I encourage all other traders to follow their gut and instincts, along with a healthy dose of traditional technical chart theory and indicators, with minimal attention to trying to understand what the indicators actually mean, or whether chart patterns have any scientific basis. And if this approach doesn't work, then just be brave and keep throwing even more money at the market. I guarantee you that sooner or later one of your bets will pay off, making you feel it was all worthwhile. At the very least, you will have contributed to the efficient working of the market, keeping many other traders employed.
what you describe sounds like a classical over-training problem: one must not judge the performance of a mechanical system (indicator) using the same data-set that it was trained with. this in no way inflicts on mechanical systems in general. so instead of changing the marketing approach, the OT developers should have simply fixed their product. - jaan
Hi jaan, I've used OT for 3 years now in it's various versions. 3.0 -2001. Tested almost every kind of settings possible with the EOD versions. Don't know, whether you are familiar with this software. It's not really comparable with any blackbox-systems or the neuro-network kind of software, using "training periods" etc. OT uses 60 daily and 60 weekly standard technical indicators ( i.e. RSI, STO, MACD, trendline breaks in various timeframes, some candlestick patterns etc ) basically the ones you can find with MetaStock or any other standard-tradingsoftware.( Infact, OT was the first "automatic" MetaStock program ). In standard mode, OT runs a backtest for each available indicator against each stock in your watchlist and calculates the performance of each indicator for each stock. ( If you would do that within MetaStock, it would take ages ) Than OT selects the 6 best working indicators ( daily and weekly )based on filter conditions, takes the signals of each of these indicators and creates consensus signals ( long - short )on a "voting line". These consensus signals are selected by quantitative and qualitative analysis of their performance ( hitrate, annual profit ). Consensus signals are the final trading signals and the historical performance of this "indicator-combination" is displayed in hitrate ( how often did the consensus signals work ) and annual profit ( now much profit could you have realised, applying this indicator - combo on this particular stock in the past ) Like in TradeStation or MetaStock, you'll get a detailed summary of the performance of each trading-system and indicator for each stock. Of course, normally, there's no need to do a backtest every day after having updated the data. I just mentioned that for illustration purposes. Usually you only run a backtest in order to find better working indicators every month, or in volatile markets every 2 weeks. This sounds reasonable, because if a stock transits from sideway's action into a ascending or descending trend, oscillator systems start to perform poorly, while trend following systems can produce better results. ( and vice versa ). OT is, at least, able to find the best performing indicators - for the selected backtest period. OT follows the signals of the currently selected 6 ( or more )indicators until a new backtest brings up new, better performing indicators to use for the system. This means, the votingline ( which is the final tradingsystem, consisting of the currently selected indicators ) shows you every day whether to stay in the trade, follow reversing signals, or when one of your pre-defined stop-loss conditions has been triggered. You could use OT also in non-auto - mode : Just select your 2 or 5 favourite indicators ( be it patterns, MACD's RSI's etc, or a combination of all of them ) and let OT use only these for signal generation.( means force to vote for a signal, even if your choice of indicators would not generate the best possible performance ). It will still generate trading-signals on the voting line ( if there are any produced by the indicators you selected ) and will not change the "tradingsystem" unless you run a backtest again with other indicators added. In addition, you can apply several automatted studies on OT charts , in order to get info about support & resistance landscape, certain patterns, short to longterm trendline breaks etc. The portfolio module allows you to keep track of your trades directly in OT. In this regard, OT offers completely mechanical tradingsystems, consisting of rules / indicators defined partially automatically by OT and partially by the trader himself by entering his stop-conditions, slippage, comissions, entry & exit delay conditions, selecting performance filters, types of indictors to use, time-rame of trade ( short to longterm ). This sounds all very good and is certainly of much help for any novice trader, who's looking for an organized, mechanical trading approach. But relying only on this type of mechanical tradingsystems would be not possible for me, since I'm simply not the kind of personality who accepts mechanically generated trading- signals whithout questioning them. I use OT for creation of trading ideas, because it signals opportunities, draws my attention to stocks, which I may otherwise have overlooked. But that's about it. Maybe, another trader, who likes this kind of mechanical trading approach, may find with OT exactly what he was looking for. Just trade the voting-signals, stick to his stop-loss an be fine. Could work - I don't know.
