BlueHorseshoe

That sounds like an excellent suggestion - please can you post your findings, along with some general information about how you tested the strategy (timeframes etc)? I look forward to seeing your results. Thanks.

Tams, I find it quite fascinating that when you post a useful volume indicator on here the "volume hawks" give it virtually no attention whatsoever . . . Perhaps next time you should post some vague, wishy-washy, and totally subjective claptrap about volume spread analysis and what a 'big' bar means? Your post might get some more attention that way. Anyway, thanks for the code!

Hello, You're not misunderstanding 'percentage profitable' at all, so it's a very good question. It highlights something un-stated in how I backtested to get these results . . . In addition to exiting positions according to the relevant scaleout rules, the strategy also exited positions when an opposing entry signal was given. In other words, to enter long, the strategy did not require that you were flat (and similarly for shorts). So while there might have been a higher percentage of 2 point targets met than there were 4 point targets met, a proportion of the positions entered with a 4 point target would have closed out for a profit without that 4 point target ever having been met. Please bear in mind that although I tried to choose something really simple with which to demonstrate my point, the way in which the different elements of even a very simple system interact with different parameters can be complex. You could quite reasonably complain that my example is far from ideal in that respect.

Of course it's not a 'proof'. No matter how many examples I provide in favour of a generalising statement of this kind, then it will still take only one single counter example to disprove it. It's called the 'falsification principle'. I made it abundantly clear in my post that I was aware of this fact. I studied the philosophy of science for three years at one of the best universities in the world, so you'll excuse my being a little short tempered with your patronising and unnecessary need to point this out. So I can't provide a proof for my claim, but what I am requesting from others, a proof of the incorrectness of my claim, could be very easily provided. While I can assure you that I didn't sit down and deliberately search for a set of parameters that supported my claim, it may be that I have inadvertently done so. But where are the posts from any of you giving a concrete example of where scaling out does work? You can optimise away all you want, it doesn't matter - one concrete example and I have to shut up and go away.

I think you (and plenty of others on here) are confusing 'eliminating risk' with 'eliminating the opportunity for profit'. No risk means no reward guys, there's no two ways about it. If you start jamming your stop in at break even all the time then you're not doing yourself any favours, however good it feels. If you're trading systematically, then there's no need to worry about 'viewing the market objectively'. Your views of the market will become irrelevant.

I wasn't suggesting that you could sidle up to any local and they'd reveal exactly how they trade. I simply said that you had access to them. Are you trying to tell me that watching one of the best traders on the floor of the CBOT and trying to model them wouldn't be beneficial to most newbies? If I came and stood next to you in the pits, then you wouldn't even speak to me, sure, but maybe I'd stand and watch you lose, watch how you responded to that loss, and think to myself "hmm, look at this guy, he just dealt with it calmly and now five minutes later he's got a profitable position on; last time I had a losing trade like that I just let it go indefinitely against me and hoped it would come back, maybe that wasn't the right thing to do . . ." Sounds like a learning experience to me.

Hello, If you're able to program EasyLanguage then you could have a look at TradeStation. You could open and fund an FX account (without needing to trade it) and that's going to give you historical data for most stocks. Another little known charting package to look at if you only need end of day data is ProRealtime. Charts are streaming (no download or installation) and it's free. You can program and backtest - I'm not sure what the language is, but its not hugely different from EL so you'd pick it up pretty quickly, and there's plenty of support documentation. Hope that helps!

I made this same point in another thread last week, and received a very aggressive response from a guy claiming he operated a successful self-built HFT system from a server in a closet. I didn't think at the time to ask whether the closet was located in the CME building or not - perhaps he's a janitor there?

Hopefully I made it clear in my post (the final para) that I wasn't doing this, and that I was fully aware that I was citing a specific example of a backtested system, and the limitations of doing so. The 'generalisation' comes not from that specific example, but from the fact that it is typical of the same tendency in the many hundreds of systems that I have examined. You say the profitability of scaling out is 'system specific' - but to what system is it specific!?!? Despite several days of pressing this question across two different threads, and receiving responses such as the one above, not one single person has responded by giving a concrete example of where the scaling out of profitable trades actually works.

You're probably correct and I hadn't read through the earlier posts clearly enough - I was certainly talking about 'generally' rather than 'always'. I wasn't referring to arbitrage, however, which would suggest buying the one and selling the other, but to simply using the futures as a leading indicator for the cash index.

