It is up to the subscriber to determine the number of contracts. What happens if it reaches $500,000, should the system stay with no more than 10 contracts?

The equity curve so far shows excellent control by the vendor.

To create a realistic track record in C2, system vendors should apply certain restrictions in systems, somebody trading with real money would apply.
In case of this system, several posters and subscribers tend to agree, that not only the total number of contracts, given account size is too high, but also given the trade times of the system, the number of contracts used seems unrealisticly high.
This obviously windowdresses the return in C2 and at least looks questionable for real life trading.
Certainly risk levels are different, but certain money management and market rules still apply.
Just to create attention with a new system, which trades the way this system trades, is fine as long as subscribers are aware of these facts.
That systems, which have compounded returns are trading more contracts is certainly suggested, but it makes a big difference, at least in my understanding, whether we have a system trading very liquid contracts in peak hours, or a system, which trades outside regular trading hours market untypical size for a scalping trade.
My point is simply that, the system performance posted in C2 and real trade achievable results, in this particualar system are 2 totally different animals.
And as MK clearly stated, the trade size is only reflected in the RF of C2, but not in the results posted.That could surely create a lot of wrong expectations among potential subscribers.
Imagine that 10 subscribers trade 5YM each, all of a sudden, you might see a lot of uuppss in the market, once the famous autotrader software converts limit orders into market orders.

if the overnight volume does not reflect reality, then C2 should be inflicting punishment via RF. Systems that flaunt reality don't tend to last. But saying that the #contracts should never exceed 10 is not a valid request. Personally, my own backtesting on various futures systems throughout the years seems to suggest using 30-35% of account funds for positions held overnight. That seems the sweet spot between acount growth and disaster prevention. And that is using several uncorrelated futures...

Several vendors have recently stated that when they start modifying their systems according to various subscriber inputs (Renato I think was one? Don't remember), that their confidence and performance tends to suffer.

I think vendors can consider the input, but feel free to throw it away if not interested. A subscriber has the right to follow a system, but the vendor bears ultimate control of the signals. If a subscriber is not happy, then there are supposedly 500 other systems there to take the money. There are few enough reliable ones.

This one has at least some promise, for the moment. But the likelihood seems like sometime between 2 weeks and 2 months from now, it crashes. At least, that seems to be the C2 pattern.

Some very valid points there Ross, which i like to comment further.
1.I would immeadiatly stop any subscriptions once a system vendor starts to accomodade subscriber requests.
2.Your point of crash and burn, often observed, matches exactly my previous argument.Since system vendors want to create attention in C2, they overleverage to produce returns and a steep equity curve.Than the usual loosing trade comes in, sooner or later it will for sure,and the overleverage which collected a few points everyday for the records gets wiped out with one single loosing trade.Than everybody jumps on the system and states:
A another crash and burn system.............

In my opinion system vendors, who try to windowdress with high overleverage their track records, must be punished that way, but on the other side, as this doing creates false expectations on the upside, it also does on the downside and might exclude good systems from potential subscriber due dilligence.

Example here:
Lets get real.............which trader would trade in an outside regular market hour system 25 or 50 YM?
And if so, would that be in a 100k USD account to just scalp 5-10 points.
So, as Ross mentioned, 30-35% margin use, is the maximum, often used by future traders and this is already aggressive.
In case here...........a little less aggressive leverage due to order types, trading hours and liquidity vs profit target.
Lets use on a 100k account 5 YM contracts maximum.

Outcome................much much lower returns, and probably no attention in C2, but Ross mentioned C2 phenomena of crash and burn, would be much smaller.
It seems to me, that system vendors, who trade real life leverage in C2 get little or no attention, which leads to the other famous discussion, where have all the good systems gone .......
They are here in C2, but get little attention, because these system vendors trade low leverage and never make it on the hot systems page of C2, but this was not their intention in the first place.

I disagree somewhat.

Matthew has systems decently ranked so far as I can tell, with some obvious exceptions that creep in via calculation errors.

Annual return does not seem heavily weighted in any of the "best" categories. Please point me to any system you have seen that displays a poor annual return (by C2 Fantasyland standards), but is good enough in other categories to be tradeable, which isn't already ranked highly on the "best" lists. I'd definately like to check it out.

Rather, I think what everyone is seeing is that it's extremely (Extremely!) difficult to design and implement ANY profitable trading strategy. The smartest people in the world aren't working at NASA -- they're trying to figure out how to beat the market tomorrow.

