Reasons for failur in automatic trading systems

I'm surprised to see that in the Automated Trading forum most of the people think that automatic machines are very rare in the trading fileld and very hard to develop, if at all.

A little about my method and why I think it will do good in forward testing:
- Intraday only.
- Based on the support & ressistance break throughs rational.
- Taking in account:
- 1 minute ticks.
- Mainly one minute low-high spread + volume.
- Daily Low & High.
- Previous days support & resistance.
- Previous days volumes.
The above info (with some more) is mixed up to produce the potential of the index to go long / short at any given point.
This is a very basic and old concept, which is not complicated to implement.

The thing is, I've developed it using information for the last 3 months only, and then backtested it on the past 2-4 years.
It did great - 10-50 pips a contract per month on 6 main indices without any big drowdowns.

Allthough I understand why, It's not rational that it will not give good results in the near future (1-2 years ahead).

Comments....
 
im toyin' with da idea of automatin' me openin' ranges trades, spreadin' me bets on gappers that meet me requirement..i got to da conclusion that it wud be impossible for da system to be fully automated...most difficult thing wud be placin' progit tgts, da most important paramenter of 'em all for this kinda strategy. for da moment i cannot figure a functional bot that cud be able to fully replicate me discertionary approach; as maxpi said, way to many imputs impossible to translate into a language my pc cud understand...on da other hand it cud be very much feasable to create somethin' that requires assistance but that wud not make much sense given that i still wud have to monitor entry/exits'n'override da bot's decisions.
 
Quote from cohvi:

I'm surprised to see that in the Automated Trading forum most of the people think that automatic machines are very rare in the trading fileld and very hard to develop, if at all.

A little about my method and why I think it will do good in forward testing:
- Intraday only.
- Based on the support & ressistance break throughs rational.
- Taking in account:
- 1 minute ticks.
- Mainly one minute low-high spread + volume.
- Daily Low & High.
- Previous days support & resistance.
- Previous days volumes.
The above info (with some more) is mixed up to produce the potential of the index to go long / short at any given point.
This is a very basic and old concept, which is not complicated to implement.

The thing is, I've developed it using information for the last 3 months only, and then backtested it on the past 2-4 years.
It did great - 10-50 pips a contract per month on 6 main indices without any big drowdowns.

Allthough I understand why, It's not rational that it will not give good results in the near future (1-2 years ahead).

Comments....
More...


Try to backtest it on many other indices. Raise the commish and slips (even though for such a low t/f it will probably make a diff). If it holds up, you may be onto something. Also, may I ask you how you implemented it?
 
Your going to get killed.

nitro
Quote from cohvi:

I'm surprised to see that in the Automated Trading forum most of the people think that automatic machines are very rare in the trading fileld and very hard to develop, if at all.

A little about my method and why I think it will do good in forward testing:
- Intraday only.
- Based on the support & ressistance break throughs rational.
- Taking in account:
- 1 minute ticks.
- Mainly one minute low-high spread + volume.
- Daily Low & High.
- Previous days support & resistance.
- Previous days volumes.
The above info (with some more) is mixed up to produce the potential of the index to go long / short at any given point.
This is a very basic and old concept, which is not complicated to implement.

The thing is, I've developed it using information for the last 3 months only, and then backtested it on the past 2-4 years.
It did great - 10-50 pips a contract per month on 6 main indices without any big drowdowns.

Allthough I understand why, It's not rational that it will not give good results in the near future (1-2 years ahead).

Comments....
More...
 
Quote from robertstone:


If your model is failing, how soon will you me made aware that it is a model failure, vs. say expected drawdowns, and once failure has been identified, how soon can it be fixed.
--RS
More...

Thx for saying that.... It's always interesting to ask discretionary traders about that....
 
Quote from Cheese:

The obvious question on backtesting appears never to occur to backtesters.
Are the raw numbers of Open, Close, High & Low a sufficient data input per day for sequential days?
Thats the question.
The answer is 'No'.
:)
More...

The answer is : Yes.

But having more signals is quite helpful too and helps to diversify.
 
Quote from cohvi:

I'm surprised to see that in the Automated Trading forum most of the people think that automatic machines are very rare in the trading fileld and very hard to develop, if at all.

A little about my method and why I think it will do good in forward testing:
- Intraday only.
- Based on the support & ressistance break throughs rational.
- Taking in account:
- 1 minute ticks.
- Mainly one minute low-high spread + volume.
- Daily Low & High.
- Previous days support & resistance.
- Previous days volumes.
The above info (with some more) is mixed up to produce the potential of the index to go long / short at any given point.
This is a very basic and old concept, which is not complicated to implement.

The thing is, I've developed it using information for the last 3 months only, and then backtested it on the past 2-4 years.
It did great - 10-50 pips a contract per month on 6 main indices without any big drowdowns.

Allthough I understand why, It's not rational that it will not give good results in the near future (1-2 years ahead).

Comments....
More...

Cohvi,

I fully sympathise with your opinion here, I think it is nonsense for people to say future market performance has no relation whatsoever to past performance. If that were true then the markets are purely random and you would be just as well off going to a casino. The trap most people fall into is they over optimise and curve fit their systems on too small a range of data, what you are looking for is a system that doesn't produce brilliant results on past data - but produces reasonable results on both optimised data and data the system has never seen before.

I've been running a daily auto-system successfully (although I manually enter the trades it recommends) for several months now - however it does require tweaking at regular intervals to accomodate changing markets, but the tweaks are usually minor. But I've also automated the tweaking process so I now regard it as an integral part of my auto-system.

I'm currently at the final development stages of a fully auto FX trading system which trades on the 1min data. The actual trading algorithm has been finished for some time and I've been tweaking the parameters using as much data I can get my hands on. I now have a system that can deliver reasonable results on any historical data I throw at it. Now I plan to leave it alone and see how it performs on the data over the next few weeks. If all goes well, I'll go live and do monthly tweak to keep it running smoothly.

I do not expect it will make profit week in week out, however over time the general direction will be up.

Keep the faith!
 
Quote from cohvi:

I'm surprised to see that in the Automated Trading forum most of the people think that automatic machines are very rare in the trading fileld and very hard to develop, if at all.

A little about my method and why I think it will do good in forward testing:
- Intraday only.
- Based on the support & ressistance break throughs rational.
- Taking in account:
- 1 minute ticks.
- Mainly one minute low-high spread + volume.
- Daily Low & High.
- Previous days support & resistance.
- Previous days volumes.
The above info (with some more) is mixed up to produce the potential of the index to go long / short at any given point.
This is a very basic and old concept, which is not complicated to implement.

The thing is, I've developed it using information for the last 3 months only, and then backtested it on the past 2-4 years.
It did great - 10-50 pips a contract per month on 6 main indices without any big drowdowns.

Allthough I understand why, It's not rational that it will not give good results in the near future (1-2 years ahead).

Comments....
More...

Ask yourself:

1. Does my system have an edge?

2. To answer yourself of the edge... What makes the system profitable? What "condition" or "character" of the market has allowed the system to sustain it's profitability?

3. Does the list of things above really matter to the 2 prior questions?

...

Maybe I've wrote too much... LOL
 
You must log in or register to reply here.
Back
Top