BlueHorseshoe

Hello, The strategy uses the RSI on the timeframe being traded, yes. The filters are MAs across this and two higher timeframes (nothing at all original there). A buy limit order is placed at the low of the signal bar (to do so and then instruct the backtest to only fill orders where price trades through the limit is the only way I have of knowing absolutely that I would have been filled without adverse slippage - the strategy works equally well with a order to buy at market at the close of the bar, if you can be confident that slippage is not a huge issue). Another filter rule that I recall working well for the S&Ps was to only buy following a down day, and visa versa. Another interesting indicator to look at for this type of strategy is the CCI. Because this is already quite sensitive and 'jumpy', there isn't the need to shorten the lookback to the same extent as with the RSI. Let me know if there's any more info you'd like. Bluehorseshoe

A trend following approach could work for you if you have sufficient capital. Successful trend following funds nowadays tend to be incredibly well diversified and have a far more sophisticated trading model than many would lead you to believe: if you've got that kind of cash then I would suggest that you might be better simply placing your capital with a trend following fund with a long track record of decent returns. A good place to start looking would be Micheal Covel's book on trend following. Another option you could consider is 'swing trading' - operating in that vague and nebulous space between the cool-handed position traders and the hyper-active day traders. Swing traders typically make decisions from daily charts and put their orders in on the close. They often hold positions for less than a week. There are many strategies based around trading pullbacks in trends that can work well that you could investigate. Your number one rule needs to be this though: decide clearly what your goal is and then find the best way to achieve it. If your goal is to make the best returns on your capital and placing your money with a fund seems the best way to do it, then don't be seduced by the 'glamour' of trading. Hope that's helpful.

Thanks for your reply. I remember reading about the Larry Connors set of strategies and thinking that, though they sounded very plausible and made sense in terms of things I'd already discovered for myself, they'd most likely be disappointing. So I was suprised when I started testing them and seeing very positive results. Nevertheless, Connors leaves a hell of a lot of unanswered questions, and I have no idea whether he has any sort of a consistent track record as a trader. And of course, his focus is on ETFs and individual stocks, so a leap of faith is needed to start looking at his ideas if you want to trade corn or lumber. Fortunately there is some commodities overlap, as ETFs exist for things like oil and gold. I have used the 2 period RSI strategy you mention (also with additional filter criteria) to create a strategy for intraday trading. I've been monitoring this since last November, and it seems to be performing about as well as could be expected. Unfortunately I don't have the cash, the technology, or the guts to trade it, so I'm sticking with swing trading for now. I also took a look at the RSI(2) for the Euro this afternoon - it seems that it can be made to perform much better than I recalled, but only with significant optimisation. I think the filter MA needs to come down to about 120 and the overbought/oversold levels at +70/+30. I think when I examined this last year it was maybe backtesting with a 4 period RSI (another Connors idea), hence the discrepancy. So if you do find the time to take it further, I'd much sooner hear about how this performs in Bonds (I can't access t-bill charts in TS as I haven't subscribed for the data feed!!!) Oh, and one final thing to note - if you're using EasyLanguage for testing, the TS RSI formula is some rather weird modification of the original, and does produce slightly different results. Will look forward to hearing more from you about this type of trading - it's somehow reassuring to know that there are other traders out there looking at the same kind of things as I am!

I think that what you describe above has led to an increased directionless-ness (did I make that word up?) across all timeframes. There are more and more market participants, and fewer and fewer of them are doing the same thing at the same time. Whilever this is the case, I think that the market will be less and less able to sustain trends in any timeframe. It's worth noting that two recent clean, low volatility trending periods on the S&Ps daily charts (from mid2010-mid2011, and then the current trend) have been marked by low volume - where trends are concerned, 'too many cooks spoil the broth'. Just an opinion though.

I think that what you describe above has led to an increased directionless-ness (did I make that word up?) across all timeframes. There are more and more market participants, and fewer and fewer of them are doing the same thing at the same time. Whilever this is the case, I think that the market will be less and less able to sustain trends in any timeframe. It's worth noting that two recent clean, low volatility trending periods on the S&Ps daily charts (from mid2010-mid2011, and then the current trend) have been marked by low volume - where trends are concerned, 'two many cooks spoil the broth'. Just an opinion though.