This is just a fact of trading life. Systems are going to come and go here at C2, with maybe 1 out of 5,000 worth trading. I've only seen two systems I would allocate capital to in the last six months. Both trade the ER2, which means they can only support a small number of contracts before slippage eats away profits.

I haven't seen a single system that trades only very liquid markets that would be worth trading.

"WhyNot" is close. However, it took all of 5 minutes to reverse-engineer that system to within 90% accuracy. Having done so, I see many problems and I would never trade it. Most importantly, it doesn't take reversal signals. WhyNot? If the logic is sound, reversals should be taken. Otherwise, it's wait and hope combined with optimized testing. Additionally, the bond contracts are some of the most efficient markets in the world. They don't lend themselves well to technical trading over the long haul.

"Additionally, the bond contracts are some of the most efficient markets in the world. They don't lend themselves well to technical trading over the long haul. "

I am convinced that the best place to trade is generally the stocks/futures that are neither the most liquid nor the least liquid.

For example, in any system I have subscribed to or personally created, the S&P/Nasdaq price action behave one way, while meats/softs/grains behave in another. Since technical analysts almost seem blind to the underlying instrument (price action & indicators alone), this seems partly explainable by the popularity & size of the S&P / Nasdaq markets. For every 50 S&P systems, there only seems to be 1 or 2 that trades physical commodities. Making regular profits in the index futures is a difficult thing.

On the other hand, some here expound the inefficiencies of really illiquid market. But as that overnight 50 YM trade shows, who wants it? Slippage can be a killer in these markets. If you scalp $50 on a brief trade, why spend $30 of it on slippage and commissions? It is hard enough to make the $50 in the first place... When I enter a trade, I don't want to be sweating the fill due to a wide bid-ask, IN ADDITION to the other issues surrounding a trade. Who needs the extra grief? Like, who would like to fill a 30-lot trade in the lumber futures pit? You have to wake up the market makers first!!

And you should see the volume difference between pit & electronic Feeder Cattle. The electronic version is like watching the heart monitor signal of a freshly-dead patient.

My real preference, is the major physical commodities. I only have to follow a few dozen. With pure stocks, there are thousands, and I don't have the energy to chase them. Commodities gives me easy entry/exit and all the leverage I need. Forex trading I have never tried; I stick to currency futures.

An individual investor's research on a company has more value if no one else is paying attention to the firm. According to data presented by Jeremy Grantham, illiquid stocks outperformed liquid securities by 2.1% per year for the last 40 years, and Warren Buffett wistfully stated several years ago that he could still compound returns at 50% per year if he could run a (relatively tiny) $1 million portfolio.

My conclusion is that 'value investing' is a classic strategy that has passed the test of time because values are not determined by fiat not by majority vote.

ps. Just as the number of adherents is not a proof of an idea's truth or falsehood, of an art work's merit or demerit, of a product's efficacy or inefficacy - so the free market value of goods or services does not necessarily represent their philosophically objective value, but only their socially objective value, i.e., the sum of the individual judgments of all the men involved in the trade at a given time, the sum of what they valued, each in the context of his own life.

Dustin:
All you said (Most importantly, it doesn't take reversal signals.) have no sense.
I know I could "reverse-engineer" your system and all of the best systems here on C2 but the very important things is manage a trade and a system.
Ciao

I know I could "reverse-engineer" your system and all of the best systems here on C2 ...

Please attempt to back up this ridiculous statement, thanks.

I agree managing a trade is equally as important as trade entry. However, I don't see how setting a profit target and stoploss, then walking away counts as "managing" a trade. Please clarify your statement. Your trade management seems nothing more than "wait and hope".

And for non day-trade systems: for reversals, if system logic is sound then signals will be taken in the reverse direction of open positions as they occur. Why would the most recent system signal not be the most valid? This WOULD in fact be some trade management beyond waiting and hoping. The only reason you don't take reversals is because it works out better that way in back-testing. There is no logical reason -- in fact, the reverse is the case. This is an example of nothing more than curve-fitting to past data.

Dustin:
hear my advice, maybe you have to change work, you could be a good prophet :-), You could gain more money doing it.
ciao

I don't know if Mario meant reverse engineer as if HE COULD, or that EVEN IF HE COULD

Obviously, some systems are based on discretionary trading and there is nothing to reverse engineer...

>This is an example of nothing more than curve-fitting to past data.

