How much analysis is enough? Does complexity equal profitability?

Nick Radge said:

ASXG,
You've done a lot of work on Edge Ratio. Its not something I agree with specifically in regard to Curtis Faith and the Turtle system. What's your thoughts?

Click to expand...

Yeah, I've coded it up and Amibroker, and worked with it a fair bit in a losely structured way.

I think that the intention is good, that being to compare entry effectiveness in a statistical and scientific kind of way. But based on what I've come to believe about the purpose of a good entry, I think it's usefulness is limited to identifying the short-term effectivness of a good entry 'trigger'.

If the purpose of a good trigger is to enter a trade at a point when you are most likely to drag your trailing stop up past 1R, breakeven and beyond, then some MAE/MFE analysis represented as an e-ratio can be an effective way to see whether during the first weeks of a trade being open, if one entry trigger and/or set of parameters was historically more likely to achieve this than another.

Framing a good entry 'setup' is more likely to help with identifying which markets are likely to sustain a trend though, IMO.

theasxgorilla said:

Yeah, I've coded it up and Amibroker, and worked with it a fair bit in a losely structured way.

I think that the intention is good, that being to compare entry effectiveness in a statistical and scientific kind of way. But based on what I've come to believe about the purpose of a good entry, I think it's usefulness is limited to identifying the short-term effectivness of a good entry 'trigger'.

If the purpose of a good trigger is to enter a trade at a point when you are most likely to drag your trailing stop up past 1R, breakeven and beyond, then some MAE/MFE analysis represented as an e-ratio can be an effective way to see whether during the first weeks of a trade being open, if one entry trigger and/or set of parameters was historically more likely to achieve this than another.

Framing a good entry 'setup' is more likely to help with identifying which markets are likely to sustain a trend though, IMO.

Click to expand...

Some good thoughts there ASXG.

I would like to know your opinions as well Nick. This E-Ratio has given me alot of curiosity since it was revealed.

I particularly appreciated how Curtis managed to validate this concept by running a random entry through different timeframes and somehow get an edge ratio of roughly 1.0.

ASXG, do you think you can repeat Curtis's exercise and see if you get similar results?

theasxgorilla said:

Framing a good entry 'setup' is more likely to help with identifying which markets are likely to sustain a trend though, IMO.

Click to expand...

I suppose that could include earnings growth in a given company, as in Shane's example above.

Temjin said:

ASXG, do you think you can repeat Curtis's exercise and see if you get similar results?

Click to expand...

Probably not. My testing has been exclusively on ASX equities...and his test was on futures. Futures, from what I understand, have a tendency to trend in both directions quite happily. Equities take a long time to climb up the stairs and a short time to fall out the window. Any testing on equities during the last 10 years has a much higher probability of sampling random entries during the bull phases, as that's what the market has spent the most time doing.

bingk6 said:

Putting it another way, if I were to generate 1000 different random entries. how may traders would be prepared to say OK, I'll run with these 1000 random entries and would be confident enough to say they would expect their performance to be on par with their existing systems ? I suspect not many. So therein lies a potential problem. Therefore, for the traders that truely believe that trade management and exits are the key and that entries (via T/A or F/A or voodoo or whatever else) offer no better than 50/50 (and I include myself in this category) why would they not be prepared to trade on random entries ???

Click to expand...

I basically would (and actually do with real money). The entry for my long term trend following system is laughably simple and barely different to random entry.

Blindly following its rules over the last 12 months has seen me currently fully pyramided into coal, iron ore, steel, fertilizers and oil and well out of financials which I was heavily into 12 months ago.

And blow me down, what sectors are over and underperforming at the moment...?

Nick Radge said:

...its been awhile since I made $22 mil in a year!

Click to expand...

Yeah, these dry spells are a bitch.

Re Random entries:

The 50/50 success rate we all experience would certainly indicate a random entry would suffice, but I have this nagging doubt.

Let's say we take a give set of exit conditions, shouldn't matter what they are so long as consistent. If we test differing entry conditions with our set exit conditions, we shouldn't see much difference in the bottom line.

