would be grateful if anyone provide pointers.? When I compare my model performance after end of day using 3rd party tick data, I find the paper trading account and my own backtest significantly differs (I do account for slippage in time for my model computation and when the trade actually get placed) Thanks kris
mark collins All Messages By This Member #46715 To really understand this you'll need to account for possible differences in the exchanges, as well as the timing differences. Lastly you need to know that paper trades in the IB system are approximated, you may find real money trades differently. As a start I would want to validate that the prices reported are possibly tradable at the times reported in both places. Best wishes, M toggle quoted message Show quoted text On Wed, 10 Mar 2021, 20:10 , <kaykay.krish8@...> wrote: Hello Can anyone tell me how the IBKR paper trading via APIs is simulated ? This is how I am testing my system. I have a 3rd party tick data that I consume and build my predictions real-time (so I am not using any of the data feed from IBKR). When I place MKT order, is IB paper trading going to assume top of the bid/ask or is this last trade price or something else. I am not able to find any details on this. I would be grateful if anyone provide pointers.? When I compare my model performance after end of day using 3rd party tick data, I find the paper trading account and my own backtest significantly differs (I do account for slippage in time for my model computation and when the trade actually get placed) Thanks kris
kris k All Messages By This Member #46717 Thanks Mark. I do account for timing differences in my backtest (i.e I assume there is a time delta between when my model signal is generated and when the orders are actually placed. So in backtest, I look for the price traded at that specific time and evaluate model performance). I even assume worst?of bid/ask at that time.? But IB paper trading ends up far worse toggle quoted message Show quoted text On Wed, Mar 10, 2021 at 12:42 PM mark collins <mark.collins@...> wrote: To really understand this you'll need to account for possible differences in the exchanges, as well as the timing differences. Lastly you need to know that paper trades in the IB system are approximated, you may find real money trades differently. As a start I would want to validate that the prices reported are possibly tradable at the times reported in both places. Best wishes, M On Wed, 10 Mar 2021, 20:10 , <kaykay.krish8@...> wrote: Hello Can anyone tell me how the IBKR paper trading via APIs is simulated ? This is how I am testing my system. I have a 3rd party tick data that I consume and build my predictions real-time (so I am not using any of the data feed from IBKR). When I place MKT order, is IB paper trading going to assume top of the bid/ask or is this last trade price or something else. I am not able to find any details on this. I would be grateful if anyone provide pointers.? When I compare my model performance after end of day using 3rd party tick data, I find the paper trading account and my own backtest significantly differs (I do account for slippage in time for my model computation and when the trade actually get placed) Thanks kris
mark collins All Messages By This Member #46718 I would strongly recommend that you record the pricing that IB believes is available at the time, it will tell you whether it is a data issue or a simulated trade issue. At least if the data lines up you know where to focus, if it does not, well, you have your answer? Best wishes, M toggle quoted message Show quoted text On Wed, 10 Mar 2021, 20:59 kris k, <kaykay.krish8@...> wrote: Thanks Mark. I do account for timing differences in my backtest (i.e I assume there is a time delta between when my model signal is generated and when the orders are actually placed. So in backtest, I look for the price traded at that specific time and evaluate model performance). I even assume worst?of bid/ask at that time.? But IB paper trading ends up far worse On Wed, Mar 10, 2021 at 12:42 PM mark collins <mark.collins@...> wrote: To really understand this you'll need to account for possible differences in the exchanges, as well as the timing differences. Lastly you need to know that paper trades in the IB system are approximated, you may find real money trades differently. As a start I would want to validate that the prices reported are possibly tradable at the times reported in both places. Best wishes, M On Wed, 10 Mar 2021, 20:10 , <kaykay.krish8@...> wrote: Hello Can anyone tell me how the IBKR paper trading via APIs is simulated ? This is how I am testing my system. I have a 3rd party tick data that I consume and build my predictions real-time (so I am not using any of the data feed from IBKR). When I place MKT order, is IB paper trading going to assume top of the bid/ask or is this last trade price or something else. I am not able to find any details on this. I would be grateful if anyone provide pointers.? When I compare my model performance after end of day using 3rd party tick data, I find the paper trading account and my own backtest significantly differs (I do account for slippage in time for my model computation and when the trade actually get placed) Thanks kris
kris k All Messages By This Member #46726 Thanks Mark. That was a great suggestion that did give me insights into the real issue (that you alluded to subtly). There is strong correlation between IB prices and what I expect to so based on tape. So turns out my edge was small enough (close to spread value) that it was not real!? But still getting better fill is important. I don't know how well midprice or Adaptive orders could be simulated in paper trading for example. I'd greatly appreciate if you have any pointers to explore getting better fils.? (my system is momentum based, so will be taking liquidity away).

Previous Topic Next Topic