I agree. I have no doubt that all systems at Futures Truth and C2 are curve fitted. In my mind, curve fitting means either using different systems for different markets, or using different parameters of the same system for different markets, and this is not valid technical analysis. Instead, one should trade the moves, rather than markets. Some traders hold on to a position, and keep changing their systems to fit it - other traders hold on to their systems and keep changing their portfolios to fit it.

If a system works on Bonds and not on Beans, this system is curve fitted over a specific set of data (Bonds) and it loses all statistical validity. To believe it will work in the future as it has worked in the past is very dangerous. I therefore take exception to any system, that either only trades one specific market (stocks or forex or bonds or beans) or group of markets (Energy), or trades different markets using different parameters or rules of the same system. All this proves is what has worked best in the past, and this will usually not continue to work in the future, as there is no correlation under this scenario as history wont ever repeat itself exactly.

ps. I meant most systems at C2 and all systems at Futures Truth.

"If a system works on Bonds and not on Beans, this system is curve fitted over a specific set of data (Bonds) and it loses all statistical validity"

Not at all. I find systems will work across physical commodities, but behaves very differently on index equities and vice versa. What works on S&P seems not work on Corn. However what works on Live Cattle is often possible with Platinum...

If a system works on Bonds and not on Beans, this system is curve fitted over a specific set of data (Bonds) and it loses all statistical validity.

This is the most false thing I have never heard.

yes, isnt it

Hahah!

This is the worst response to the best post I never have seen.

> If a system works on Bonds and not on Beans, this system is curve fitted over a specific set of data (Bonds) and it loses all statistical validity.

Again you seem to be lost in your own theoretical abstract. Of course trading MSFT isn't the same as trading Corn.

BTW, some of the best stock and index traders use breadth, trin, tick, relative strength, inter market correlations, etc., etc. etc. These items are unique to stock and index trading. This has been done long before Pal ever placed his first trade..... wait, has Pal placed his first trade yet? I couldn't help but notice he was trading December indexes back in July. Trying to save $1 in commission while risking 100's in slippage? This is the mark of someone who has never traded in real time.

On that note, I see many C2 index traders failed to "rollover" into December last Thursday. This always makes me wonder if they are actually trading since most brokers call or email out Wednesday prior to rollover. Folks trading the big ND and SP will see near zero volume and a host of restrictions after rollover.

In my mind, curve fitting means either using different systems for different markets, or using different parameters of the same ssystem for different markets, and this is not valid technical analysis. Instead, one should trade the moves, rather than markets.

Historical testing via computer means feeding a set of numbers (open, low, close prices), and receiving back an output set of rules that hopefully will make money trading. The numbers themselves do not have names, and the computer doesn't recognize the difference between 'Beans' or 'Bonds'. For a system to be valid, it must work on all numbers tested, not just those with certain names and not others with different names.

If a system works on Bonds and not on Beans, this system is curve fitted over a specific set of data (Bonds) and it loses all statistical validity. To believe it will work in the future as it has worked in the past is very dangerous.

Also, different markets do not have different personalities. Again, they are reduced to just being a set of numbers or a bunch of algorithms. If a channel breakout (or any other) method is successful, then the same parameter must be used for all the markets, for the same reasons as above. You cannot use a 20-day channel in Silver and a 40-day channel in Corn, this also falls under the crime of curve fitting.

I therefore take exception to any system, that either only trades one specific market (stocks or forex or bonds or beans) or group of markets (Energy or Grains or Cattle), or trades different markets using different parameters or rules of the same system. All this proves is what has worked best in the past, and this will usually not continue to work in the future, as there is no correlation under this scenario and history wont ever repeat itself exactly (though those who forget history are condemned to repeat it.)

This is not specifically written to condemn any particular system vendor. This is a clarification of my definitions of 'optimizing' and 'curve fitting', and a warning as to what types of trading systems may be valid and what to stay away from.

Although I have never personally worked with any of the systems covered by Futures Truth, I have no doubt that they are all curve-fitted. Any 'system' that purports to specialize in one market is optimized for that particular set of data.

Some people will say that different markets have individual characteristics or personalities. This may be true to a limited extent. However, in testing, a computer doesn't 'know' what market it is examining. All the computer knows is a bunch of numbers (highs, lows, closes), from which it attempts to produce an algorithm to explain or predict price behavior.

For a system to be valid, it should work over a given set of numbers (data). Whether those numbers have a name such as 'Beans" or 'Bonds" is (and should be) irrelevant to the data and to the testing program.