Yet, when I was doing a lot of testing, different entries made a lot of difference.

Can we explain this by being "fooled by randomness", or is there a difference in results from different entries.

If so, that puts a dent in the random entry theory.

I have splinters in my @rse on this issue.

wayneL said:

Re Random entries:

The 50/50 success rate we all experience would certainly indicate a random entry would suffice, but I have this nagging doubt.

Let's say we take a give set of exit conditions, shouldn't matter what they are so long as consistent. If we test differing entry conditions with our set exit conditions, we shouldn't see much difference in the bottom line.

Yet, when I was doing a lot of testing, different entries made a lot of difference.

Can we explain this by being "fooled by randomness", or is there a difference in results from different entries.

If so, that puts a dent in the random entry theory.

I have splinters in my @rse on this issue.

Click to expand...

Read this thread if any of you haven't seen this thread on TradingBlox Forum yet. It's quite interesting on how these ppls were trying to replicate the random theory. (to no success of course)

http://www.tradingblox.com/forum/viewtopic.php?t=3637

This is another interesting one.

http://www.tradingblox.com/forum/viewtopic.php?t=3779

Apparently (at least from his results so far), a specific entry with a random exit gives 42 profit runs out of 50. Something to ponder...

Temjin said:

Read this thread if any of you haven't seen this thread on TradingBlox Forum yet. It's quite interesting on how these ppls were trying to replicate the random theory. (to no success of course)

Click to expand...

The testing demonstrated does not prove what it purports to prove.

Problems;
1. The universe is wrong. Random entry + ATR stop only works on equities since they have a long term upwards bias to exploit. No other market exhibits this upwards bias. A crude example illustrating this is buy and hold - do that in equities and you'll end up ahead eventually. Do that in Forex or Futures and you won't.

2. A 3.0 ATR stop on daily equities data is too tight and will whipsaw to death. 3.0 ATR will work on weekly data, but not daily.

For those interested in Dan Zanger, here is the article in which he was first featured (Fortune magazine, Dec. 2000).

wayneL said:

Let's say we take a give set of exit conditions, shouldn't matter what they are so long as consistent. If we test differing entry conditions with our set exit conditions, we shouldn't see much difference in the bottom line.

Yet, when I was doing a lot of testing, different entries made a lot of difference.

Can we explain this by being "fooled by randomness", or is there a difference in results from different entries.

Click to expand...

I too have done some testing comparing a range of different entrys to a common exit and have arrived at the same conclusions, they do fluctuate a lot. I am not talking about singular runs of each entry/exit combination, but monte carlo runs as well as using GP's random skipping of trades and the averages (incorporating a large number of runs) fluctuate a fair bit. This tells me that certain entry/exit combinations do tend to work better than others.

In fact, I go as far as to say that it would almost be miraculous if anybody was able to show results from a large number of entries, being matched to a common exit showing results that showed little variance.

It certain does put a dent on the random entry concept and is the reason why most traders do not use it IMO.

With regards to the edge ratio, my belief is that this has good potential in that it allows for the evaluation of the effectiveness of an entry, independent of any exit. That is not to take anything away from the importance of trade management and exits, but it does offers an opportunity to develop a superior entry/exit combination that will hopefully outperform a system incorporating a random entry.

nizar said:

Is the word "edge" only related to entry?

Click to expand...

Sorry Nizar, I missed this.

I think talking about 'edge' is contextual. We can zoom in on entry and try to quantify 'edge' there, or we can zoom out to 37,000 feet and try to quantify 'edge' overall.

As an example, an 'edge' in gambling is of course the 0 or 00 on a roulette wheel. That's the house's edge. Your edge is something that allows you to beat the house in spite of their edge (2,63% in Europe and 5,26% in the US).

Your edge in trading is something that allows you to win inspite of the fact that you are instantly in the red by an amount equal to round-trip brokerage (and perhaps interest) every time you enter a trade.

