[Journey] - trading bot that may actually work

[AlexAkimbo]

Here is some food for your thought

https://github.com/jimtin/how_to_build_a_metatrader5_trading_bot_expert_advisor

 

[francisirving]

bot: a_2_b
timeframe: 3M
no stop loss, no take profit

heavily skewed trading this morning. last night wasnt great (maybe down $7) but going into Fridays is usually slower, at least I've been seeing. But then the 8:30am est jobs report came out, sent the market down and the bot had me in a short at the time, so, yay.
NASDAQ
18 Losses -$9.75
6 Wins $ 20.6
total change +10.85

I mean, technically that's what the bot is for; catch the runners. but still a nice, lucky break

there's potential with the AB for overnight, but it's also not a bad bot for live scalping from what I've seen (using trend analysis and support/resistance) but that's not what this thread is about. I'm going to look into using MT5 do get some indicator data this weekend like AlexAkimbo mentioned and maybe I can get a few more bots running concurrently without using pyautogui

 

[francisirving]

Not much to report today
good news: the bot is working pretty damn good from a technical standpoint
bad news: not profitable at this time
For the rest of the week I'm just going to let it run and collect data, because that's what I really need right now. I can backtest to the moon but its just not going to be as accurate as forward testing with a demo account. Sunday and Monday nights weren't great because I had a problem with the 'a2b' bot because I threw it together to only track 1 trading pair, oops. A quick fix once I realized it. Tuesday night was pure data collection, make some modifications to stop loss and take profit, and last night was data collection with the adjusted parameters. Here's a snippet of what happened last night into this morning:


XAUUSD and GPBAUD were profitable, the rest were not. As you can see EURUSD was a disaster, but XAUUSD and EURUSD do not have a take profit set, just a stop loss. I'll dig through the data and see if adjustments should be made.

At this point I just need to let the bot run, track the results and make TP/SL changes to see how it goes. I'm pretty bummed that letting a trade run from the start of an indicator to the start of the opposite indicator didnt work out as I'd hope, but, if SL/TP ends of being profitable I dont care how it gets there.

 

[SyntaxSurge]

seems that the bot is not profitable most of the time, perhaps you might need to change the strategy and logic. Add more confirmation like RSI, EMA, Bollinger band or maybe you can based on order books

 

[Ratangus]

Not much to report today
good news: the bot is working pretty damn good from a technical standpoint
bad news: not profitable at this time
For the rest of the week I'm just going to let it run and collect data, because that's what I really need right now. I can backtest to the moon but its just not going to be as accurate as forward testing with a demo account. Sunday and Monday nights weren't great because I had a problem with the 'a2b' bot because I threw it together to only track 1 trading pair, oops. A quick fix once I realized it. Tuesday night was pure data collection, make some modifications to stop loss and take profit, and last night was data collection with the adjusted parameters. Here's a snippet of what happened last night into this morning:

View attachment 290860
XAUUSD and GPBAUD were profitable, the rest were not. As you can see EURUSD was a disaster, but XAUUSD and EURUSD do not have a take profit set, just a stop loss. I'll dig through the data and see if adjustments should be made.

At this point I just need to let the bot run, track the results and make TP/SL changes to see how it goes. I'm pretty bummed that letting a trade run from the start of an indicator to the start of the opposite indicator didnt work out as I'd hope, but, if SL/TP ends of being profitable I dont care how it gets there.

You are making good progress, keep it up.

 

[White Night]

what kind of ema rule you are using?
why the results is different too much for each pair?

 

[francisirving]

seems that the bot is not profitable most of the time, perhaps you might need to change the strategy and logic. Add more confirmation like RSI, EMA, Bollinger band or maybe you can based on order books

the indicator I'm currently using (SWING CALL) uses ema, sma, and an overbought/oversold RSI to determine when to announce a buy/sell. On my test machine I've been playing with more indicators to see what happens. To be fair to the SWING CALL indicator, when I set the RR to 1 it was averaged about a 45L/55W rate. Not great, but the fact that its near 50/50 is a good start point for me as far as determining stop loss and take profit.

 

[francisirving]

what kind of ema rule you are using?
why the results is different too much for each pair?

on the SWING CALL indicator its currently using these inputs:
ema_value = 5
sma_value = 50
RSI overbought = 80
RSI oversold = 20

Why are the results so different for each pair? That's a great question. My guess would likely be differences in trading volume affecting the volatility. For example, NZDCAD has a lot less swing volatility then NAS100. That's not the best example to compare because NAS100 is like a roller coaster ride, but, based on my testing and 1 minute or 3 minute timeframes, NZDCAD just doesn't swing enough to make the wins worth more than the losers.

