Please find latest aggregated positions attached.
So far all the data releases this week have pushed price in a favourable direction.
Currently the system has 156 open trades and floating P&L is in positive territory.
Going into the UK and EUR Interest rate decisions a breakdown of the aggregated position by currency can be seen below (in lots):
EUR 2.20
CAD -2.30
GBP 6.70
AUD -2.90
JPY 0.00
USD -6.00
CHF 2.30
See the attachment for a more detailed breakdown.
So far all the data releases this week have pushed price in a favourable direction.
Currently the system has 156 open trades and floating P&L is in positive territory.
Going into the UK and EUR Interest rate decisions a breakdown of the aggregated position by currency can be seen below (in lots):
EUR 2.20
CAD -2.30
GBP 6.70
AUD -2.90
JPY 0.00
USD -6.00
CHF 2.30
See the attachment for a more detailed breakdown.
The Nets had a stellar day yestrdy, meaning the system is up 20%.
It was perfectly positioned to take advantage of the USD move down and GBP & EUR moves up.
This translates to current equity being $144k (up $24k) and is way ahead of weekly target of $9k a week. However we have massive event risk of NFP today which could reverse some (or perhaps even all) of these gains. Also I've seen the system do this before and over the next 2 weeks we may see a pull back which after 3 weeks may still indeed average out at $9k profit a week.
Please see attached the detailed breakdown of the system's current open positions (there are approx 190 open trades).
It was perfectly positioned to take advantage of the USD move down and GBP & EUR moves up.
This translates to current equity being $144k (up $24k) and is way ahead of weekly target of $9k a week. However we have massive event risk of NFP today which could reverse some (or perhaps even all) of these gains. Also I've seen the system do this before and over the next 2 weeks we may see a pull back which after 3 weeks may still indeed average out at $9k profit a week.
Please see attached the detailed breakdown of the system's current open positions (there are approx 190 open trades).
Pat, tbh I'm just showing off the system to keep that buzz whilst I open a larger account with a partner. I'm not trying to sell the system, I'm literally trying to show the many doubters out there that there are neural network systems out there that actually work... its just that they'll prob never hear about them or have any insight into its operation because those people aren't sharing that information.
Its a fully automated system. The beauty in it is that you don't need to start manually fiddling with the trades. This has been my downfall in the past, you cannot second guess the neural nets.
I already know the optimum hold time for each currency pair / strategy combination. I know how much each strategy should make over the year if I hold it for x amount of days.
In the above posts you can see how I was positioned going into ECB and BoE interest rate decisions, and also how I was positioned going into NFP. I can almost guarantee you that whatever you would have manually changed going into those events you would have ultimately lost money compared to just leaving it to the system. And even if on the one occasion you may have made all the right changes and made more money, over the course of the next 11 NFPs I can almost guarantee that you would end up losing more money by playing around, than you would have had you just trusted in the maths and stuck with the optimum hold times.
I'll give you a real life example, going into last NFP on my real money account, I had amassed a huge total trade volume and I thought that going into that event the "prudent" thing to do would be to de-risk and halve, close out some of my positions, or even disable some strategies to stop additional trades going on. I didn't fiddle around too much but I closed off a few trades and disabled a few strategies in the run up to NFP.
After the announcement it appeared to go in my direction and over the next 2 weeks I made about £1500. However I went back and calculated how much the trades that I stopped going on would have made (long GBP/JPY, long EUR/AUD) and also how much the trades I cut short would have went on to additionally make (short AUD/USD)... to my horror I found out that I lost out on approx £5400. A massive move, way above my average take over 2 weeks.
I then knew that the next week I'd prob experience a pullback after such a massive gain, but unfortunately instead of having £6900 in the account to absorb the impending losses I had only £1500 and when the pullback came and I lost £2500 over the next 2 weeks I was ultimately left down £-1000 when I should have been up £4400.
So my point is the systems strength is in trading every single signal, and leaving it for the optimum hold time, unless there is another statistically proven reason for amending any trades and in that case that behaviour has to be consistent and can be coded to be automatic anyway.