A slightly off-subject question, but does anyone know the official designation of the language used by the ProRealTime software (and hence many of the spreadbetting firms who white label this charting package)? Thanks, Bluehorseshoe

This is one of those arrogant, patronising posts that I know people will deservedly give me grief about but . . . MightyMouse, go and look again at the second equity curve that jswanson has diplayed. Have a good look at it. Think about what has been required to produce that equity curve - how simple and robust the trading rules are. Take a good long look at that equity curve. Now why on earth would you want to add a stop-loss? (It's a rhetorical question). Bluehorseshoe

Hi jswanson, Thanks for yet another interesting post and study. The swing strategy that I trade is very similar to the one you use in your 'buying weakness' example. My entry criteria are a little more sophisticated (or that's what I like to tell myself!), but my equity curve in the S&P bears very close resemblance to the one you display, complete with the big one trade drawdown from last August. So although I know you're only presenting it as a study, I think that with very little tweaking it can form the basis of a good mechanical strategy. Incidentally, the precise criteria you use - 200SMA trend filter and 3 higher/lower closes - has been offered as a complete ETF trading system by Larry Connors. I believe that you'll most likely find the optimum filter MA length in your test above to be 180 periods, or thereabouts. Which leads to the inevitable question - should this be optimised? Then there's the subject of other markets. The Euro, which is one of the few super-liquid markets still capable of supportant a breakout trading style, is a no-go, although you would have been profitable with this strategy in the Euro until about five years ago. Oil was similarly profitable, though here the equity curve flatlined more recently, yet without any significant drawdown. Gold has been very kind, and in the past twelve months there have even been trading opportunities to the downside! The Dow has been great, and with suprisingly less correlation to the S&P than you might expect. European indices tend to be mean-reverting as well, and the Nikkei can also be good, although I don't trade it. Does anybody else trade with this type of strategy on a daily chart, and if so, what does your porftfolio contain? What I would also love to see is a study of this type of strategy in Interest Rate products - it's on my 'to-do' list but I haven't got around to it yet . . . Thanks again, Bluehorseshoe.

Thanks for an interesting post! If you take the approximate dates when your research shows that trend-following in the S&P has clearly begun to fail, and then look at some of the major trend following funds, you'll notice that many of them underwent 'major system modification' around this time. Dunn Capital is a great example - having traded over about a twenty-five year period using exactly the same strategy they were finally forced to reassess how they operated. Obviously most of whatever they changed remains hidden from public knowledge, but the information that is available shows that they massively increased the number of markets traded. Other newer firms such as Winton Capital have taken a similar approach - they are ludicrously well diversified. On a slightly different topic, I think a fairer comparisson could have been achieved by using the optimal lookback for each market. As things stand it could simply be the case that a 50MA is a totally unoptimal setting for the ES, but a perfectly optimal setting for the Euro. I know this isn't the case, and that the point of your argument holds true, but it's worth mentioning.

Thanks for the reply. It's been on my long list of 'things to look into'. It just went to the bottom of the list! I think my assumption was less to do with how it might impact theoretical strategy performance, and more about whether it would improve execution - obviously several of you have tried this in live trading and found no substantial improvement though. On a fairly closely related topic, has anyone experimented with or becktested swing trading strategies using (obviously very large) range bars, as opposed to daily?

If this is the same ELD file that I have (bound to be - the programmer eKam mentioned in the thread rings a bell), then to view the calculations the indicator performs, take the following steps: - Open the EL Development environment from the TS 'View' dropdown menu. - Select 'open'. - In the drop down, select 'Functions' - Find the relevant 'VChart' function file and open it. And now you're looking at the Value Chart calculation. The link to the Stendahl article is a useful intro, however there are almost certainly better ways of using it than those he suggests (ironic, seeing as he's the creator!).

That translates perfectly in to English - I know exactly what you mean

This is very, very easy to do with EL, and I'm sure if you're programming with Java it won't require much explanation. The code to create an order condition would look like this (assuming by 'positive for the day' you mean trading above the SMA): If C of Data1 > Average(C of Data1, 6) and C of Data2 > Average(C of Data2, 6) then . . . The important bit to grasp is that you must set up your chart so that you've inserted the two relevant timeframes as Data1 and Data2. This is very easy - just right click, insert symbol, add the symbol which will appear as a subpane, right click, and adjust to the appropriate timeframe. If this second timeframe is only required for reference by the strategy and you don't need it for visual on-screen reference, then you can hide it. Obviously Data2 could be a different symbol, allowing you to make trading decisions based on more than one instrument. However, you can't mix symbols with different delays in the same window, so here's a handy tip: find another derivative that is heavily arbitraged against the one you cannot insert and use that as a surrogate reference. For example, you can't mix the $TICK (based on NYSE data feeds) with the @ES futures (based on CME data feeds). But you could substitute the SPY Exchange Traded Fund (NYSE) but still use this to trade the @ES futures in most cash-session scenarios. Hope that's helpful.