Hi MightyMouse, I don't trade intraday and almost certainly haven't spent as much time as you or many others on this forum studying price action or watching depth of market info. So I'm pretty much 'on the fence' with this, but I think there are people who would disagree with you, and I don't think it would be true to say that their arguments don't make sense. If you take the stance that a liquid market exists to efficiently facilitate trade, then you would expect it to gravitate to the price levels where the most trade would occur. I'm sure you've read countless times about floor traders supposedly moving the market to areas where they know a confluence of stop orders are placed, such as the prior day's high/low, in order to trigger these. Do you not think that the market behaves like this anyway to some degree, without the deliberate manipulation of the floor traders? On a similar note, can I ask, would you consider a support level to be an area where many buyers are willing to step in (or a few buyers bidding for many contracts), or an area there is little interest or price agreement from market participants (ie something like the opposite of 'fair value')? Finally, can I ask how you reached the undertanding you described (ie is it the result of observation, testing, other people's theories, etc)? Look forward to hearing your response.

Although I wouldn't go as far as to say that this edge is always present, it's the existence of edges of this type that the HFT firms tend to exploit. The market doesn't leave this type of edge open to all participants, though. To exploit this you would need: a) to have the capital to take some type of weighted position in a basket of 500 stocks. b) the capital to make small price differentials profitable. c) the means to identify and respond to such opportunities with great speed and precision. I'd be curious to know how the big S&P contract relates to the cash market and the e-mini in this respect.

I haven't thought this point through in any great depth yet, but wondered about the following: If System A trades 10 contracts and gives a yearly net return of 100K without scaling, producing a 50% drawdown, and System B trades 10 contracts and gives a yearly net return of 50K with scaling, producing a 25% drawdown . . . Then surely the trader of OPM could simply use System A but only trade 5 contracts, thereby generating a 50K return with 25% drawdown? Of course, I'm talking in terms of invented figures here . . .

Surely this depends on your beliefs about the auction process? In the example above, I would suggest that the thing that may move the position against you is not increased 'selling pressure', but lack of buying interest. In fact, if you were to place a large sell order, then there is a strong chance that the market will gravitate upwards to try and fill it, benefiting your original long position. Even if we assume the market perspective that you discuss, surely you'd have to be trading a rather large number of contracts to have any significant impact on any liquid market? Look forward to hearing your thoughts on all this . . .

I thought it was about time that I did what I’m asking others to do, and put forward some actual figures to support my claims about the inadequacies of profitable scale-out strategies. So here are the results of testing a simple stop-and-reverse day-trading system, with various dollar profit exit methods. The system sells when a 3 period moving average crosses above a 30 period moving average, and buys when a 3 period moving average crosses below a 30 period moving average (ie it ‘fades’ the MA crossover). It always trades 2 contracts, and it always uses a 4 point per contract stop-loss. The back-tests all use the @ES e-mini futures contract over a 5 year period. 30 minute bars were used. No commission or slippage has been deducted. IMPORTANT: The profitability of this system is almost entirely due to the fact that the MA length parameters have been heavily optimised. It is used here as an example. Unless you want your broker to fall in love with you, please do not attempt to use it in live trading or you will most likely lose money. In the 4 months following the back-test period this system has consistently lost money. Here is the performance when we scale out of one contract for a 2 point profit, and the other for a 4 point profit: Total Net Profit: $60,925 Profit Factor: 1.17 Percentage Profitable: 58.7% If anyone wants to see equity curves etc then let me know and I will upload them. Next are the results of scaling out of the first contract for a 4 point profit, and the second contract for a 6 point profit: Total Net Profit: $79,812 Profit Factor: 1.16 Percentage Profitable: 57.93% Finally, we have the results of simply exiting both contracts for a 4 point profit – in other words, not scaling out at all: Total Net Profit: $166,800 Profit Factor: 1.23 Percentage Profitable: 65.48% I would like to point out that the ‘un-scaled’ 4 points per contract profit target is not the optimum target (this would have been 7 points per contract, which goes some way to explaining why the 6 point late scale-out fared better than the 2 point early scale out). This is why I think Tim Racette's advice in this thread to scale out is poor - I believe that you will find that what you see above repeats itself in pretty similar form for any strategy. And before anybody starts quoting Karl Popper’s falsification principle at me, I am fully aware that I can produce dozens of examples to support my argument, but that it only takes one counter example to discredit it. . .

A very helpful post - thanks for the links!

My understanding was that the futures always led nowadays?