Forgot to mention : For those new to mechanical trading, looking for a gameplan or education on how to apply certain technical indicators, support & resistance etc. within a gameplan, have a look at the new Sixer.com website before you commit money to expensive software. ( 14 day free trial , after the trial, very reasonably priced subscription - I think it's < 10,00 USD/month ) Although sixer covers "only" 600 stocks, their screening tools, stockreports and gameplans for different trading styles are worthwhile a look. http://www.sixer.com Example stock report : ( of course, it looks much better on their website ! ) Just took AOL for this example - no particular reason - "AOL (America Online Inc. ) 8/23/2001 4:01PM Exch.: NYSE Sector: Technology Industry: Computer Ser Last: 40.08 Change: .58 Percent Change: 1.47% RS Rank (13wk) 34 RS Rank (3wk) 34 Open: 39.00 High: 40.22 Low: 39.00 Volume('000): 9616 Today's Range: 39.00 to 40.22 52 Week Range: 31.50 to 62.27 siXer analysis : Support:36.75 Resistance:44.48 Breakout:47.70 Breakdown:37.55 Volatility:1.84 siXer Indicators: America Online (AOL) is showing Bear(Mod.) trend in Short-Term time frame and Neutral trend in Long-Term time frame.. AOL posted Inside Day with decreased volume. Last bullish alert was given on Jul.26. Last Bearish Alert was given on: Aug.09. MACD is below the centerline and trending up. AOL closed below 13-Day MA (41.69) and below 50-Day MA (47.00). AOL 21-Day Range is 38.05 to 47.20. RSI is:31.66 and trending up. Volume %Change is:-30.55%. Day Trading : Daily Pivot/Support Pivot 39.77 Support 1 39.31 Support 2 38.55 Resistance 1 40.53 Resistance 2 40.99 Med/Long Term trading : Entry/Stop/Obj. Close 40.08 Entry 40.28 Stop 38.46 ST Obj. 42.99 LT Obj. 50.34 Fundamental analysis America Online is the world's leader in interactive services, Web brands, Internet technologies, and electronic commerce services. The company operates two worldwide Internet services, America Online and Compuserve, several leading Internet brands, including ICQ, AOL Instant Messenger and Digital City, the Netscape Netcenter and AOL.COM portals, Netscape Comunicator and Navigator browsers, and AOL MoviePhone, the nation's largest movie listing guide and ticketing service. Market Cap: 208,344.10M Book Value: 35.12 Revenues: $7,703.00M Beta: 2.58 P/E: N/A Earnings per Share (EPS) -.03 Whisper Number .2200 Earnings Date: 07/18/01 Analysts as of 08/17/01 Avg. Recomm: 1.4 (Strong Buy) Avg. Target Price 65.23 Strong Buy Buy Hold Moderate Sell Sell 19 9 1 0 0 Technical analysis Targets 51.2 [Project3] 47.0 [MA50] 46.7 [Project2] 46.6 [MA200] 42.3 [Project1] 41.7 [MA13] 40.1 [Price] 37.8 [Retract1] 35.7 [Retract2] 34.4 [Retract3] Price & Volume Action : Price AOL closed up .58.(1.47%) at 40.08. 21-Day Range: 38.05 - 47.20. 52-Week Range: 31.50 - 62.27. Day Type: Inside Day. Volume 9386600 shares traded. Volume made a -30.55% move down. Moving Averages: (13-Day MA: 41.69, 50-Day MA: 47.00, 200-Day MA: 46.61) [Moving Averages act as Support and Resistance levels. They help measure the direction of the trends and to determine if trend change is taking place.] Stock is trading -4.01% below 13-Day MA and -17.26% below 50-Day MA. With its 13 and 50 Day Averages sloping in negative direction, the stock may trend down. There are no Moving Average crossovers. Relative Strength Index (RSI): 31.66 [Relative Strength Index measures the buying/selling pressure of a stock.] RSI rating of 31.66 suggesting the stock may be in neutral zone as 14-Day RSI Direction is up. MACD Signals: -2.12 [MACD and ADX measures the trend of a stock.] MACD is positioned below zero-line and moving up. Momentum: -4.77 [Momentum measures the current price trend of the stock.] 10-Day Momentum is at -4.77 and positioned below zero-line. Current position of Momentum is in-line with the general current trend of the stock. Stochastics: %K: 20.47 %D:20.71 [Stochastic oscillator compares a stock's price relative to its price range over a given time period.] SlowK is at 20.47 and is in Ovesold condition (down direction). ADX: 32.63 DMI+:21.26 DMI-:30.20 [ADX measures the trending quality of the market. ADX gives whether the market is moving directionally or not. DMI+, DMI- components of ADX, can aid in identifying direction.] DMIPlus is trading Below DMIMinus and is in Bearish condition." Usually, one checks this report after either having screened for particular trading signals / patterns, or after having used results of sixer pre-screened stocks via their "active gameplans" - as there are : Daytrading, swing-trading, Longterm stocks, breakouts, breakdowns, shorts, chartwatch, morning watch Hope this ( plus site content ) gives the newbies a brief impression about how to apply technical indicators in trading.