 

[francisirving]

bot: a2b
1M timeframe
various stop losses based on the pair, but no take profit, just letting the trade close once the opposite trade signal pops up.



A good trading night over night. XAUUSD actually looks better than what I posted because I took this sample from 18:00 to 09:00 EST, and by the time I put the results together XAUUSD made another +$9 on a good trade.
I dropped NZDCAD, GBPAUD and CADJPY as the small swings just aren't profitable enough for me at this time.

The a2b bot just seems to work better on the 1 minute timeframe with no take profit and hard stop losses based on trading data. The nicest thing about the a2b bot (other than when you get a trade that's a runner, like this)

is that when you get sideways moment (series of open trades/close trades in a row), is that the close trade signals are usually less than the hard capped stop loss.
For example, on XAUUSD the hard stop loss was set to 120 ($1.20 at 0.01 lots)


As you can see in this snippet, the hard stop loss only hit once (-$1.10) and the rest closed at the next buy/sell signal for considerably less so. In this case, tracking each trade shows me that the 120 cap could be reduced to 90 since when these trades fail, they usually fail fast. I also made slight adjustments to the stop losses for NAS100 and EURUSD

On to next week now. Once I get a little more settled in what strategy I want to stick with I'll rest my demo account and start the totals from scratch.

 

[francisirving]

just a little note about how the bot can be useful in my live trading. below is a trade (circled) that the bot took when it saw the 'sell' indicator pop at around 19:55 UTC+3

The purple square is the 1 hour chart support level that I drew up:

I manually closed just below that level, and this time it started to retrace. It wont always do that at the support level, but support/resistance is a strong risk management tool. The bot would have closed at the next signal (the blue 'B') so either way this would have been a profit, it's just bigger since I closed below the support level vs waiting for the indicator to trigger.
(image edited out ticket #, prices, etc)


Support/Resistance is what I'll likely start working on next at some point, I just have to wrap my head around how I'm going to do it.

 

[francisirving]

Been a mixed bag the last few days. I ended up dropping everything but NASDAQ as its the only consistent winner, or at least the biggest mover with the most gains and losses. I held on to EURUSD for a while but it's just not a big money maker, and profitability is quite low. You'd really have to crank up the trading lot size to match NASDAQ which I don't care to do.

Today was a nice day. I let the bot run from Wednesday 6pm est to Thursday 8am est and it was actually more profitable than using it as a live trading assistant during the 9:30am NASDAQ open, but I'm finding letting it run overnight is not very profitable most the time.

This first section is letting the bot do its thing until about 8am est:

Indicator: VMC Cipher_A B&S Signals

bot: a2b
stoploss/takeprofit: none
volume: 0.02 lots
wins: 5 +30.06
loss: 6 -10.98
gain/loss +19.08


I had 2 big winners overnight that make it very profitable. 1st is the $7.80 winner, and the 2nd is the $15.24 winner



Those worked out great but are not as common as I'd like. This is one of those situations where had I set a 60pip take profit I would have made about the same, so, I'm still going to run with no take profit for tonight. Looking at today's results I might set a 13pip stoploss, it will shave a little off the losers but nothing significant overnight. For day trades, it will help a lot more as daytime is a lot more volatile.

Next up is the trading assist/manual trades starting after 8am est. If you are familiar with NASDAQ there is usually a big opening at 9:30am everyday that goes wild at first then usually slowly simmers down until market close at 4pm

bot: a2b
stoploss/takeprofit: none
volume: 0.02 lots

First, here's the chart to show you the results



wins: 4 +$28.09
loss: 2 -10.06

There's a 'buy' signal and 'sell' just after the +33pip win that didn't get picked up for a trade. I think that was a case where the signal popped to far after the fact for the screen capture to pick up, and/or I might have been tweaking the bot at that time. Worked out fine for today.

As you can see, I had 2 manual stops. The 1st one (+32.7 pip) would have likely been a small loss if I hadn't stopped it but would have made almost double if I added a 60pip take profit. Not going to for now, but I might add a hard take profit if the data backs it up over time. The 2nd manual stop would still be going for, again, an easy 60pip win.
The -18.8 and 31.5 pip losses could be better mitigated with a 13pip stop loss, which I'll be adding tonight.



update: shit, she's still going, maybe the bots just a better trader then me.



Overall, a really nice day.

