System Messages

This forum is for members of the Apiary Investment Fund. To learn more about how to get involved, sign up now or fill out the form below and preview the training right now!

MS ML Callout

 
Back to Forums
 
 
mike_smith

MS ML Callout

https://github.com/gomlfx/apiaryfund

Legal Disclaimer: My Callout is for educational purposes only. It is not a recommendation to buy or sell. That is a personal decision that should be made by you only.

The ML callouts represent the evolution of my current trading skill. I don't claim they exist in a state of perfection or any given callout will be perfect.

 

Wed, 10/09/2019 - 1:17am
 
mike_smith

Exit buy orders and sell TP 1.155. Trending down. I'm changing strategy here from trading the uncertainty interval to trading just the gradient.

-fbprophet

 
FXGold54

Hey Mike,

Just wanted to give you a heads up that Monday I will be discussing the Type 1 trade set-ups that, I believe you really like,

Hope to see you in the room.

Allen

 
mike_smith

Thanks Allen. Ya I've been looking at the different set ups. I didn't know you had so many. But ya I scope the Type 1 on GBP.

 
mike_smith

For example: having luck on daily high and low prediction. Use it as you will. I propose I supply and members reply how they used it.

14.09.2020 00:00:00 GMT prediction:

High: 1.19071

Low: 1.18086

Supplementary predictions. Still measuring the probability for this. If you're unsure if it'll adversely affect your risk using below predictions then use above only.

Today is classified as high larger than low.

Close is predicted slightly lower than open at 1.18395. This number can be used as a break out point or to suggest down trend.

 
mike_smith

"As we came closer...small lights started flickering and flashing..."

2020-09-13 18_28_44-MS ML Log _ Apiary Fund.png
 
mike_smith

fbprophet has literally turned belly up! I changed my sell target on that to 1.18.

2020-09-14 23_04_42-fbprophet_H1_mt5.ipynb - Colaboratory.png
 
mike_smith

I will focus on random forest more since I was able to make stops easier with it vs fbprophet. I still haven't thought a way to make good stops with fbprophet.

 
mike_smith

I have a 9-5 job so sometimes I'll write the 0-24 gmt prediction late. Today writing it at 9 gmt. Just to keep record.

High: 1.19127

Low: 1.18187

bias to high.

I prefer to use 1 risk : 2 reward. And move SL to break even if approaches TP.

 
mike_smith

I'm going to reduce random forest to H1 interval.

 
mike_smith

1.175 ya that's deep. But I'm looking to buy 1.195 to 1.21. Now.

2020-09-16 21_04_47-fbprophet_H1_mt5.ipynb - Colaboratory.png
 
mike_smith

I want to start taking a look at random forest H1 predictions. I'll keep this comment as the log:

GMT | bias | prediction | profit
07 | sell | 1.17419 | F
08 | sell | 1.1720 | F

I'll try this as standardized. I think it can be used. Then high low differences used as TP/SL and no classifier prediction.

9 GMT:

2020-09-17 01_29_37-RandomForest_CLF_RGR_risk_reward.ipynb - Colaboratory.png
 
mike_smith

H1 analysis was wearing me out fast! I'm changing to D1 here was today's forecast:

2020-09-17 02_11_43-RandomForest_CLF_RGR_risk_reward.ipynb - Colaboratory.png
 
mike_smith

yesterday was a perfect example of the strategy; direction = open vs close, stop = close, target = either high/low.

 
mike_smith

Today's numbers for the record. I'll enter some other pairs here too.

Since open > predicted close TP 1.19.

2020-09-17 16_42_18-RandomForest_CLF_RGR_risk_reward.ipynb - Colaboratory.png
 
mike_smith

A nice change if you don't prefer fbprophet weekly forecast (left), a daily forecast for expected high/close/low (right). In this instance all three predictions are above the current price indicating buying pressure. Since eurusd is even under the low midline prediction, I would expect to target the high blue prediction midline (1.195). Subject to daily adjustment. I believe it's possible to use all three lines as targets along the way to the high. I'd like to note the daily vs weekly prediction congruence.
https://github.com/gomlfx/apiaryfund/blob/master/fbprophet_daily.ipynb

2020-09-18 23_43_53-fbprophet_H1_mt5.ipynb - Colaboratory.png 2020-09-18 23_45_11-Untitled1.ipynb - Colaboratory.png
 
mike_smith

On a 1 year graph for buying and selling oppurtunity, even before the pandemic fbprophet shows that buying USD opportunity was rare. Below are overbought/oversold graphs for eurusd.