There are a few problems here: One is that on the current bar[0] you are issuing an instruction to buy/sell on the next bar[-1] based on the previous bar[1]. What you should be doing is issuing an instruction to buy/sell on the next bar based on the high/low of the current one. Another problem may be that you're not specifying the point size - for instance in the @ES you have a 'big point value' (a point), and a 'small point value' (a tick, or quarter point). The 'begin . . . end' bracket seems redundant. I am not sure why you are bothering to specify the number of contracts in the code if it remains fixed throughout your strategy - just issue a buy/sell order and use the default settings on the strategy's formatting to determine the fixed number of contracts to trade. I am guessing that something like the following would produce the result you want: If Marketposition = 0 and date = 1110621 then Buy next bar at high+1 stop; If Marketposition = 0 and date = 1110621 then Sellshort next bar at low-1 stop; if Marketposition = -1 then Buy next bar at high+1 point stop; If marketposition = 1 then Sellshort next bar at low-1 stop;

Hi Tim, Whist I can see why this would reasonably appear to be the case (I even did it myself for a while), I am totally convinced that if you do the maths (or better still, get a computer to do the maths), over a significant number of trades you will find that it does not enhance performance. Trailing a stop only reduces your risk after you have trailed it. Your initial risk is unaffected. So what you need to pay attention to is how many times the position moves against you without ever giving the opportunity to trail the stop loss and reduce the initial risk. Trailing stops is another topic entirely, but once again, I would argue that they significantly reduce the performance of a trading system. Even the performance of very long term trend following systems is inhibited by the use of trailing stops, and for traders, who have to work around intraday volatility they can be lethal.

I totally agree with what you're saying here. Discretionary trading is a different matter entirely, and if you can make scaling in, out, or just about anything else work, on a discretionary basis, then that's great. Personally my own initial attempts at discretionary trading were dismal to say the least, and I have been deeply envious since of anyone with any kind of intuitive knack for trading. My suspicion that it is far harder to trade intraday without this knack was one of the reasons that I moved to longer term systems where mechanical approaches work well.

I totally agree with the sentiment of your post. However, finding people who are consistently profitable in their trading, who are willing to provide evidence of this, and who are willing to give any kind of mentoring or guidance to new traders, is extremely difficult. I think we are at a far greater disadvantage in this respect than traders in previous decades: twenty years ago if you were really serious about becoming successful then you moved to Chicago, leased a seat at one of the exchanges for a few months, and gained instant access to some of the most profitable traders in the world. What is the equivalent of that in 2012 for the screen trader?

Hi Cory, I gave the CCI as an example really (preferring to avoid revealing exactly how I trade!), and I don't use it in my own trading, but I have backtested this before and it does perform well. Basically, most overbought/oversold indicators can provide an edge in this way if their length settings are greatly reduced. I would recommend avoiding Stochastics though, as for some reason it doesn't perform as well. Another thing to be aware of is that there is an asymmetry to performance. I know that many eminent traders have argued that their is no justification for a system that has an inbuilt bias towards long or short positions (either Dennis or Eckhart has a real rant about it in one of the Market Wizard books), and this is probably the case for most systems in most markets. However, in the stock indices there is a very clear and easily demonstrable tendency for long positions to outperform short positions with this type of strategy. In other words, the tops of upward corrections in long term downtrends tend to be harder to identify than the bottoms of downward corrections in long term uptrends. It isn't too hard to guess why: in stocks buyers are quick to step in when they see 'bargain' prices, whereas the indices tend to top out due to lack of buying interest more so than due to increased selling pressure. This can justify asymmetrical settings for the indicator you're using (eg. Buy if the CCI falls below -100, but Sell only when the CCI rises above 130). What's more, this asymmetry applies more to the values of exits parameters than it does to entry parameters, and exits where what your original question was about . . . Basically, before you try and do anything with any of the above information, please test it thoroughly for yourself!

I know exactly what you mean - I experience the same frustrations as I only trade the S&P cash session. If you happen to be using TradeStation then there is a simple way around this problem (but unfortunately I can't help you with any other charting package)?

Hi Cory, I'm sure that you probably already know most of the common methods for doing this - basically using any kind of dynamic, adaptive indicator, such as ATR, Standard Deviation, or even a moving average to provide a target and stop. Personally, I tend to rely entirely upon overbought/oversold indicators to make trading decisions. I use the same indicator that gave me an entry signal to provide an exit signal - this will give the exit signal the same regardless of whether I exit the position for a profit or a loss. An example of a long position when the long term market trend was upward would be something like: - Entry: On the close when the Commodity Channel Index falls below -100. - Exit (for profit or loss): On the close when the CCI rises above 0. Therefore I don't know my exit when I enter a position. Because the CCI is an indicator that adapts to changing market conditions, so will my exit criteria. Please note that I also use a 'catastrophe' stop - an actual hard stop placed in the market in case of extreme events. This has only ever once been hit (and if it hadn't been there, I would actually have exited at a better price), when the markets came off at the end of last summer. Also note that I swing trade - this type of approach will work intraday, but with far greater volatility of returns. I hope that answers your question and is some help to you in your trading.