Behemoth - All pairs Обсуждение

This is the new Behemoth test on FX Open.

Nice opening trade on USD/CHF.

Risk is set at 1% per trade.

Pairs are

Aud/Usd
Aud/Jpy
Eur/Usd
Eur/Jpy
Gbp/Usd
Usd/Chf
Usd/Cad
Usd/Jpy

Got caught in the non-farm payroll release today. Could have gone either way but it didn't go too bad. I think I need to add a news filter to prevent new trades being created during the news releases.

Just discovered another bug in my code as well. Testing new version at the moment.

Account got massacred on Friday due to the appalling spread widening with USD/CHF at one stage moving up to as much as 12 pips!

I've now added in a spread filter to the account so that trades are not placed if the spread is too high, this will also help with avoiding trading on news spikes.

I could start another demo account but I think it's going to be interesting to see it recover from this position.

Hi Steve,

I see you've switched to a custom start date to hide the bad start...how come?

If you're only risking 1% per trade and trading 8 pairs, how did NFP cause a 23% drop?

Best of luck.

Hi CasTrader,
I'm hiding the bad start so I can more easily assess where it's up to from today. Otherwise I can't assess the performance of each pair with the new version.

The bad start wasn't caused by NFP, although it did contribute three losing trades on the Eur/Jpy spike but was mainly caused by the spread widening as the markets moved towards closing.

Some pairs use as little as a 15 pip stop loss and often get into a new trade after the stop loss is hit. As the spread on USD/CHF widened to 12 pips this gave the EA a 3 pip range to play around in! The account lost about $7000 in the course of several minutes towards the end of Friday's session as a result of this. I hadn't appreciated that the spread goes so mental on the close of the markets as I don't usually trade around this time. Now I know!

As well as adding in the spread filter, I'm planning to add a cutoff for Friday afternoon from a certain time and also a gradually shrinking trailing stop in order to close off any positions that are still open as there is a chance with this EA of trades being open for some time.

It should recover over the course of this week as markets start to move. The EA will be undergoing constant improvement in order to address the differences that occur between live trading and backtesting. Once I've addressed all these issues and have the working version I'll run a new test with version 1

Yesterday's sluggish markets due to the US Bank Holiday didn't do any favours.
Mental note to self - Don't start tests on NFP Friday, especially not when a bank holiday follows!

Seems to be recovering nicely now fortunately.

Steve,

You are funny. Thanks for sharing all your honesty as it is greatly appreciated. :) Most other people would 'oh, I had a technical issue w/ my broker (scumbags) and I decided that I will start again in 2 days etc.'.

Gil

Am I seeing things, Steve? I swear I saw this up at 5% growth and over 250 pips yesterday. Did it have a rough NY session?

Still trying to work out if this is working as the backtests project it will or not Cas.

The EA is designed to get into a trade if it thinks things are going in the right direction but using a tight stop loss, generally between 10-20 pips depending on the currency pair. If it hits this stop loss and the conditions are still favourable it will get into another trade, at the higher position. So it can suffer several losses in a row on the same pair if it gets choppy.

If it gets it correct it can earn as much as 300 pips - I'm sure you can work out what percentage that would add to the account on a 10 pip stop loss!

From backtest it should get it correct somewhere between 40-50% of the time. Which is about where it is. It just hasn't caught many big trades yet.

It doesn't win every day in backtest. Neither does it even win every month on each pair. Compound the results of eight pairs and it's possible that ranging conditions on the markets in general will cause it some upset. I'm comfortable that it's treading water for now.

As trades progress it will give me more information to work with to improve this alpha version. If I can find a good way to stop it getting in during some of the ranging conditions that it's been in ( UsdChf has been particularly harsh ) then it will perform better. However from testing so far, trying to avoid the ranging markets just leads to it missing the trend when it starts!

I'm also going to be setting up an account with more conservative settings that trades only the best four pairs and has a less aggressive approach to taking trades in the next few days.

Thanks for the comments Gil. I decided I would put this in the public domain as it might be interesting for people to see the process that is gone through on the development of an EA. So many of these things get done behind closed doors and then people expect everyone to have trust in their 'little black box'.