screengrabs for some strategies i have developed.

[Metamega]

Did you have enable Intra-Bar Generation with LIBB. If not results are flawed from the get-go. Nothing to do with data.

As for advanced bar type, just like MA's, they are garbage to begin with. Price is price. Period. Fantasy works in Hollywood not Wall Street.

I've never used Tradestation but I was assuming theirs probably some setting like you mentioned.

You really need to understand how whatever backtester your using works.

For instance I use Amibroker I generally keep my orders to the open/close as I can reference which came first and delay my closing orders to the next bar. Their are circumstances where you can use other values but you have to make sure their not forward looking bias.

Another thing is when using indicators ( special candles included) as to how their created. Many indicators are forward looking and their historic values will change as new data is introduced. A quick example I can think of would be Zig Zag. Even a simple moving average system can easily become forward looking.

 

[rtw]

No arguments from me on your conclusions as to day-trading.

What do you mean by "advanced" bars please?

tradestation refers to any bars other than minute or daily as advanced (range, momentum, kase, renko and others).

 

[rtw]

Maybe im not fully understanding this, but if the OP did his tests on data that is not provided in tradestation, can we not instead try to use the data he used but in real time? the results of the backtesting, which apparently includes commissions and slippage, look phenomenal. can we not bend over backwards to get the data we need live?

thanks.

all i would need to get some of these strategies to deliver fairly positive results would be 10 - 20 years of tick data for these instruments. the only issue i have is the position management after the strategy has decided to open a position.

 

[rtw]

Run your backtests on high-quality data from https://www.tickdata.com/ or a similar vendor.

that would be all i needed, but only if their data was compatible with tradestation, which surely is not the case. i also strongly suspect they charge thousands of usd per year of tick data, and that would add up to a lot of money very fast.

i have a couple of possible adjustments that could help my strategies work on real, live data, but i'm really burned out from this endeavor right now. i will get back to it at some point in the future.

 

[rtw]

Do the results from the live trading in the sim account match the results when you run your backtest over that same period?

no. that's the issue. the entries and stop loss orders work fine for the most part in either case, but breakeven stops and trailing stops get triggered all the time on a chart being drawn from live, tick by tick data. when the same chart is closed and reopened, it redraws from historical data and tradestation treats large bars like chunks of straight, uniform price movement up or down, so it can't see all the noise that triggered all the breakeven and trailing stops.

 

[rtw]

Did you have enable Intra-Bar Generation with LIBB. If not results are flawed from the get-go. Nothing to do with data.

As for advanced bar type, just like MA's, they are garbage to begin with. Price is price. Period. Fantasy works in Hollywood not Wall Street.

the platform disables ibg and libb when charts have more than a couple weeks of data. and, tradestation only saves the latest 6 months of tick data, on 10+ year charts like i use there's no use enabling ibg or libb when there's no inside bar data for the platform to evaluate.

 

[rtw]

I've never used Tradestation but I was assuming theirs probably some setting like you mentioned.

You really need to understand how whatever backtester your using works.

For instance I use Amibroker I generally keep my orders to the open/close as I can reference which came first and delay my closing orders to the next bar. Their are circumstances where you can use other values but you have to make sure their not forward looking bias.

Another thing is when using indicators ( special candles included) as to how their created. Many indicators are forward looking and their historic values will change as new data is introduced. A quick example I can think of would be Zig Zag. Even a simple moving average system can easily become forward looking.

the issue here is the historical data.

in tradestation strategies evaluate the conditions to enter at the close of each bar, so all entries are triggered at the end of each bar. stop orders are the ones evaluated tick by tick.

i'm using very pedestrian indicators and leaving 4, 5 years of data out of sample when optimizing and evaluating, so these are not strategies that can be easily faulted for being forward looking.

 

[fan27]

no. that's the issue. the entries and stop loss orders work fine for the most part in either case, but breakeven stops and trailing stops get triggered all the time on a chart being drawn from live, tick by tick data. when the same chart is closed and reopened, it redraws from historical data and tradestation treats large bars like chunks of straight, uniform price movement up or down, so it can't see all the noise that triggered all the breakeven and trailing stops.

I suggest reprogramming how you configure your break even and trailing stops so that live trading will match the historical backtest.

 

[SunTrader]

the platform disables ibg and libb when charts have more than a couple weeks of data. and, tradestation only saves the latest 6 months of tick data, on 10+ year charts like i use there's no use enabling ibg or libb when there's no inside bar data for the platform to evaluate.

LIBB is not disabled with more than a couple of weeks of data. And although TradeStation only saves 6 months your PC has no limits.

And yes you are correct if there is no inside data there is no way to use LIBB. That is obvious. And without using tick data software assumes where entries and exits happen based on certain criteria which sometimes may or may not line up with real world conditions.

Therefore it should also be obvious to you your test results are less than useless - without use of tick data and LIBB.

 

[kevinkdog]

Ugh, I feel sorry for you OP. Nothing more frustrating to work on something, only to find that some flaw/issue makes all the conclusions and results useless.

For probably the first 5 years I used Tradestation, I trusted nothing except standard time based bars, and market orders (no stops, no limits). Save me a lot of frustration.

I tell everyone who backtests: "If it looks too good to be true, it probably is."

That advice won't help you, but maybe it will help someone else...