So for tonight:
- adding a 13pip stop loss
- letting the bottom run overnight again, just NASDAQ

 

[francisirving]

Another good day, especially since the bot missed a couple trades, trades got stopped out and I quit for the day early.
bot: a2b
stoploss: 13pips, no take profit
pair: NASDAQ
lot size: 0.02

Losses: 7 -$18.22
Wins: 6 +28.08
gain/loss: +$9.86

A few interesting trades to look at. This first screenshot shows 2 different trades the bot took in the early morning:

The 1st trade, the 13.6pip loss, stopped out where I told it, but seeing where the 'buy' signal popped this trade would have closed anyway and likely at a slightly less loss then 13.6 pips.
The 2nd trade, the 13.2pip loss, was a big, missed opportunity. The trade stopped out at 13pips like I told it to, but if there was no stop loss set this would have closed around the blue arrow, just after it saw the 'sell' signal at the top. This would have been around a 38pip win had there been no stop loss. The risk is, of course, with no stop loss is if the next signal was missed by the bot, or if my trading bot machine lost power and the trade just kept running into who-knows-what it could be big losses.
This is just trial and error for now, as the days go on the output will tell me what a good stop loss will be.

Another missed opportunity:


Lol this one is pretty heartbreaking. The stop loss was hit after a few minutes, like I told it to, but this would have ran for about 153 pips if there were no stop loss. Right now I'm thinking of doubling the stop loss to 25 pips, just in case something bad happens like I said above (missed close signal or loss of power).
Also, based on testing most of these losses at night are minimal when the close trade signal pops. They been averaging between 3pips on the low end to 40pips on the high end, most around 10-14 pips losses just letting the buy/sell signals dictate the opening and closing of positions.

The rest were uneventful; a small gain, manual closed loss and manual closed gain.



Onto next week. I'm feeling good about the progress. I might make adjustments to the screen capture area this weekend to see if I can tighten the signals up. Right now I'm scanning a small x area from top to bottom, like this:


But on the bot machine I've change the buy/sell indicators from these large arrows at the point of entry to smaller lines that appear at the top of the screen, like this:


So I should only have to scan this top area now, sorta like:



I'll get that going and ready for test start of the trading week Sunday 6pm est.

 

[francisirving]

Discover the use of bot for trading crypto in some days through the search engine, although I don't know how perfect the bot in gaining or returning profit

Thats the big trick. I finally caught on after a year of and half of learning how to trade what Nick Sean on youtube has been saying from the start:
Don't try to predict if the price is going to go up or down, that's a fools errand. You need to find a strategy that will get you about a 50/50 win/loss rate and tweak your trade to earn more when it wins and lose less when it loses. That's it, but it's a lot of work to get there.

Nick uses support/resistance, I'm using this 'VMC Cipher_A B&S Signals' indicator on Tradingview to get the trade entry signals.
My bot does nothing special at all, you could manually trade this by setting alerts in Tradingview when the signals appear and close them via stop loss or when the next signal appears. A lot smarter people then me can do this a lot more efficiently with a better bot, this just works for me.
If you look at the screenshots that show the long and short position outcomes, you'll start to see why I'm setting my stop losses or take profits where I do by backtesting them. Right now I'm not setting a take profit, but that could change as the data comes in day by day. For me this is a marathon, not a sprint, so I'm happy to get as much trade data I can before I try this on a live account.

 

[Esvityme]

Another good day, especially since the bot missed a couple trades, trades got stopped out and I quit for the day early.
bot: a2b
stoploss: 13pips, no take profit
pair: NASDAQ
lot size: 0.02

Losses: 7 -$18.22
Wins: 6 +28.08
gain/loss: +$9.86

A few interesting trades to look at. This first screenshot shows 2 different trades the bot took in the early morning:
View attachment 293000
The 1st trade, the 13.6pip loss, stopped out where I told it, but seeing where the 'buy' signal popped this trade would have closed anyway and likely at a slightly less loss then 13.6 pips.
The 2nd trade, the 13.2pip loss, was a big, missed opportunity. The trade stopped out at 13pips like I told it to, but if there was no stop loss set this would have closed around the blue arrow, just after it saw the 'sell' signal at the top. This would have been around a 38pip win had there been no stop loss. The risk is, of course, with no stop loss is if the next signal was missed by the bot, or if my trading bot machine lost power and the trade just kept running into who-knows-what it could be big losses.
This is just trial and error for now, as the days go on the output will tell me what a good stop loss will be.