The full system derives its strength from putting on trades on an hourly basis, hence it needs to be traded with micro-lots and has to be managed automatically because by the end of next week it will have on approx 450 trades.
However I actually have a stripped-down simplified version of the system that's been running on Collective2 for the last 7 weeks that would be easier to manage manually. Its currently up about 17% despite some early teething problems I had trying to make the changes to get it to work on there:
Neural Net FX - a forex trading system
So after a massive week for announcements and data releases last week, the Nets have started this week as I expected, trying to hold onto the gains. With nothing big in the calender until the latter part of the week it will be interesting to see what new trades are generated. I'm still awaiting the first GBP/JPY trade and with major JPY announcements tonight and tomorrow this is something to look out for.
The system is approx $19k (15.8%) up after one weeks full trading.
We now have a more even mix of winning and losing positions and post NFP it's quite likely that these losing trade will now stay unprofitable and eventually close for a loss once their hold time is reached.
This is fine, there will inevitably be losers too. I expect the system to eventually settle on a 60-65% win rate.
Please see latest positions attached.
Looks like the Nets are building up a massive long GBP position and the Short USD/CHF from last week closed off for a nice profit as the strategy signalled a reversal at the perfect moment to now be long USD/CHF and in profit.
The system is approx $19k (15.8%) up after one weeks full trading.
We now have a more even mix of winning and losing positions and post NFP it's quite likely that these losing trade will now stay unprofitable and eventually close for a loss once their hold time is reached.
This is fine, there will inevitably be losers too. I expect the system to eventually settle on a 60-65% win rate.
Please see latest positions attached.
Looks like the Nets are building up a massive long GBP position and the Short USD/CHF from last week closed off for a nice profit as the strategy signalled a reversal at the perfect moment to now be long USD/CHF and in profit.
The Nets continue to outperform even my expectations and are currently up +$22k (+18%).
Considering the system has built up a very large long GBP position, the NIESR GDP estimate this afternoon poses significant event risk (or opportunity) so it will be interesting to see how it affects P/L.
On a side note the system should be performing slightly better than even listed but this demo account doesn't have GBP/AUD which the Nets have been giving long signals for for quite some time now.
Please see aggregated positions by currency below and detailed breakdown attached.
EUR -3.70
CAD -2.30
GBP 17.20
AUD -2.30
JPY 0.00
USD -4.00
CHF -4.90
The are currently approx 290 open trades.
We're 3 days away from point of equilibrium, where trades will be getting opened at around the same rate as they're being closed.
Considering the system has built up a very large long GBP position, the NIESR GDP estimate this afternoon poses significant event risk (or opportunity) so it will be interesting to see how it affects P/L.
On a side note the system should be performing slightly better than even listed but this demo account doesn't have GBP/AUD which the Nets have been giving long signals for for quite some time now.
Please see aggregated positions by currency below and detailed breakdown attached.
EUR -3.70
CAD -2.30
GBP 17.20
AUD -2.30
JPY 0.00
USD -4.00
CHF -4.90
The are currently approx 290 open trades.
We're 3 days away from point of equilibrium, where trades will be getting opened at around the same rate as they're being closed.
So yet another impressive day for the nets. Looks like the system is trying to push on rather than consolidate as the current account equity stands at $146k, up $26k (21.6%).
The Nets have been positioned nicely to benefit from good GBP data over the last 24 hrs, however the event calendar is pretty light for the rest of the day.
The system currently has over 350 open trades.
Please find attached the aggregated positions and also the account equity on an hourly basis since the demonstration started.
So far max drawdown is about $11k which is just under 10%. This is inline with my expectations.
The Nets have been positioned nicely to benefit from good GBP data over the last 24 hrs, however the event calendar is pretty light for the rest of the day.
The system currently has over 350 open trades.
Please find attached the aggregated positions and also the account equity on an hourly basis since the demonstration started.
So far max drawdown is about $11k which is just under 10%. This is inline with my expectations.
