TDesk Whispers

[rapple]

My holy grail will be a shortcut and at the same time go beyond - The multi-pair TDesk Optimizer. Fed with a signal history, it will simulate tons of variations of the relevant TDesk inputs (those that you can change by clicking buttons in TDesk Player). It will run for hours or days and at the end tell you the outcome of each setting for the given signal history on the multi-pair basket. Once we have all the simulation data, we can display it as graph or chart or whatever for each variation.

It will be a long way to get there, but doesn't that sound like the future?

That would be truly awesome ...I am no statistician but I think that Monte Carlo analysis can be used to sample variables and determine best setup

Hopefully we have a statistician on the team to help guide us :youknow:

Cheers Richard

[cozybooks]

Thomas, thanks for the invite!

I am mostly on the road this week but looking forward to digging into this much deeper.

With regard to the veto signals, I was thinking along Leon's line of thinking (I think! ). I'm actually using my veto on lowest timeframe. My example:

Chart sending four signals to tdesk:
Three Baluda SS timeframes (4-hour, 1-day, 1-week) plus the Gann indi on a 15 minute TF. I'm looking to buy at 75% solo buy signals, but only if the Gann indi is set to buy (in effect, it is my veto signal, even though it is the lowest timeframe). I'll buy every time the Gann Indi changes from sell to buy, as long as at least two out of the three ss signals are buy (I don't care which ones, at least in the testing I am doing now).

Then I close as soon as Gann changes back to sell.

On the other hand, now that I read your explanation, I'm beginning to look at everything from the opposite end. Hopefully it will expand my thinking!

I've been working at my regular job for the last 15 hours, so if this doesn't make sense, hopefully it's only from my brain being tired and not too many deceased brain cells.

Endless thanks for everything you are doing!

Steve

[Ehrenmat]

My holy grail will be a shortcut and at the same time go beyond - The multi-pair TDesk Optimizer. Fed with a signal history, it will simulate tons of variations of the relevant TDesk inputs (those that you can change by clicking buttons in TDesk Player). It will run for hours or days and at the end tell you the outcome of each setting for the given signal history on the multi-pair basket. Once we have all the simulation data, we can display it as graph or chart or whatever for each variation.

It will be a long way to get there, but doesn't that sound like the future?

Cheers
Thomas

Hey Thomas!

What an awesome dream: The TDesk Optimizer ! ! !

And at this point, I have an idea for a little moonshot... Okay, maybe only the stratosphere but listen up:

Step 1: fxtestify
--> Check
Step 2: TDesk
--> Check
Step 3: Enough promising test results of TDesk setups on fxtestify
--> needs more time and motivated fxtestify participants
Step 4: All the history data of the most promising test results
--> needs the motivated fxtestify participants to cooperate

Step 5: Channeling/Uploading all the history data of the most promising TDesk setups on fxtesifiy

Step 6: Compute the history data centrally with the upcoming TDesk Optimizer

Result: The the most optimized TDesk setups of all time






What do you people think?

Cheers,

Mat

:hi:

[cozybooks]

Hi all,

I am throwing this out there just in case it spurs ideas in others. I am looking to run a few different variations in Tdesk that involve trend trading and stacking trades. I remember a post by Bob a few years back in which he said he only started making good money when he started stacking trades, and it has stuck in my mind.

I'm very excited about the idea of gathering history from different strategies and what it might give us in the future, but it can take a lot of history to really have valuable information. My two weekly AUDUSD charts cover about eleven years of data. Imagine having Tdesk back in the early days and running trending trade tests. In 2008 and 2009 (bottom chart), you would have made a killing and thought you had the perfect system - 3000 pip runs on AUDUSD! Imagine spacing 10 trades 100 pips apart at the start of either of those runs. Ten trades each making an average of 2500 pips profit on just one pair that usually doesn't move much.
2014 would have brought another great run that lasted over a year. Compare that to if tdesk was running trend trades in 2016 and 2017. You might have just given up on a system that did nothing or lost money for two years straight.

I'm hoping to get a few tests set up for the start of next week. It will be interesting to compare having a TP on the trades vs only closing on a signal change. My first thought is that pure profit wise, it's better in the very long term to only close on a signal change, but psychologically ,it might easier to combine the two strategies.

Great work, Thomas, as always!

Steve

[Dragonian]

Hello Thomas,

In TDesk does the input "BlockBeforeNewsMinutes" only block signals for "x" minutes before the news, as the name suggests, or does is also block signals for "x" minutes after the news? If the former it would be useful to inlude a parameter for "x" minutes after the news.

What do you think?

[Bruster400]

Leon, you gave me something to think and I thank you for this. I will reflect on the Excluding/Including thing (especially as I have seen an error in my respective code right now). Most likely Veto Signals should always be excluded from the Solo and Multi signal calculation.

You other masterminds, what do you think about that part?