Privateer, thanks for thorough explanation of OT. well, i think it is. "training period" is not specific to neural nets, it is just an alternative expression for "data used for optimizing the system". hence, if you say: then this effectively means that the period of historical data used in selecting the best indicators becomes the training period for the system. as such training will inevitably gain retrospective hindsight into the training data, the performance numbers for the training period will not be representative of the actual expectancy of the system. in order to get the actual expectancy of the system, it is necessary to run the system (ie, in OT case, the set of best indicators) against data that the system has not "seen" before. i hope that did make sense - jaan
Hi guys This is something my father Dr. Van K. Tharp -had studied quite extensively. (I know a lot more than 2 traders who are consistently profitable).... What he discovered is the very intuitive traders took years to become that way. The developed a sixth sense about the market, an unconscious competence. To gain that type of skill though takes years of practice. I'm finally starting to get that level now. What He recommends is for a trader to start with a fully mechanical system with certain rules which can be backtested. There are quite a # of profitable traders who are purely mechanical. The intuition comes after seeing the market for a long while and knowing when to alter the system or use a different system. There were quite a few pages about this in his Psychology Course he sells to traders. Rtharp http://communities.msn.com/rtharpsland
i fully agree with you jaan, testing against out of sample data is the only way to ascertain the viability of a system. i also agree with with rtharp that a mechanical system makes sense under certain circumstances such as an add on, to be able to follow more instruments, for quantitative methods etc, however i disagree when you say that a mechanical system makes sense for new traders. i believe that the only people who can make a mechanical system work are those that also have the ability to trade on a discretionary basis, ie they really understand markets as opposed to fooling around with optimised parameters on tradestation etc. would also add that every system is nevertheless discretionary in the sense that the money management part of your applied risk management utilised in determining your position size is always determined by the trader, and not the system. and that is the most important size, as that determines your overall profitability. cheers
Hello jaan, indeed, what you say makes sense. It's possible to do exactly that in OT. You can define backtest and forward test periods ( either by exact date or by using the number of trading days involved in each test period ). Indicator-optimizing can be restricted, if desired, in several ways : - by turning the optimization function for each individual indicator off, - discard the performance filters. - turning any "confirmation filter" off - forcing all selected indicator's signals to the voting line, regardless of their performance. OT would, in this case, test all selected indicators ( standard = all 120 ) and would create voting signals by just using the direction of the majority of signals generated. i.e. : 70% of the indicators generate a long signal or are in a long trade already, 30% generate short signals or are in a shorttrade - the final voted signal would be "long", since the majority of signals is long. It's simply a quantitative filter for generation of voting signals, regardless of the performance of each indicator. You could also set this cut-off level at 60% or 50% - and add some " confirmation - filter ", like "daily signals have to be confirmed by weekly signals" or "drop daily signals which are in conflict with weekly signals" or with "voted weekly siganls " and so on. Say you let OT "look back" 750 trading days of data to find the best performing indicators for this period. Than you "forward-test" this set of indicators on 250 days ( just an example - you can set any period you like, provided, you have sufficient data ). You'll get the statistics how the systems have performed during backtest and forward test. As usual, most "systems" which fared well in backtest, don't produce any satisfying results in the forward test period. This is not surprising, since we all know, that no one indicator or trading system works well at all times and for all markets. For what concerns those systems who actually did well in both periods : These succesful systems consist only of a few ( mainly 3 -5 ) standard, time-tested indicators, like Bollinger Bands, Stochastik, MA's, MACD, and trendline breaks in various timeframes. And the underlying market hasn't changed it's personality very often during the back & forward test period. BTW - in MetaStock, there's a tradingsystem called "The ultimate System" ( not the same as the Ultimate Oscillator !! ). It has been developed just