Learning how to program fluently enough to perform accurate backtests, and also learning about all the pitfalls of backtesting (there are very many indeed), will require quite a bit of time and effort. The good news is that it will probably be the most profitable and worthwhile thing you'll ever do in your development as a trader, and it can also be great fun (as well as tremendously frustrating!). Good luck!

I'm not a fundamental trader, so I can't give you any advice on selecting markets to trade. Given that you'll be makin g decisions based on the strength of your convictions (unless you're taking a more quantitative fundamental approach), then I would imagine you would want to trade those markets about which you have the strongest convictions. One key thing to develop will be your 'philosophy' about particular markets. A revealing question to ask yourself is: 'do I, on balance, expect that this particular market is more likely to continue in the direction it is going, or reverse and head back where it came?' For me, for example, for the Euro I would consider the former to be the case, and for the S&P500 the latter. So I want to try and trade trends in the Euro, and trade counter to the trend in the S&Ps. Most new traders have completely unrealistic expectations about what they will achieve and how soon - you don't seem to be burdened with such expectations - so you're already off to a great start! Good luck!

Are you sure, BlowFish? I'm sure I once saw a John Carter vid where he said that it is a magical number . . . On a more serious note: one thing that jumps out reading back through this thread is that everyone keeps referring to the same time frames: 1, 5, 10, 15, 30, 60 mins. Now given that most traders lose money, and assuming we set aside preconceptions about self-fulfilling prophecies etc, then wouldn't it make more sense to employ reference timeframes that are not commonly used by other traders? In other words, why haven't you configured your charts for 4, 7, 18, 43 minute bars? Or, for that matter, 144 minute bars! Hope that's thought provoking (ps I lifted this idea from an old Joe Krutsinger book where he discusses 180 min bars).

Following my previous post I thought I would try and provide something to back up and hopefully clarify what I was describing. The following is code for an EL strategy that buys the ES at 1700 GMT - or the midday 'doldrums', in other words. It's only a demonstration, so the entry time was chosen at random (as was the decision to buy, rather than sell). The important part here is that the stop loss is outside of normal volatility (at 3 standard deviations), and the profit target is within it (at 1.5 standard deviations). I've also added in an instruction to buy only when an RSI shows that the market is oversold - this is not essential. And then a few suggestions for other optional filters (trend and 'headroom') are given in square brackets. You can run your own backtests, but over a ten year period the Profit Factor was 1.44 and the average profit per trade was $32, with total net profit of about $16k (that's without any deductions for slippage or commission). { ** Copyright © 2012-2013 BlueHorseShoe. All rights reserved. ** INTRADAY VOLATILITY SCALPER - @ES 15MINS - VARIABLE STOP AND TARGET DISCLAIMER: This EasyLanguage strategy has been provided as an educational example to demonstrate a potentially profitable aspect of the relationship between intraday market ‘noise’ and exit orders. It is neither a solicitation nor an offer to buy or sell any type of security or financial derivative. No representation is being made that any account will or is likely to achieve profits or losses similar to those described. The past performance of any trading system or methodology is not necessarily indicative of future results. FUTURES TRADING CARRIES A SIGNIFICANT DEGREE OF RISK. TRADE ONLY WITH CAPITAL THAT YOU CAN AFFORD TO LOSE. } inputs: n(3); if t=1700 then begin if RSI(c,2)<20 {and c>average(c,1000) and h<highest(h,20)-((0.5*n)*(StdDev(c,10)))} then buy this bar; end; setstoploss(n*(50*(StdDev(c,10)))); setprofittarget((0.5*n)*(50*(StdDev(c,10))));

Thanks. Pressumably your trading methods aren't dependent on closing prices, so you don't need to worry about discrepancies between backtesting and intra-bar real-time trade entries?

This is a fascinating study and exactly the sort of thing that I joined this forum for. Might it be worth posting this as a seperate thread to avoid pissing off TheRumpledOne? Do you have any ideas yet about how you'd go about buying close to the daily low? Have you looked at how the average low-to-close ticks are impacted upon by the long term trend? Have you considered a study to generate a volatility-neutral average low-to-close tick reading? Looking forward to reading more - please PM me if you do move to a seperate thread.