Nets continue to perform well, although account equity has dipped slightly to $139k, up 19k (15.8%).
This morning we've seen the first GBP/JPY trades with the nets going short.
Also started seeing some trades closing off now that their optimum hold times have been reached. Currently the system has over 400 open trades.
Please find aggregated positions and equity chart attached.
This morning we've seen the first GBP/JPY trades with the nets going short.
Also started seeing some trades closing off now that their optimum hold times have been reached. Currently the system has over 400 open trades.
Please find aggregated positions and equity chart attached.
Vasco da Banana
Established member
- Messages
- 699
- Likes
- 121
Interesting thread, Intel_Bvc. Keep up the good work.
c6ackp
Well-known member
- Messages
- 441
- Likes
- 28
HiIntel
Why does the algo have to trade every hour? I think folks are, in general, a bit sceptical of systems which have hundreds of open trades in a few pairs.
What is the performance like if you only take the first new directional signal in each pair and hold that to your optimum hold time settings and then repeat?
Thanks,
c6
Why does the algo have to trade every hour? I think folks are, in general, a bit sceptical of systems which have hundreds of open trades in a few pairs.
What is the performance like if you only take the first new directional signal in each pair and hold that to your optimum hold time settings and then repeat?
Thanks,
c6
Vasco da Banana
Established member
- Messages
- 699
- Likes
- 121
But isn't the essence of the idea to trade a large number of small positions? I doubt simplifying it would work well, if at all, but no doubt Intel has looking into that aspect.
c6ackp,
Thanks... I've been waiting to start getting some good questions!
One of the criticism's people came up with when I started this thread (before they even saw the system perform) was that Neural Nets are slow to respond. I've also seen other threads on this forum with people complaining and using all sorts of mathematical jargon to explain why they're no good.
But the truth is no-one's doing what I'm doing.
I'm doing something a little bit unconventional and different and therefore my system is performing in a way that no-one else has managed to demonstrate, and I've done it live for anyone to log in and see at any time.
The simple answer to your question is that the nets a trained with new data (on sliding window principle) and predictions are then generated after training, on the hour, every hour.
Therefore when there's a shift and the existing "pattern" changes then the nets can respond very quickly. Another factor is the "density" of the signals. There are some "signals" that accumulate a large trade size and others that don't. There's obviously a reason for this although I don't fully understand it, all I know is that it produces results.
I do have a version of the system similar to what you are suggesting and its been running on collective2 for over 7 weeks Neural Net FX - a forex trading system. Its also making good profit but it doesn't seem to have the edge that the full system has because the full system can maintain this complex web of hedges that no-one truly understands.
If I kept this demo running for one year I expect it to make over $400k, have a win ratio of 65%, I expect my winners to be much larger than my losers and I expect the max draw-down to be no more than $12k (10% of initial capital). Those are phenomenal stats.
I understand people may be skeptical , and I understand that a lot of people may not have a clue exactly how it all works. But this is a revolutionary system. I'm showing you something that you'll not see again once I get bored and stop the demo.
So feel free to make the most of it and ask questions.
Thanks... I've been waiting to start getting some good questions!
One of the criticism's people came up with when I started this thread (before they even saw the system perform) was that Neural Nets are slow to respond. I've also seen other threads on this forum with people complaining and using all sorts of mathematical jargon to explain why they're no good.
But the truth is no-one's doing what I'm doing.
I'm doing something a little bit unconventional and different and therefore my system is performing in a way that no-one else has managed to demonstrate, and I've done it live for anyone to log in and see at any time.
The simple answer to your question is that the nets a trained with new data (on sliding window principle) and predictions are then generated after training, on the hour, every hour.
Therefore when there's a shift and the existing "pattern" changes then the nets can respond very quickly. Another factor is the "density" of the signals. There are some "signals" that accumulate a large trade size and others that don't. There's obviously a reason for this although I don't fully understand it, all I know is that it produces results.