My opinion is that the USD was being involuntarily overbought before the March 2020 correction. And March 2020 was clearly the breaking point where all that stopped and eurusd resumed buying EUR.

2020-09-19 08_09_49-fbprophet_daily.ipynb - Colaboratory.png
 
mike_smith

I changed code to run only high and low predictions. So attached says I should buy and start selling after 1.187.

2020-09-19 08_21_42-fbprophet_daily.ipynb - Colaboratory.png
 
mike_smith

Or more precisely, if I'm under 1.18671 I should buy and over 1.190271 I should sell (...with 100 pip grid) (;

2020-09-20 14_52_05-fbprophet_daily.ipynb - Colaboratory.png
 
mike_smith

As eurusd digs down, it still has to meet that high prediction line someday, which is currently predicted at 1.19.
Let's see what happens to this. BTW this should be compatible with Mr Pyne's swing wobbling.

2020-09-21 22_06_18-fbprophet_daily.ipynb - Colaboratory.png
 
mike_smith

As for the weekly forecast seems went flat after yesterday's dip. I'd like to note the trend component (bottom photo) seems to eerily change before the price action. I'll record more of it to see if it's significant.

2020-09-21 22_32_01-fbprophet_H1_mt5.ipynb - Colaboratory.png
 
mike_smith

Trend PCA change. I'll check it out today.

below snapshot: trend principal component analysis

2020-09-22 14_43_57-fbprophet_H1_mt5.ipynb - Colaboratory.png
 
mike_smith

Did PCA trend on High/Low predictions too. Average anticipated high now 1.185. If I was using swing wobble technique, I'd buy again at today's open to target 1.185.

2020-09-22 16_18_19-fbprophet_daily.ipynb - Colaboratory.png
 
mike_smith

I think the PCA trend component isn't anything I should follow. I'll keep looking at the current price under low gradient for sell and over high gradient for buy. Hopefully that translates to trading with the trend.

2020-09-25 13_54_18-fbprophet_H_high_low.ipynb - Colaboratory.png
 
mike_smith

For example the USDCHF. Look how current price in the uptrend avoids the low prediction line.

2020-09-25 14_15_55-fbprophet_H_high_low.ipynb - Colaboratory.png
 
mike_smith

The cool thing about trading is I only need inversion for other direction.

 
mike_smith

This is why I like the weekend. I can talk about cool s$$$ I have no idea works.

2020-09-25 15_16_38-fbprophet_high_low_M1.ipynb - Colaboratory.png
 
mike_smith

Looks like that was right. Here's the next 24 hours. Technically, I should do this at 0 GMT.

I use M5 one week of data and predict 1 day. This is 1440 train : 288 predict data points.

I am wondering if I could use this with risk management and fib I talked about. Something like 1sl:4tp, -.618:2.618.

2020-09-30 09_14_33-fbprophet_high_low_M1.ipynb - Colaboratory.png
 
mike_smith

It may be more useful for me to feed the model less data (like above), make a longer prediction, so to let it remain more dynamic. Then apply risk management to counter the variable risk. Just need to find a good balance for input vs output data and risk at that point. The key is mixing the variability of the prediction with risk management.

 
mike_smith

It's definitely changed its bias.

2020-10-01 07_57_46-fbprophet_high_low_M1.ipynb - Colaboratory.png
 
mike_smith

Seems still nothing's in the way of that eurusd drop yet.

2020-10-01 19_55_16-fbprophet_high_low_M1.ipynb - Colaboratory.png
 
mike_smith

Party could be over. Have a good weekend!

2020-10-01 23_26_46-fbprophet_high_low_M5.ipynb - Colaboratory.png
 
robunited

NFP's coming tonight. There will be some parties for some, for sure ;-)

For me also not, will do some analysis and try to stay out.

Have a good weekend!

 
mike_smith

Next week is still looking down. Forgot to throw that in. I think Friday is reflecting that.

 
mike_smith

Using tabular prediction is a lot less confusing. I could use TP>SL risk management in direction of the predicted numerics. The model is still highly nonlinear so I can update frequently. I found the more nonlinear, the more relevant. For example current prediction:

day: unknown, weekend
week: sell
month: buy
year: sell

This can be used in conjunction with the alignment trading (entirely different strategy) I just submitted in MS TA long term Log.
The below prediction numbers are never expected values...just values to give direction relative to the current open price.
Unlike the MS TA strategy where fib values are expected, to count risk on stops and targets.

2020-10-03 13_27_07-fbprophet_d_w_m_y.ipynb - Colaboratory.png
 
mike_smith

Ideally I'd like to have at least the day and week aligned. Looking for up now until today shows down.