Lots of systems make money when they trend and lose money when they don't. This is not surprising. The best that you can hope for is to create a system that is profitable over time over a wide range of markets/asset classes. Systems such as the Turtles use, makes money when the markets trend and loses money when they don't (no surprise).

Since trendiness is a proven characteristic of commodity markets, given a long enough sample period (i.e. 20 years) almost all the markets yield positive results.

However, in any given year, since there are only a few good trends, most of the markets will prove unprofitable. This is not a reason to abandon the system, or to eliminate (temporarily) unprofitable markets from the portfolio. In fact, the markets that have lost the most money recently (due to being in a consolidation) will probably be the best in the future (when they finally hit a trend).

Trend following is basic to life. Sir Francis Bacon (1561–1626)noticed runs and consolidations 400 years ago. The aim of diversification is to cancel out the short term noise and enjoy the overall signal.

1. By definition, if you make money in the markets, you are on the right side of a trend. No trend, no profit, period. So it comes down to how do you get on and off a trend at the right time, and how heavy do you bet so you don't stub out during the corrections.

Buffett uses fundamentals ... and missed the great run up in tech stocks. Of course, you can use many methods to miss markets, so I don't think he has any unique ability in that area. Mostly, Buffet has a great attitude; he has the seemingly infinite patience, courage of, well, a Warren Buffett. Plus he has a secret advisor in Charlie Munger.

2. There are at least as many ways to approach the markets as there are traders. I feel every successful strategy has to exploit some or another trend.

Trend followers look for trends directly in the price (and get some whipsaws).
Fundamentalists look to predict inevitable trends by looking at underlying factors (and are often early and / or wrong).
Arbitrageurs and swing traders feed on small imbalances (and sometimes find out these imbalances are just the start of a bigger trend.)
Contrarians look to take the opposite side from the public (and sometimes find the public just keeps the trend going and going.)

2) Also, does a trend follower need to be in the market all the time in order to catch the big move?
2. No, he just needs to be in for the big moves, and it's also nice to be out during choppy whipsaw markets.
3. The trend following strategy does not anticipate anything. The trader might anticipate a top or a bottom, and stick to his system anyway.

Well, if it's on the news, then you can bet on it. Of course, you might very well lose your bet, or miss a good one. Trend followers do not trade in anticipation of anything, nor do they try to figure things out. They simply go with the trend.

Figuring-out and anticipating events are things that fundamentalists, and weathermen do. They engage the processes of analysis and anticipation, with similar results. The best ones learn simply to predict the trend will continue for a while. Long-term trading has an advantage, in that the transaction costs are small relative to the average move. But, some traders might find it difficult to sit tight through prolonged corrections.

Sounds like you are coming around to the idea of trading moves, rather than markets. Some traders hold on to a position, and keep changing their systems to fit it - other traders hold on to their systems and keep changing their portfolios to fit it.

Most markets creep along most of the time and then make a nice move, sooner or later.

Finally, If one could learn to tell when the markets will trend and when they will be in a trading range, they wouldn't need to know much else to make money. One can visually eyeball a chart and tell if it's in a trend or consolidation, but that still doesn't tell one much about the future; a valid case for expert judgment (based on a strategy) in trading and investment.

Happily, there is some wisdom in here. If you spoke to the issues and offered your analysis and stayed within the realms of cold market logic, dialogues would be a lot easier.

Reality is, it is all about making consistent money over time, without getting your arms hacked off. Everyone wanders around with theories and beliefs, but if they cannot generate a decent looking equity curve, then I don't care whether they are Elliotticians or Gannites or Fibonacci lovers.

I trade futures/commodities of all types. But traders have to wander around in a sea of manure (spam, systems, hype and other garbage) generated by believers in weird theories. Show me the good equity curve, and you have my attention. Otherwise, they are no better than a cult.

My specialty is this. I have done exacting, hard-nosed statistical testing or close scrutiny of all the systems that seem to have been worth testing over the last 10 years (As I said, I am somewhere between 800-1200, not counting C2). If it appeared in Futures Mag, Tech Anal of S&C, Futures Truth, Striker, email/postal mail advertisements, or other places, I have run them through the ringer. I got to the point that I can tell you what is wrong with a system just based on its wording, presentation, or other facets. That is one of the reasons I am within C2. It makes it a lot easier to flush the fertilizer.

I am 98% skepticism and 2% optimism.