I think you're right, the way I see it is that the veto signal should be excluded from both the solo and the multi signal calcs. I feel that the veto signal should be used as the overall "Trend" of the system - either long only or short only. The solo and multi signals can then be faster (generally lower timeframe but not always) trigger signals for entries into that trend. Once the faster "trigger" inputs combine in such a way as to give a solo or multi signal, they should be checked against the overall "Trend" (veto signal) and if the solo or multi signal is against the trend then it is rejected/filtered out. A change in the veto signal would therefore be a change in the overall trend and should also be used to exit any open trades in the opposite direction.

Having the veto signal work in this way would allow us to test whether a higher timeframe trend filter actually helps improve a systems performance or not. I've used lots of overall "trend" filters in the past that have actually reduced the overall performance!

[cozybooks]

I remember in my early days asking Bob an HGI question and showing a chart that I thought showed EURJPY on a two week trend. I thought I was so smart! Bob replied with a chart showing a true EURJPY trend, in which it moved pretty in one direction for months and months. I think I didn't post anything for six months after that!

I think that when you get a true trend that goes on for months (or a week or two on a 15 minute chart would I suppose be the equivalent on a scalping timeframe), you are very, very happy to have the trend filter and it saves you from a lot of losses and lets your winnings pile up like crazy. When you are not in a long drawn out trend, the trend filter keeps you out of a lot of trades that could have made some money.

Where's the balance? The problem is, none of us really know when a very long trend is going to hit or how long it will last. I live in Massachusetts, where John Henry used all of the millions he made trading commodities to buy the Boston Red Sox years ago. He believed in always being in a trade, either buy or sell depending on some long term buy or sell signal he had developed. Some years he would lose 10, 20% or more when price ranged, but they would always be followed by years of massive gains once a trend really hit. Now there's patience for you! I also remember him saying that he started truly making money when he realized that nobody out there really knows where price is going next, so you can ignore all of the experts. He would just follow his trend for as long as his signal lasted, then close all of those trades and start trading the other way.

That's my new goal with tdesk and all of the testing. Different timeframes, different triggers. It will be great if we can start getting data from a lot of testing. The brilliance of what Thomas is doing is that you never know when all of those results could trigger a brilliant idea from someone just plodding along who sees what someone else has done and just tweaks it a little bit. Happy trading!

[c1borg]

Hi Richard.

That is exactly the path I am thinking along. The problem is: Clicking on one of the buttons would start a calculation process that could take quite some time. TDesk Player would have to go through all 1-minute data points, re-calculate entries and exits, track the trades etc. Phhh - this will get time-consuming.

My holy grail will be a shortcut and at the same time go beyond - The multi-pair TDesk Optimizer. Fed with a signal history, it will simulate tons of variations of the relevant TDesk inputs (those that you can change by clicking buttons in TDesk Player). It will run for hours or days and at the end tell you the outcome of each setting for the given signal history on the multi-pair basket. Once we have all the simulation data, we can display it as graph or chart or whatever for each variation.

It will be a long way to get there, but doesn't that sound like the future?

Cheers
Thomas

This would be amazing and save soooo much time optimising a strategy. Just a thought could we feed the data collected into https://rapidminer.com/ or would you like to keep everything in Empty4?

[bishopdotun]

A change in the veto signal would therefore be a change in the overall trend and should also be used to exit any open trades in the opposite direction.

Having the veto signal work in this way would allow us to test whether a higher timeframe trend filter actually helps improve a systems performance or not. I've used lots of overall "trend" filters in the past that have actually reduced the overall performance!

The trick would be to add veto signal to the exit operator list.

That way, Veto can be used as an exit condition by Trade partner EA.

[shanying]

Hi,firstly great thanks to Thomas for the inviting. and also thanks to steve and leon.you has really created a holy grail for the members of SHF.I run three demo accounts using Leon's setup from January 3,all accounts have a same deposit 1000$ and the fixlot 0.01,close on opposite signal only,but each account started at a different day. Those demos showing incredible results in the past month. one account which started from January 14 now grow up to1650$,one account which started from January 3 now grow up to 1325$,and another account which started from January 20 grow up to 1213$.Maybe I'm lucky at the time started,but the results have shown great success of TDSK and DESKY.I carefully observed the profit and loss situation of each day and I found some Valuable things that worth to share for our Tdesk whispers group.From the past tradings I found each account will keep a 3%-8% negative floating after the initial batch of orders opened. then those orders waitted for the trend to happen,and the initial floating losses will turn into positive floating profits in about a week,after that there is no negative floating appeared.The maxmimum positve floating profit of the early two accounts once rose to $900.so the Profit to risk ratio of the global trading could be 9 times.These amazing results motivate me to think about a monthly trading plan.If we open the trading system at the beginning of each month, stop trading when the monthly profit reaches 30% or 50%,etc. Waiting for the system to be turned on again next month, so that the cycle will continue to accumulate, what kind of effect will it be at last?
This is just an immature thought of trading plan for share. I will carry out long-term tests later. If there are good results, I will report here and on FXTestify in time.and thank Thomas and steve again, please forgive my stupidity.

Cheers
Shanying