2020-10-04 22_05_53-fbprophet_d_w_m_y.ipynb - Colaboratory.png
 
robunited

Good plan. Could you plot these lines on a daily chart and visualize them?

 
mike_smith

So you can fade them? Good luck. Sure but these are constantly changing.

 
robunited

;-) Come on, just put them on a chart to see it at glance.

It's difficult to go through a table of numbers ... make it visually easy for your subscribers ;-)

 
mike_smith

It's from this. I'm looking to get at least two in alignment. Right now nothing is in alignment.

2020-10-04 22_27_59-fbprophet_d_w_m_y.ipynb - Colaboratory.png 2020-10-04 22_14_06-fbprophet_d_w_m_y.ipynb - Colaboratory.png
 
robunited

The price levels from your table ... not a wave ...

Screen Shot 2020-10-05 at 12.14.59 PM.png
 
robunited

What's in alignment to each other? I can't even see a TF for both screenshots?

Screen Shot 2020-10-05 at 12.19.39 PM.png
 
mike_smith

Those waves are actually the direction to expect if nothing changes. Time ranges are on the bottom of the chart. You mean like the attached chart?

2020-10-04 22_20_10-2100683976_ JAFX-Demo3 - Demo Account - [EURUSD,Daily].png
 
mike_smith

Rob,

The dates are on the bottom of each chart. The model's very sensitive so I follow the day and week prediction mostly. The day chart can turn any time.

 
robunited

"The dates are on the bottom of each chart. "

On a span of 900 to 9000 pips? 1.19 to 1.10?

What exactly are the levels you suggest to trade?

mikes graph.png
 
mike_smith

You have to be used to reading the fbprophet. Forecasting is not a literal translation. I see you're trying to read the fbprophet literally. Think of it more as directional expectation of a time range. At the moment the day is expected up, the week down, the month up, the year flat. The model is very sensitive. When the day prediction aligns with the weekly prediction (could happen any time) I'll apply risk management.

 
robunited

"Those waves are actually the direction to expect if nothing changes. Time ranges are on the bottom of the chart. You mean like the attached chart?"

Good chart ... if I would make a prediction to the down side, this is how I think it would look like, without a computer.

No one knows what the markets really drives the next couple of days (as always) or what is going to happen.

I'm still baffled how your code can predict an upside for today and a downside 2 days later?

Based on what exactly?

easy chart drawings.png
 
mike_smith

"I'm still baffled how your code can predict an upside for today and a downside 2 days later?"

Longer data range on history (train) showing a different predicted direction.

 
mike_smith

Rob,

For your layer diagram I definitely think that would be right. Predictions tend to converge to current price.

 
robunited

Following with interest ;-) I really want to know what that computer program does different than just drawing the obvious support/resistance lines & price cycles on the chart.

 
mike_smith

"When the day prediction aligns with the weekly prediction (could happen any time) I'll apply risk management."

Which now is a time!

2020-10-04 23_13_41-fbprophet_d_w_m_y.ipynb - Colaboratory.png 2020-10-04 23_14_13-fbprophet_d_w_m_y.ipynb - Colaboratory.png
 
robunited

My M30 & H4 alignment for today and going forward, if to the downside ... (if not, to the upside the same thing other way around)

M30 alignment.png H4 alignment1.png
 
mike_smith

I think that forecast needs less data to be more flexible. I'll use 1 train : 1 predict. |:D

On the bright side if I was trading asia session breakout in line with the fbprophet forecast, I never would have entered an order because fbprophet was expecting down and asia session sell breakout never happened, keeping me out of ordering.

btw having fun going through the 4,000 pages of CMT I,II,III. A lot of risk management discussion. Thanks for the initial stimulation you gave me on risk management Rob! I look forward to apply what I learned about it from the CMT material.

Recently as I've been exposed to more risk management I have mulled over how to profit that way using statistical data supplied from other things. It's like trading numerically not visually.

Attached my CMT workstation layout. Have you heard of the Iron Triangle of Risk Control? It uses the 2% per trade rule on beeline. I realized a problem I had before joining apiaryfund. I had no problem creating new strategies but I never used risk control in any of it! I had no idea about using risk control and management. I was not pro level. That's all changed. I always analyze risk first now.

CMT Level III page 114: "Beginners jump into trades without giving them much thought. Intermediate-level traders focus on market analysis. Professionals dedicate a massive proportion of their time to risk control—and take money away from beginners and amateurs."

Yup that was me before apiaryfund--never spent time analyzing risk control. Hopefully I flip that around now and take money away from beginners and amateurs.

