Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
i want to share some screengrabs of performance reports for some strategies i have developed on tradestation.
i'm interested in reading the opinions and impressions of experienced - profitable traders. i'm also interested in finding out whether there's anyone who has veritably accomplished to develop day trading strategies that really make money, being that at this point i would be inclined to think profitable day trading automated strategies for retail traders simply do not exist.
over a really long time i have developed some strategies that generate spectacular results on backtests. these strategies were optimized from january 2006 to december 2013 and all the trades from 2014 onwards happened out of sample and could be considered to be highly realistic and very promising. these reports include values of around 6 usd in commissions per trade and also very credible values of around 30 - 40 usd in slippage per trade. after other members have had a chance to share their opinions i will disclose more information on these strategies.
seeing that you are using advanced bar types, I would suggest to check that your system results are not just caused by the way TS constructs the historical bars or handles systems on these bars. The first step I'd suggest is to compare bars created in realtime to how TS creates the bars for the same period historically.
I can't speak to the accuracy of Tradestation's back-tests / SIM compared to the real market, but I can give you some details on NinjaTraders and some general observation about the walk between testing historical data via replay, back testing and comparing these to the real market.
1. Any sort of basic back testing that uses simple HLOC bars without advanced volume fill assumptions and without knowing precisely the sequence of how the bar was constructed will be considerably overstating your results. There are a number of reasons for this, but the biggest two are:
Entries: Limit orders can get perfect fills with no negative slippage on every entry. Which is not how the real
market behaves.
Exits: Profit Targets and Stop Losses will fill at roughly the same frequency on first touch every time. This is also
a very incorrect assumption. A profit target being a limit order will need 2-3 touches or a full price pass-through
in order to fill whereas a stop loss which is a market order will get filled immediately. So assuming you have a perfect entry every time and a 5 tick PT and 5 tick
SL, most all simple back testers will give you exactly a 50% / 50% win to loss rate, whereas in the real market you will see instead around 20%- 30% win rate and
a 70%- 80% loss rate.
2. As ABCTG mentioned using any sort of exotic bar types such as line break, renko or unirenko bars can trick a simple backtester and make the HLOC bar ignor all the granular action that occurred during the bar being formed. In just about every case I tried, I could beat any backtester with the most basic strategy using exotic bar types. These will never work like this in the real market. I am not sure what type of bars you are using in your system, but this is something to watch out for, because backtesting will not be able to accurately process these and your results will be skewed in your favor.
3. Latency is not considered at all in back testing. So most good entries and exits that perfectly touch the price you want will get filled almost immediately in back testing / SIM. In the real market there are lines of people at every price level and unless you were in line for a while more than likely you will not get that price unless it trades through it. This applies to both entries and exits.
4. Instead of using the back-testing feature in NinjaTrader I recommend using market replay. It is a day and night difference in accuracy and therefore it tends to be a day and night difference in how well your system will perform. NT 7 market replay is fair, but it will fill your profit targets too easily. It does show that you need in most cases 2-3 touches to fill, but I still see that a simple scalping strategy with few view bells and whistles can beat it in most cases, and in the real market it is not nearly this easy. In NinjaTrader 8 they come very close to getting it right. If you enable tick replay in the market replay and use a standard bar type with nothing exotic these results are in line with what I see when trading the live market. There are volume considerations on fills, and they are conservative on touch vs. pass-through with limit orders. They don't simulate people with larger orders jumping in line in the queue though, I think this is about the only thing that they miss. But as far as what I have seen this is the closest you will get to an accurate testing environment to compare to the real market.
And as far as bridging the gap between retail traders and institutional traders it is not as large of a gap as you may think. To improve your latency, you can get a server close to your data provider for a couple hundred dollars a month. To reduce your commissions you can get a seat license at the exchange for a few hundred a month (+ a one time non refundable fee of a couple thousand bucks). To increase your odds of getting a better spot in the queue you only need to trade around 3-5 contracts to be at a size where you will likely start skipping ahead. If you think about this like a business, this is not going to cost 100K to do. This is more of a 10k cost. In terms of a start up business cost, this is very cheap.
Need more detail but looks like scalping on very small TF given the amount of trades , unless its been backtested on tick by tick data i am dubious its repeatable in real trading . The slippage in 08 after lehmans collapse was of the charts . Once again with without more detail any meaningful opinion is difficult . Live testing is the real numbers
The best possible advice would be something like this:
Take some real money, put it into a live brokerage account, and run the strategy doing real trades. Don't use more than you could afford to lose, if it came to that. Have a plan about what to do in case the real results don't match the straight upward line of the simulated results.
Theoretical results are just that, theoretical. There's always a big danger with running theoretical trades in a simulation.
- the first thing i'll say is that anyone who was a piece of crap would no doubt try to use these strategies to defraud novice traders for big money. they can be adjusted to produce spectacular historical results on pretty much any symbol one chooses even when very significant slippage and commission costs are included while relying exclusively on market orders. as i have said, this spectacular performance holds up even for extensive out of sample data.
- the results when executing these strategies live (on tick by tick data) are quite different from historical results and it is all because the historical data that tradestation charges for and provides is useless. i have done extensive tests with my strategy on simulated and live data and results don't match the results on historical data at all.
i have used range, momentum and kase bars to develop these strategies and on historical data tradestation assumes these bars have been built in a completely unrealistic way. these bars are treated as if price had gone in a straight line (either up or down) for the range one specifies. this is really helpful to filter noise out and to better identify big trends, but this also makes historical data completely unreliable.
so, anyone who was thinking of also using range, momentum or kase bars to build their strategies around them should not waste their time with them, they are useless.
- the specific issues have to do with both breakeven and trailing stops.
my strategies use both of these kind of orders to manage trades once open. first i have a stop loss order when a position is opened and if price moves in my favor, a breakeven stop order replaces the stop loss and then a trailing stop is used to secure profits.
the thing is that when using these "advanced" bar types, it seems like breakeven stops would only be triggered in useful circumstances like when a trend reverses immediately and permanently, but on live data, breakeven stops are triggered all the time as it is common for price to move in one's favor for some ticks and then go back and cross the entry price once or multiple times. therefore, on live data all this noise which "advanced" bar types just ignore causes breakeven stops to be triggered continuously and the strategy to pile up a huge number of minuscule and small losses while killing practically all trades that would have been profitable.
trailing stops are also problematic when used with "advanced" bar types as trailing stops which are too tight show great results on historical data. they seem to lock up the largest profits possible (once again because of the assumptions with which range, momentum and kase bars are built) but on live data the biggest winning trades which end up being what make the strategies profitable would always be killed prematurely.
- reaching this conclusion took me months of tests and has meant months of work and development on this strategy have been for naught. it wasn't until i posted screengrabs of the difference between historical and live trades for my strategy and complained in tradestation's fora that they disclosed how "advanced" bars work and are built. tradestation only keeps the latest 6 months of tick data for any symbol, which is too short a period to develop a consistent strategy, all the years of other historical data they charge for and advertise are useless. i'd love to have tradestation advertise their data services like that: - only 6 months of minimally reliable data and years and years of useless trash -.
