Sure - I've had it running on Windows Server 2012 and on Windows Server 2008r2.
I've also encountered a whole lot of inconsistency on Strategy Analyzer and Playback, and then in comparison to real-time sim and actual CQG market emulation, unfortunately. Hard to list all of them, but a few:
> After running a handful of backtests on the same strategy, I change the test window (from 2 weeks in October to 2 weeks in November, for example) and rerun. Strategy Analyzer then runs... and gives me results based on the ORIGINAL 2-week window, even though I've changed the dates. At first I thought, "Wow, that's some consistent results" but then I go into individual trades, and see that the trades were from the previous timeframe. This seems to happen after doing ~20+ tests on various symbols... so it doesn't happen the first time. But after that, it seems to happen consistently... like NT8 can't get out of that one timeframe.
> I've also run a lot of backtests in NT8, using Order fill resolution set at High / 1 tick. These are mostly enter/exit on same bar strategies (but with high-res backtesting, so there's some expectation it'll work). Then I use Playback with the same timeframe, same strategy, and the results are wildly different. SA says its a pretty good winning strategy, Playback reports significant losses. I did expect some difference between the two - I get that SA expects one contract (or lot) to be available at bid and ask, but the SA and Playback results are nowhere near each other. Which leads me to conclude: one of them isn't working properly. On NT7, I've also had MR provide promising results (even though it has 1-sec resolution), then real-time sim/emulation accounts provide unprofitable results... nowhere near SA or MR. And this is after nearly 1 year of MR testing, one week at a time, thousands of tested trades.
And overall, on NT8 SA and Playback still appear just unstable and unreliable. Plenty of times I try to start Playback, and the little Playback window will be gray and just won't come online (meaning: won't let me begin playback, the Play button is gray). So I close it, and sometimes it won't close. So then I close NT8, and have to kill the Playback process. Then I start all over again, hoping this time it'll actually come up and work. So... unstable. It seems like the database being used either isn't stable, or NT's interaction with it isn't robust, or something. Hard to feel confident about any of the results with so much glitchiness. I find myself toggling between NT7 (not granular enough w/1-sec MR resolution, but stable) and NT8 (granular SA and Playback, but such unreliability that it becomes unusable, and overall, low confidence that its results are anywhere near real, and then its general instability). It's a little frustrating that it's so shaky, frankly.
I know it's Beta and I'm hoping it improves, but these seem like fairly fundamental underlying database issues. Maybe not, or maybe I'm expecting too much or backtesting too much, but I sure like to be thorough and depend on my test tools. So, I'll keep providing feedback as my time allows... thanks to you, sledge, for all the feedback you're providing as well.
I've also encountered a whole lot of inconsistency on Strategy Analyzer and Playback, and then in comparison to real-time sim and actual CQG market emulation, unfortunately. Hard to list all of them, but a few:
> After running a handful of backtests on the same strategy, I change the test window (from 2 weeks in October to 2 weeks in November, for example) and rerun. Strategy Analyzer then runs... and gives me results based on the ORIGINAL 2-week window, even though I've changed the dates. At first I thought, "Wow, that's some consistent results" but then I go into individual trades, and see that the trades were from the previous timeframe. This seems to happen after doing ~20+ tests on various symbols... so it doesn't happen the first time. But after that, it seems to happen consistently... like NT8 can't get out of that one timeframe.
> I've also run a lot of backtests in NT8, using Order fill resolution set at High / 1 tick. These are mostly enter/exit on same bar strategies (but with high-res backtesting, so there's some expectation it'll work). Then I use Playback with the same timeframe, same strategy, and the results are wildly different. SA says its a pretty good winning strategy, Playback reports significant losses. I did expect some difference between the two - I get that SA expects one contract (or lot) to be available at bid and ask, but the SA and Playback results are nowhere near each other. Which leads me to conclude: one of them isn't working properly. On NT7, I've also had MR provide promising results (even though it has 1-sec resolution), then real-time sim/emulation accounts provide unprofitable results... nowhere near SA or MR. And this is after nearly 1 year of MR testing, one week at a time, thousands of tested trades.
And overall, on NT8 SA and Playback still appear just unstable and unreliable. Plenty of times I try to start Playback, and the little Playback window will be gray and just won't come online (meaning: won't let me begin playback, the Play button is gray). So I close it, and sometimes it won't close. So then I close NT8, and have to kill the Playback process. Then I start all over again, hoping this time it'll actually come up and work. So... unstable. It seems like the database being used either isn't stable, or NT's interaction with it isn't robust, or something. Hard to feel confident about any of the results with so much glitchiness. I find myself toggling between NT7 (not granular enough w/1-sec MR resolution, but stable) and NT8 (granular SA and Playback, but such unreliability that it becomes unusable, and overall, low confidence that its results are anywhere near real, and then its general instability). It's a little frustrating that it's so shaky, frankly.
I know it's Beta and I'm hoping it improves, but these seem like fairly fundamental underlying database issues. Maybe not, or maybe I'm expecting too much or backtesting too much, but I sure like to be thorough and depend on my test tools. So, I'll keep providing feedback as my time allows... thanks to you, sledge, for all the feedback you're providing as well.
Comment