I am by no means a mathematician and need some help on the following. On my previous thread we identified that with my strategy (which required limit orders and limit stops) that the best way to backtest this was making criteria that price had to move through my limit order and only then i can assume my order was filled. Now i am testing if my entries have predictive power or edge so i have simply used the ES market and applied 4 point stop and 4 point target for my strategy. e.g. just for sake of example lets say signal was: entry criteria(0) > high(-5) - gives entry signal so lets assume limit order to buy short limit order was waiting at price high(-5), therefore if it trades through then entry price will be: high(-5). However, if i am simply determining whether or not the entry has any predictive power then should i not just consider: entry criteria(0) >= high(-5) - therefore entry is taken when prices touches high(-5). Again, i am just curious in terms of mathematical principle behind predictive power of an entry point (considering zero slippage and commissions). This is all being tested on intraday basis on 1min chart and goes back to 2007 so i have more than -500,000 bars of 1min data. I then have gone on to get 1200 randomly selected numbers between -500,000 and -1 and then used these randomly chosen bars as entry for my random tests. Max number of trades for this example on both random and system test was 600 trades executed. I have then performed: 50 different random back-test on same set of data - with same time parameters as my actual system and only max two entries per day (same as my system). I have used same 4 point target and 4 point stop. The results were that 45 of the random back-tests were negative. Now my question is if when the random bar is selected then if high(0) > high(-5) then short and entry price will be high(-5). Now again this means price had to move through my limit order point. However, if the entry is purely random and the entry has no predictive power then should it really matter if price has to trade through or not as winners and losers will be random?? Therefore, setting random data entries at high(0 >= high(-5) should have zero bearing on profit over a 50 random samples of data because the entry has no edge and odds are if we were filled at exact limit order every time then some random sets of data might have even more losers and some will have more. When i ran the random backtests with different variables even for entries e.g. sometime i used the close, open, low. I have found this exact result that in some limit order filling every time price is hit sometime improved performance and sometimes made the performance even worse. Now with my actual system, if use entry criteria(0) > high(-5) example then i still have upward trending curve but which is choppy (bear in mind i always thought the targets and stops in my strategy were far more important that my entries) but if i choose in our example: entry criteria(0) >= high(-5) then i get a very very smooth upward sloping equity curve. Mathematically would this mean my entry on this particular market has an edge. Can someone please explain to me if it matters whether price has to trade though my limit order or does this not make a difference just as the random test proved (again i just want the maths behind no overall system design basis)??
Also, guys i was wondering why anytime small stops and targets were chosen in random backtests with random entry (assuming zero slippage and commission) that they all tended to produce negative returns. I assumed that if market intraday was random as some say then surely excluding commissions and slippage that winners and losers should be somewhat the same. I obviously come to the conclusion that the choppy nature of the intraday ES market destroys any strategy with small stops. I then tested random entry again with zero targets and zero stops and exit end of day and again i had mainly losing random strategies. Why is this? Can someone explain this to me?
Yes, it does matter - please note that for products traded on a single exchange, this is enough to guarantee you a fill (in live trading conditions, ie. you have a working order and price trade through the limit). For products traded on multiple exchanges (stocks, forex), this is no guarantee, for a variety of reasons I will spare for a future discussion. If price doesn't trade through the limit, you only get a fill if your place in the queue was close enough to the head of the queue, which requires your order to be working at the exchange for already a "long" time (longer than most everyone else), and enough volume traded at that price so that your order effectively becomes head of the queue and get filled. For backtesting purpose, assuming this never happens is a safe bet.
Hey Dom, thanks. I think i did not make my post clear, what i am trying to get at is purely in terms of back testing my entry versus a random entry. I just wanted to know why it would make a difference statistically in the test i outlined in my first post. I think we discussed at length regarding overall strategy and price trading through my limit order in a previous thread i started in which you was again most helpful.
you need to learn how different types of orders work in markets. Try Harris and the various pamplets put out by exchanges. fro axample, my trades are always ahead of yours. I make money when you get left at the post, untraded. Start over, your efforts have been wasted so far.
Jack Hershey please go away. As for the rest of this thread i have stopped even sampling intra-day random entries as i understand it is pretty much -2 tick (+commissions) mean - i think it was a big waste of time doing this for intraday data. Saying that my strategy with the basic small stop and target displayed a positive gain which even under worst case fill scenario is more than 3 sigma, therefore i must conclude that even though the strategy still requires plenty of work i can be somewhat assured the entry part of he system is at-least statistically significant in its predictive power. Now what is the best way to test if my stops have any predictive power? I find this difficult because obviously the system is heavily reliant on entry therefore it is not possible to test my stop with a random entry. Obviously i will test random stops with my system entries and compare this against performance of system stops but is there a good way to test the performance of my stops?? Any help guys?
============ 2 problems/opportunitys ; best /brightest people in ES. ES is rather choppy/trands sloppy, but it is reel liquid......Not a prediction wisdom is profitable to direct
I am currently toying with a system prototype which uses the same signal for entry/exits ... that signal tells me "Long / Short / Flat" ... consecutive same-side signals can be used to scale-in (or be ignored, as in the attached summary). No stop whatsoever. P/F isn't the greatest, but after 1-tick slippage per entry / same per exit / and commissions, the net avg/trade is still pretty decent. You could try something similar, as an alternative to using stops.
average loser is larger than average winner and with only a 56% chance of winning (if the future looks like the past...and that's a humongous "if"),....there is absolutely no room for error....that system will blow up in real time trading.
-1 Performance statistics like these are typical of a mean reversion strategy (as opposed to a trend following/momentum strategy). See http://arxiv.org/abs/physics/0508104 It is true that a string of consecutive losers creates a big drawdown that it takes some time to win back; so the trick is not to trade too leveraged, so that you are still around for when the good times come back.