Trading multiple instruments

[mikeyork]

mikeyork,

I don't know of any other way to explain this. If the next bar's open is 17 seconds later then your order will be submitted when that bar opens.
There is nothing to be done differently. Even if we allowed you to submit your order at bar close it is still exactly the same thing as it is right now. It will not be processed till the next bar's open. There is no data or possible entry point before these 17 seconds. The open for this instrument is 17 seconds later and this is why your order is 17 seconds behind your other orders.

It is quite possible to "fill" a market order at the last close price when back-testing without waiting for the next bar. I understand this is "cheating" and not consistent with the way you currently do things. But if it gives a better simulation then it should be an option.

And, as I said, I do not believe delays of 17 seconds (or even minutes) before the next tick are realistic. Please address that.

[NinjaTrader_JoshP]

mikeyork,

The approach to fills is what it is. As an algorithm the fill is done this way. If you wish to change the fill algorithm feel free to create your own custom fill type, but we cannot offer any support in this regard beyond what is described in this article: http://www.ninjatrader-support.com/H...FillTypes.html

Your delays of 17 seconds or minutes is not something we can address. It is simply what your data is telling us. We cannot do anything to change the data for you. As I stated earlier, you need to run the suggested checks of either manually looking at the chart to see the timestamps between the bars in question or just use the line I suggested to print the timestamps and then run checks in that fashion.

Unless you can find a timestamp printed out BEFORE these 17 seconds for that instrument then it is what it is. That is what your data is telling us and that is the way it will behave under your particular data set.

[mikeyork]

Here are some relevant prints. Leading 'O' is from order update (time is time stamp), leading 'B' is from bar update (time is from bar).

B: 9/8/2008 9:00:10 AM $EURJPY (signal bar)
O: 9/8/2008 8:59:00 AM -- Order='NT-00000/'$GBPUSD'
O: 9/8/2008 8:59:00 AM -- Order='NT-00001'$EURUSD'
O: 9/8/2008 8:59:17 AM -- Order='NT-00002/'$USDCHF'
B: 9/8/2008 9:00:18 AM $USDCHF OHLC: 1.1311 1.1311 1.1311 1.1311
O: 9/8/2008 8:59:17 AM -- Order='NT-00002 State=Filled '$USDCHF'
B: 9/8/2008 9:00:28 AM $GBPJPY OHLC: 189.36 189.36 189.36 189.36
B: 9/8/2008 9:00:31 AM $USDJPY OHLC: 107.96 107.96 107.96 107.96
B: 9/8/2008 9:01:02 AM $EURUSD OHLC: 1.4123 1.4123 1.4123 1.4123
O: 9/8/2008 8:59:00 AM -- Order='NT-00001 State=Filled '$EURUSD'
B: 9/8/2008 9:01:09 AM $EURJPY OHLC: 152.52 152.52 152.52 152.52
B: 9/8/2008 9:01:30 AM $GBPJPY OHLC: 189.48 189.48 189.48 189.48
B: 9/8/2008 9:01:32 AM $USDJPY OHLC: 108.01 108.01 108.01 108.01
B: 9/8/2008 9:02:00 AM $GBPUSD OHLC: 1.754 1.754 1.754 1.754
O: 9/8/2008 8:59:00 AM -- Order='NT-00000 State=Filled '$GBPUSD'
B: 9/8/2008 9:02:10 AM $EURJPY OHLC: 152.539 152.539 152.539 152.539
B: 9/8/2008 9:02:30 AM $GBPJPY OHLC: 189.68 189.68 189.68 189.68
B: 9/8/2008 9:02:33 AM $USDJPY OHLC: 108.05 108.05 108.05 108.05
B: 9/8/2008 9:03:00 AM $GBPUSD OHLC: 1.7542 1.7542 1.7542 1.7542
B: 9/8/2008 9:03:06 AM $EURUSD OHLC: 1.4128 1.4128 1.4128 1.4128
B: 9/8/2008 9:03:13 AM $EURJPY OHLC: 152.672 152.672 152.672 152.672
B: 9/8/2008 9:03:21 AM $USDCHF OHLC: 1.1312 1.1312 1.1312 1.1312
B: 9/8/2008 9:03:31 AM $GBPJPY OHLC: 189.73 189.73 189.73 189.73
B: 9/8/2008 9:03:33 AM $USDJPY OHLC: 108.12 108.12 108.12 108.12
B: 9/8/2008 9:04:00 AM $GBPUSD OHLC: 1.756 1.756 1.756 1.756
B: 9/8/2008 9:04:34 AM $USDJPY OHLC: 108.08 108.08 108.08 108.08
B: 9/8/2008 9:05:00 AM $GBPUSD OHLC: 1.75565 1.75565 1.75565 1.75565
etc.

Execution times in Execution grid are 9:03:21 AM for all three fills.

What happened between the various fill times and the execution times (minutes)? There are several bar updates in that period. Why are all execution times the same when all the other times vary so much? Is this a very extraordinary coincidence?

I notice for other trades that it is quite common for two of the simultaneously ordered trades to have the same execution time, but one differes, several minutes after the orders were placed. Note also that these bar times (these are actually 1 second bars) are many seconds, even minutes apart at times when trading in the US is in full swing. Are you suggesting I have bad data? It comes from BarChart using NinjaTrader.

Another noteworthy detail: it loooks to me like the data is coming in 1 minute bars (give or take a second or two). Yet I clearly specified 1 second in the backtest dialog.

[NinjaTrader_Ray]

Sorry for the delay,

We can reproduce and this behaviour is not expected. We have potentially resolved this for NT7. We can not introduce a fix for 6.5 since this fix touches extremely sensitive areas of our code that has potential implications and thus needs to be thoroughly tested.

[mikeyork]

Three questions:

1. Is this problem peculiar to BarChart data, or does it apply to all data providers? What if I switch to TradeStation?

2. Does it apply only to back-testing or will it also arise in real-time testing?

3. How soon do you expect NT7 to be ready?

[NinjaTrader_Ray]

1. All vendors
2. Backtest only
3. Q1 2009