Horses for Courses

[[Deleted]]

Hi All

My first post. I would like to evaluate as much logic, from as many EAs as I can and put each piece of logic in a box. I'm not happy with the trade-execution logic I have seen in any EA I have looked at, and I have looked at hundreds. Most are very good interpretations of strategy and do their best as far as entry and exits, but ... taking a purist viewpoint -> they simply toss on trades and axe them fairly randomly after a close condition ... when viewed from a smaller timescale than most EAs use.

I am purposefully being harsh - my goal is 100% success and tolerances are vital. Anyone who has ever watched the tick chart on the gbp/jpy knows just how much the market can (and does) move in 15 minutes, yet we are quite content to toss on a trade at the end of a 15 minute bar without even the slightest attention to the smaller charts. This is a criticism of myself as well, please don't anyone take this personally. I have traded gbp/jpy off the 15 minute (because the moves are bigger on the 15m chart and likelihood of going wrong is less than the 1m chart), placing my trade at the optimum point and watching it dive 50 to 100 pips before continuing it's climb to sucess.

EAs that trade on accelaration are very often wrong-footed in a volatile market and badly positioned as the market tests channel support and resistance before the real breakout (or whatever the case may be). This is entirely my point - horses for courses, the appropriate strategy for the right conditions is needed.

What I would like to do is the following:

1. Determine varying conditions for various strategies and list the strategies in order of success per condition

2. Develop optimum entry & exit techniques and apply to varying market conditions

3. Devise the logic to match conditions to strategies

4. Analyse winners

5. Analyse losers

Here's how I hope to do it:

1. Build a data warehouse, extracting every conceivable piece of information neccessary for complete analysis (at every useful timeframe)

2. Build a knowledge base by analysing winning & losing trades (and their associated strategies)

3. Build a rules engine to run off the knowledge base and logic in the EA to use the rules engine, identifying the current conditions and the best suited strategy

4. Apply the optimum entry and exit strategies for the optimum trade

I've identified GRNN neural nets for the purpose of classification. The first order of business, however, is to identify all the optimum buy and sell points over an extended period and then train the system to identify which indicator readings, crossovers, divergences and the like are matched to all other occurrences. There is some good analytical reporting software that can do this out the box. Once that's done, then the system can be put through it's paces and the trade analysis begun.

... long post, sorry. Point being, there are a lot of really good EAs, all successful under the right conditions. All we have to do (other than optimise exit and entry points) is create a library of all that logic and work through the history as explained above, attempting to create the ultimate management tool for the EAs that already exist. Of course, there may be opportunity to invent the holy grail, but the first order of business would be to get the EAs that already exist, working to their optimum.

Does anyone like this idea?

Regards

Emile

[[Deleted]]

... hmm, ok - everybody seems happy with the logic "if (macdMain>macdSignal) { buy() }"

I'm not. I'll battle on on my own and post my result - if anyone has ideas on which indicators to watch and which timeframes, please post a reply with those. I'm looking for a way to specifically analyse losers, from a variety of different perspectives, so that early warning signals can be generated allowing statistically losing trades to be avoided.

Thanks