From a statistical point of view I would say that curve fitting is not a crime per s'e. It is over-fitting that is a problem. That is, if the model has more parameters than needed, then it is likely that these parameters will accomodate idiosyncrasies of the sample that cannot be generalized (to the future, in this case), thus making the predictions worse than would have been the case without these extra parameters. So having different parameters for different markets is only over-fitting if these markets actually behave according to the same parameters. If the markets actually behave according to different models or different parameters, however, then forcing them into the same model with the same parameters would be under-fitting - which can as well be a problem.

I have no idea which of these two situations is a better description of the reality of markets. From a statistical point of view there is no reason to abandon any of these models a priori. From a practical point of view, I must admit that most people tend to be more vuldnerable to the possibility to over-fitting than to under-fitting, so I think that it is generally wise to be conservative with introducing new parameters. But it is not a logical necessity that a market-specific model is bad.

I would say that the proof of the pudding is in how good the (forward tested) predictions are.

> From a statistical point of view I would say that curve fitting is not a crime per s'e. It is over-fitting that is a problem.

Yes. Indeed, Pal's own example of trend following needs at least
*a* parameter even in its simplest form: buy the breakout of X highest highs.

The nice thing about TradeStation, or similar programs, is you
can run an optimization across 20 (or whatever #) inputs and see how
they all would have done. If all 20 made good money, and a few
are a little better, than likely you have something worth a second look.
OTOH, if a few inputs make good money and the rest lose, you just
curve fitted those few #'s.

When I meant "curve-fitting," I meant the same as "over-fitting" or "over-optimization." A lot of traders fall into the trap of curve-fitting or over-optimization without being aware of it. Curve-fitting is like rolling a fair dice with 1/6 probability of getting any number from 1 to 6, rolling it 5 times, getting #6, 4 out of 5 times (80%) of time.

There is a correct method to optimize any system that is statistically valid, 30 occurrences with 95% accuracy.

The key to optimization is to select stable parameters with an equity shift less than the parameter shift after equity spikes have been eliminated. This process creates stability for optimal parameters shifts within the technical market phases. Parameter shift is always geometric, but equity shift decline relative to unstable parameter selection is usually exponential.

All systems are optimized to some degree. As soon as a trader chooses to enter a trade on the open as opposed to the high/low/close of day, he has made a decision as to how a system should be traded. Does he know the high/low/close of day entry is better than the next opening for an entry? If not why not? A potential 28% difference in profitability exists for channel system entries between opens and closes.

The purpose of trading is to consistently make money. This is done by having the best information available. If a trader does not know the best entry for his system, what is he trying to prove? That the system isn't optimized? To lose money because a trader is ignorant of his system's best parameters is foolish.

But, a lot of traders fall into the trap of curve-fitting or over-optimization
without being aware of it.

Attempts to model the markets by applying sophisticated technology such as neural networks in a simplistic manner does not work well. Present-day attempts to apply the neural-net novice approach fail because the market inefficiencies discovered by the simplistic application of sophisticated technologies have already been "traded away" by speculators who used the technology to discover and trade the inefficiencies.

Instead, a rule-based approach is ideal and may also be a "mind-based" one, relying on the individual constructing the system/method. Once observations have been made, these ideas may be formalized into computer programming instructions.

Also, check out Oct 2001 Futures Mag article "Walking a Fine Line" where the following quote is found:

"Enhanced returns were found among optimization strategies that selected the worst performing input for the prior period. For example, regardless of the criteria used for optimization - average trade or net profit - the worst input for the shorter intervals significantly outperformed the best performing input in the prior period and the average of all inputs."

What are the differences between "out-of-sample" testing and "walk-forward" optimization?

I find the terms "out-of-sample" testing and the "walk-forward" method used loosely, sometimes interchangeably. To me, "out-of-sample" testing consists of taking historical price data, optimizing a trading system on part of it, and then evaluating the performance of that optimized trading system on the remainder of the data which the trading system hasn't previously "seen." Another form of "out-of-sample" testing consists of evaluating the trading system on indices and funds other than those it was optimized on.

The trap of over-optimization or curve-fitting can be avoided if one has an
adaptive or self-adjusting system, for eg., a parameter value that is related to an ATR. Periodic reoptimization often does not work because of this trap of over-optimization or curve-fitting and the alternative is to use adaptive parameters. Dozens of indicators, trading systems and chart patterns with adaptable parameters, makes a system/method (with its ability to adapt itself to any market conditition) to be truly versatile leading to the definition of the term "robust."

The fact that "a" parameter is chosen does not lend evidence to curve-fitting, because the parameter could be chosen with the "pocket book" size in mind, for example, to withstand drawdowns, not because the results were excellent.