2020-10-05 14_20_49-264. CMT Level III 2016 The Integration of Technical Analysis.pdf - Adobe Acroba.png
 
mike_smith

Here's a risk management example how to clean up the SSI trade, only trade on asia session breakout and use a proper SL:TP ratio determined by fib,adr,etc.

As you see yesterday the ssi said buy eurusd but there was no buy breakout on asia session.

2020-10-13 17_06_40-2100683976_ JAFX-Demo3 - Demo Account - EURUSD,H1.png 2020-10-13 17_08_48-Current Ratio _ FXSSI - Forex Sentiment Board.png
 
Rookie

Mike the SSI is a counter-trend trade that players off of the London reversal.

Steve takes most of the trades and does not single one out as a stand-alone. By only selecting one pair you have a 50/50 chance of success instead of all six and their weighting of the strengths as a group. Much different odds. Also when not viewed over months of trades the consistency is not considered.

 
mike_smith

Rookie you're not factoring the risk management. My post is how to improve the risk management. And timing, by using an asia session breakout. He enters at 0 gmt with 1:1 risk control. That's too raw. Risk management is why he's making money then losing the same amount. Timing is why he's participating in losing streaks when it was totally avoidable. Also about diversifying if SSI said up but in fact was down that day using more pairs only tends to magnify the loss. Since as you know it's really USDX SSI measured here and most if not all pairs will go with usdx especially at NY session.

So the question is should one use the SSI at all as a strategy. Then if yes, how can you optimize the risk control. I have 2 comments on that. First I don't think the SSI data being used (fxssi.com) is comprehensive enough to represent the actual SSI market. Is it even possible. Forex is an autonomous market, example: volume data is unknown, how can you ever know the actual SSI values. If the fxssi represents an insignificant number of traders then is it still useful. Second, like the Continuing/Initial Jobless Claims numbers, SSI is not a technical indicator. AND both had a different sentiment for USD strength today. Conclusion: In my opinion there is too much variation in opinion when using non-technical indicators. It is said if you can't code it to a computer don't trade it. Non-technical indicators epitomize this proverb.

 
Rookie

ok got your point,
"So the question is should one use the SSI at all as a strategy."
my answer is no, because of the cycle timing issues.

But the fact remains Steve insisted he is making consistent profits with the SSI and it is his most consistent method.

 
mike_smith

I'm looking for an up day on the eurusd. I don't know if it's today but probably today or monday the up day will happen. After it happened I'll sell. The setup is an up day, then the order is sell expecting one day to complete. Yesterday wasn't an up day so I'm not selling.

 
mike_smith

And there it is. Monday 0 gmt I sell.

2020-10-16 05_18_58-2100683976_ JAFX-Demo3 - Demo Account - EURUSD,M5.png
 
mike_smith

Well that sell didn't work out. Good thing to use that stop and diversify! Also not giving up on Perry's fxssi.com strategy. Thanks Steve. Just use risk management, not 1:1. Best thing I saw for that is at least 1/2ADR s/l. For example today buy eurusd with 1/2ADR s/l.

2020-10-20 19_22_52-FXSSI - Forex Sentiment Board.png
 
mike_smith

Here's my attempt to document if fxssi works for me long term. I entered 6 orders. At 0 gmt, ADR 0.5 s/l : 2 t/p. Did anyone try using no target and exiting the order after change of fxssi direction? That's basically following the contrarian trend. Following trend doesn't last forever because it needs to reverse, and that my friend is why we don't use 1:1 risk reward.

2020-10-21 16_07_35-Alveo.png
 
mike_smith

The above entries alone all failed. I think Mr Perry should use the fxssi as a general contrarian warning for market direction, but he'll still need to use an array of other technical indicators to pinpoint precision timing for actual trend reversal. And with proper risk management.

 
mike_smith

Practicing Rex's support and resistance inversions. Today here, I am shooting for GU 1.3013 t/p, 1.3075 s/l.

2020-10-22 21_43_08-2100683976_ JAFX-Demo3 - Demo Account - GBPUSD,M15.png
 
mike_smith

Well it didn't do what I thought right away, but I think it pays to know inversion, clearest way for me to use and see support and resistance anyway.

2020-10-23 04_25_00-2100683976_ JAFX-Demo3 - Demo Account - GBPUSD,M5.png
 
mike_smith

I can confirm that Alveo is removing at least T/P values from my orders. I have several demo accounts I trade simultaneously in Alveo and go back and forth throughout the day on them. Today several targets were removed from my orders with no intervention on my part. Take care Alveo traders! Watch your orders and don't trust this platform will not leave your orders unchanged! Like Dr Elder says, he prefers to eyeball only one order at a time, this is good advice.