for statistical and comparison reasons - it's not tradeable as such. This system is the best proof of what you can achieve with over-optimization of technical indicators: If you backtest this system on any market, it finds every possible trade in the past, enters and exits all trades with perfection. This system would make gadzillions of money with each and every market you like to trade - only in theory of course. Optimizing in OT refers mainly to selection of those indicators which perform best with the current "personality" of the stock. If this "personality" changes, the currently selected indicators become obsolete and have to be replaced by better ones ( new backtest ) The longer the backtest period, the less different indicators are selected for a tradingsystem. That's also usual, since only a few basic indicators can be considered as "time-tested" on the long run. The performance of these indicators, over time, levels out at the 45% - 55% hit rate ( 50% HR means, 50% of generated signals were followed by a corresponding move in the stock-price ). My personal "hitrate" = profitable trades, on hundreds of trades in the past 3 years is still at around 62%. Many of those trades , were based on OT entry and exit signals. Unfortunately, in the beginning of my trading carrier, I had no idea about proper positionsizing, money-management, support & resistance levels etc. So I made not enough money on my profitable trades to cover the losses. Typical experience for a beginner, I'd say. Why is that so ? because a system might signal you an entry, or an exit but it doesn't tell you, how much money to commit, which is the best entry / exit price and where to set the stops with regard to the underlyings volatility, is it a shortterm or a longterm signal and so on. Performance ( or profit expectancy for that matter )for a tradingsystem is, of course, depending on the size of the pricemoves itself, the capital committed to each trade and your personal stop-loss preferences, setup in OT before running the back/ forwardtest. When using "standard" OT setup , which means a stop-loss at 3 times ATR or reversing signals, you'll find back & forwardtest performances for many volatile nasdaq stocks in the range of 150% - 300% APR. But you might have to sustain drawdowns of 30% - 50% when trading these systems. Shorter back/forwardtest periods and shorter timeframes of the selected "trading-model" i.e. medium-term trading, tend to use more "responsive systems". Forwardtests perform also considerably better with shorter periods compared to longterm test-periods. Medium term models are used for trading periods of up to 32 days, longterm for 64 days and over. Shortterm is up to 16 trading days. Well, one has to consider also the number of executed trades before judging hitrate and performance of a system. 100% HR on 1 executed trade is not a very satisfying with regard to reliability, while 85% hitrate on 10,20 or more trades, long or short, back & forward-tested, points to much higher "stability" of the system. Anyhow, once you have tested ( back & forth ) all kinds of possible trading-setups, stop-loss techniques, optimization , non-optimization etc. pp. on thousands of stocks and in various timeframes, you'll end up with the conclusion, that any "sound,timetested" mechanical trading system isn't finally any better than tossing a coin, even when applying it to markets which tend to work well with technical indicators. Trading such mechanical systems may give you the impression, that everything you do related to this trade is totally rational and no emotions are involved at all. But I'd bet, that 95% of the traders, who seek for this kind of "systematic approach" just couldn't execute it on the long-run and start to spoil those systems efficiency by setting either unrealistic profit objectives or to tight stops ( or to wide stops ) or the positionsize they trade on a particular system is not appropriate. Once in the trade, the "daily news effect" may override the previous confidence about the validity of the trading signal and influences the feelings about this trade, although no reversing signal has occured yet. These feelings may lead a trader to bail out of a position pre-maturely - just to watch the stock doing afterwards exactly what the system "indicated" before. After all, we are human beings, not yet robots or cyborgs. I have yet to see / meet a trader who makes consistently money, over years, by trading a purely mechanical system with no discretionary decisions involved.
Privateer, when you wrote: "My personal "hitrate" = profitable trades, on hundreds of trades in the past 3 years is still at around 62%. Many of those trades , were based on OT entry and exit signals. " Does the 62% refer to all your profitable trades, including those that made small profits which simply offset your losses? In other words, are you actually getting results more like 1/3 small losses, 1/3 small gains, and 1/3 significant gains? Or do you do significantly better than this?