I do have a version of the system similar to what you are suggesting and its been running on collective2 for over 7 weeks Neural Net FX - a forex trading system. Its also making good profit but it doesn't seem to have the edge that the full system has because the full system can maintain this complex web of hedges that no-one truly understands.
If I kept this demo running for one year I expect it to make over $400k, have a win ratio of 65%, I expect my winners to be much larger than my losers and I expect the max draw-down to be no more than $12k (10% of initial capital). Those are phenomenal stats.
I understand people may be skeptical , and I understand that a lot of people may not have a clue exactly how it all works. But this is a revolutionary system. I'm showing you something that you'll not see again once I get bored and stop the demo.
So feel free to make the most of it and ask questions.
counter_violent
Legendary member
- Messages
- 12,672
- Likes
- 3,787
Hi Intel,
A few questions if I may.
On the aggregated positions update how many lots total are open positions?
In the same post you say the system has over 400 open positions. 400 x 0.1 lots is 40 lots. Every new position is 0.1 lot?
What number of open positions does the system consider to be the maximum?
cheers
A few questions if I may.
On the aggregated positions update how many lots total are open positions?
In the same post you say the system has over 400 open positions. 400 x 0.1 lots is 40 lots. Every new position is 0.1 lot?
What number of open positions does the system consider to be the maximum?
cheers
The system has over 400 open trades. None of these are pending etc. The reason why I don't say exactly how much is because a handful of trades are 0.2 lots opposed to the usual 0.1. So I cannot say exactly from that aggregated positions table.
There is not a maximum as such... but I know how many trades on average each strategy / currency-pair combination typically generates and I know the optimum hold time. So I know roughly how many trades will be on.
Tomorrow, will be 10 days of this demo and that corresponds to my longest hold time, so I know that we'll be at a point of equilibrium where trades are opening roughly as fast as they're closing. This will usually be about 450 open trades. I've never seen much more than this.
There is not a maximum as such... but I know how many trades on average each strategy / currency-pair combination typically generates and I know the optimum hold time. So I know roughly how many trades will be on.
Tomorrow, will be 10 days of this demo and that corresponds to my longest hold time, so I know that we'll be at a point of equilibrium where trades are opening roughly as fast as they're closing. This will usually be about 450 open trades. I've never seen much more than this.
c6ackp
Well-known member
- Messages
- 441
- Likes
- 28
The model makes new predictions every hour, or it re-calibrates and then makes a new set of predictions?
If this works out and you are as successful as this guy...
QIM's Jaffray Woodriff: The monk in managed futures | Institutional Investor
...please come back and tell us the story.
Average hold time in the article is 7.5 days - sounds about right
If this works out and you are as successful as this guy...
QIM's Jaffray Woodriff: The monk in managed futures | Institutional Investor
...please come back and tell us the story.
Average hold time in the article is 7.5 days - sounds about right
It re-calibrates and then makes a new set of predictions.
Interesting article C6.
I'm obviously not even anywhere near his level, but the thing we do have in common is that we've approached this challenge from a different angle to others. In this instance it seems not following the crowd is reaping rewards. I've had no outside influence from anyone regarding neural networks... I had (and still kinda have) little idea what anyone else is doing in this area. Meaning my approach is 100% original.
There's along way to go, I'm enjoying this demo account because its not real money.
Trading the real money account is difficult, not because anythings different from the nets perspective, but because I still have difficulty leaving it to run and I mess up by manually tinkering with the trades through fear. Once I can get my account sizing and discipline correct who knows...
I'm obviously not even anywhere near his level, but the thing we do have in common is that we've approached this challenge from a different angle to others. In this instance it seems not following the crowd is reaping rewards. I've had no outside influence from anyone regarding neural networks... I had (and still kinda have) little idea what anyone else is doing in this area. Meaning my approach is 100% original.
There's along way to go, I'm enjoying this demo account because its not real money.
Trading the real money account is difficult, not because anythings different from the nets perspective, but because I still have difficulty leaving it to run and I mess up by manually tinkering with the trades through fear. Once I can get my account sizing and discipline correct who knows...
Similar threads