Of course if everyone is only trading long, and the market was going up anyway, then everyone gets an edge over the house. In this situation you might be more interested to know if you had an edge above and beyond that intrinsic edge. So you could compare your results with an index.

I think people would be surprised to find that there is a lot of edge in simple money and trade mangement ideas like non-market-cap weighted capital allocation. That is, investing in 10-20 shares with equal position sizing instead of 500 shares position sized according to market capitalisation.

ASX.G

wayneL said:

Yet, when I was doing a lot of testing, different entries made a lot of difference.

Can we explain this by being "fooled by randomness", or is there a difference in results from different entries.

If so, that puts a dent in the random entry theory.

I have splinters in my @rse on this issue.

Click to expand...

I haven't really thought through the idea of random entry/exit outside of a long-only LTTF system within the context of a bull-market. I did the testing because I had an overwhelming suspician that the success of any LTTF during such a period had more to do with market exposure than any well thought out entry or exit method.

Consider the attached mock up of a bull market. A good entry will present most of it's trading opportunities during the long bull phases and fewer, if any, during the short and sharp down moves.

With enough runs (which turns out not to be that many) it is possible for a random entry/exit system to out-perform the designed LTTF. This is providing that the parameters for the random system are such that it results in about the same amount of market exposure as the LTTF.

On the other hand, if the market looks more like the second mock-up, then this kind of testing will result in fewer situations where the random entry/exit outperforms the LTTF system. Here the down moves are longer, and to match the market exposure of LTTF the random entry-exit system will be forced to take trades that have a higher probability of occuring during these unfavourable conditions.

I think when you do a random entry/exit test you need to think what it is you're trying to demonstrate, and if necessary, introduce some constants eg. prevent the random entry/exit system from taking trades if price is not above the x-day EMA (match x to whatever your actual system uses).

If you code up an entry you want to test and you randomise a function within your system to keep forcing it to take different paths thru the data, and then compare the distribution of 10,000 iterations of that, against 10,000 iterations of a random entry/exit system you could say you are measuring the edge of your entry....where 'edge' might mean higher CAGR, lower max.DD, higher win% etc.

This assumes that you have enough data and your random function is able to force the system with your actual entry to take a sufficient number of diverse paths through the data.

Personally I think that this is not a very time-effective way to design a system There are too many permutations. And it seems that when you focus on one area too much you end up fine-tuning that and when you re-introduce the other parts it turns out they don't work as well together any more. A top-down approach with a bit of 'common sense' thinking about what should work and why is probably a more time-effective way of doing it.

In any case, I was able to prove to my satisfacation that position sizing, stop losses, leverage and market conditions, altogether, were responsible for the lion share of LTTF success on the ASX during the past 5 years.

RE: Random entries

If you look at breakout patterns, many are characterised by range contraction.

Volatility/range contraction has a significant impact on the risk:reward relationship of an entry signal - it shifts the expectancy of the trade in your favour because you are able to run a tight stop, relative to past price action.

So by trading breakout patterns you are only participating in trades with favourable risk-to-reward. Not to mention it often aligns with medium-term inflection points...

Compare this with a random entry: either 1. the stop is relatively 'tighter' or 2. the reward is relatively less.

If the stop is tighter, the win% will suffer, which will negatively impact expectancy. Alternatively, If the stop is adjusted to the most recent price action, the reward will be comparatively less, beause the potential upside of a random entry versus a pattern entry is exactly the same.

tech/a said:

To T/A.
Having discarded most T/A other than VSA and Patterns
I have come to the conclusion that these tools of analysis do one thing.

They supply me with Trade Start/Failure/and End points of reference for my trade management.
NOT ONLY for the three distinct points mentioned but also for continuation of trade management over the duration of a trade wether that be pulling back or belting the hell out of a trade.
With this I can and do skew the results in my favor.

The answer is in the simplicity for all who want to see it.

Click to expand...

Great post John,

The irony is that most will either never get it ... or spend many years of frustration before returning to the same conclusion.

sleepy