This makes sense up to a point, I think. When I test something I normally expect it to work on broadly adjacent timeframes. This is actually a great way to create a 'quasi-unseen data set' on which to test a completed strategy. Though the new timeframe test data will bear a resemblance to the development data, it will also provide a new set of specific data points. This can help in identifying over-optimisation, and also tends to give a more realistic representation of how the strategy might perform in real time trading. Nevertheless, there are significant differences between distant timeframes (5min and daily, say), owing to things like intraday volatility. There are plenty of strategies which, if you ignore commission and execution costs, perform the same on 5min charts as daily charts. Once you factor in the costs of trading these strategies then lose money on a 5min timeframe. That's because commission and execution costs are fixed across all timeframes but average profit per trade is not. This is the danger inherent in focussing too much on metrics like Profit Factors rather than actual dollar returns. Sometimes the above is what people may be referring to when they specify a particular timeframe for a strategy.

'Noise' is typically defined using a measure of volatility, of which the ATR is an example. You could also use Standard Deviation or, if you can program, the 'Volatility' function from which the DMI+/- is derived. There are also countless other more sophisticated possibilities, none of which are certain to perform any better. The way that I would approach this issue is to think 'I want my stop loss to be outside market noise, and my profit target to be inside it'. Such an approach should, in theory, present the possibility of profit without necessarliy having an edge in the entry criteria. Assuming that you had a 'neutral' (or random) entry, at zero, and that market noise caused price to oscillate by +10 and -10, and you were buying, then you'd want a stop at -12 and a profit target at +8, for example. Of course, in reality, this is a huge over-simplification. There may be more upside noise than downside noise. The noise level will change after the position is established. And assuming that your entry provides a significant edge, then your approach to noise may need to be different. A trend-follower, for example, places their stop within noise, and their target outside. They still have positive expectancy. Another thing to consider is that you are only looking at 'noise' within a particular timeframe. If you didn't have a stop in place and allowed price to wander significantly against your position, then this adverse excursion is a perfectly normal amount of noise in a higher timeframe, and you can therefore anticipate a corresponding amount of upside noise within that higher timeframe. I personally trade without stops (to all intents and purposes) - I have a percentage win rate that might make plenty of people on this forum jealous . . . But I don't have a PL statement that would make very many people jealous at all. One thing has a tendency to even out another.

Here's a question: do any of you who trade using closing prices use a 'line on close' chart as opposed to bars, candlesticks etc?

Yes, I can't imagine anyone wanting to trade a long term trend-following strategy gaining much from simulated trading. Or not unless they wanted to sim trade for about ten years.

I couldn't agree more, especially when it comes to intraday trading, as you say. Except that you missed a bit . . . In addition to the fact that you're trading 5mins, I'm trading 3mins, and Joe Soap is looking at a 20min chart, there's also this to consider: you and I could both be looking at five minute charts, but if my five minutes are staggered by, say, just one minute, then I'll be looking at a rather different chart to you. If price moved a lot in that 'sixth' minute, then your five minute candle will be a nice long one, and I'll have some crummy little doji. So I agree: intraday closing prices are a completely arbitrary construct in which a discrete time structure is imposed on continuous price data. And yet I only take entry signals on a bar closure. Why on earth would I do that? Because I'm working with a backtested strategy that uses the indicator value on the close. Another thing to consider: when a trader says 'this works on five minute charts', this isn't necessarily to do with the close. If it doesn't work on daily charts, then its just as likely that it would work on a five min chart using the low price, and wouldn't work on daily charts using the low price. This seems slightly logically inconsistent with my earlier statement - perhaps its because I haven't eaten all day . . .

This is a great suggestion of Tim's for an end of day exercise. Another way to use it is as follows: If you regularly trade intraday, then you'll most likely have forgotten the nuances of any given day's price action within a couple of weeks. Perform the visualisation that Tim recommends, then return to that previous week and use the cursor arrows to scroll through your intraday charts one bar at a time, and identify your setups (I believe some charting platforms - perhaps Sierra - actually offer high speed replays). Record your entries and exits. Now get out your actual trading record for that week and compare your results - did you do any better on the 're-sit'? What had you learnt? If you've got a particularly keen visual memory, then just wait longer before revisiting. If you've got a photographic memory . . . well then I can't help!