Another missed opportunity:
View attachment 293002

Lol this one is pretty heartbreaking. The stop loss was hit after a few minutes, like I told it to, but this would have ran for about 153 pips if there were no stop loss. Right now I'm thinking of doubling the stop loss to 25 pips, just in case something bad happens like I said above (missed close signal or loss of power).
Also, based on testing most of these losses at night are minimal when the close trade signal pops. They been averaging between 3pips on the low end to 40pips on the high end, most around 10-14 pips losses just letting the buy/sell signals dictate the opening and closing of positions.

The rest were uneventful; a small gain, manual closed loss and manual closed gain.

View attachment 293004

Onto next week. I'm feeling good about the progress. I might make adjustments to the screen capture area this weekend to see if I can tighten the signals up. Right now I'm scanning a small x area from top to bottom, like this:
View attachment 293005

But on the bot machine I've change the buy/sell indicators from these large arrows at the point of entry to smaller lines that appear at the top of the screen, like this:
View attachment 293006

So I should only have to scan this top area now, sorta like:

View attachment 293007

I'll get that going and ready for test start of the trading week Sunday 6pm est.

Nice journey, I'm following.

 

[JianovMarik]

nice journey. Good luck bro

 

[Conceart]

the indicator I'm currently using (SWING CALL) uses ema, sma, and an overbought/oversold RSI to determine when to announce a buy/sell. On my test machine I've been playing with more indicators to see what happens. To be fair to the SWING CALL indicator, when I set the RR to 1 it was averaged about a 45L/55W rate. Not great, but the fact that its near 50/50 is a good start point for me as far as determining stop loss and take profit.

50% win rate seems decent for me. Nice journey!

 

[francisirving]

Sunday evening 6pm start to about 8:20am Monday was underwhelming, but not a disaster. only 2 wins but 1 was huge:

The win/loss was terrible that night. It edged out a small gain with a 25pip stop loss. Meh, the stop loss can be tighter, but, then 9:30am start happened and it went downhill from there.
Below is the 9:30am start of the NASUSD. Watching this live, every ounce of my being knew to get out around the circled area, especially after that 1st big green candle at the bottom, but since I'm testing a bot on a demo account I let it go. Took about a 26 pips loss here.
That's fine, that's within the bots parameters.



What wasnt fine was that it missed the buy signal to open a long once it closed that short. More then likely the green indicator was past the screenshot area and it didn't see it.
As you can see that long went and went and went.


That would have been a 175pip win. Luckily for me thats the kind of tweaking and testing I'm working on right now, not program crashes. In this case it was likely due to the time.sleep between image snaps (for instance the program pauses around 58 seconds between checking for an indicator. I dont want to check too often because it can place the same trade twice if the 1st ones stops out. I'm shooting for 1 attempt per signal but I have an idea for that).


That put me in the red for the day. Had it caught that huge runner mentioned above that alone would have been a +$35 trade, but, oh well. Keep testing and tuning, that's why I'm here.

 

[francisirving]

Monday 6pm start to around 8:20am Tuesday AM went beautifully. Tightened the stop loss from 25pips to 18pips, set the time.sleep image capture from 57 seconds to 45 seconds.


Wins: 9 +24.36
Loss: 5 -7.62

I should have stopped while I was ahead because then the 9:30 NASUSD hit.

I've been trading the NASUSD start everyday for the past 6 or so months. I'm no expert, but today was an even more bipolar day for the NAS then normal. It lost all the gains from overnight and then some. Even as I'm writing this the punches keep coming.

So you'd think the easiest solution is to stop the program before 9am, and that may be correct. The problem is that the NASUSD usually farts around a bit at the start then takes off in one direction for a nice profit, like I saw yesterday. I'm going to put some thought into it and see if there's a good risk management solution for the 9am to 4pm trading day that can help with this. Maybe manual trades or smaller stop losses, not sure. Or maybe this was just a crappy trading session and it happens. On the plus side, the bot worked very well but I did bring the sleep timer from 45 seconds 50.

 

[coinrpn]

Monday 6pm start to around 8:20am Tuesday AM went beautifully. Tightened the stop loss from 25pips to 18pips, set the time.sleep image capture from 57 seconds to 45 seconds.

View attachment 293752
Wins: 9 +24.36
Loss: 5 -7.62

I should have stopped while I was ahead because then the 9:30 NASUSD hit.
View attachment 293758
I've been trading the NASUSD start everyday for the past 6 or so months. I'm no expert, but today was an even more bipolar day for the NAS then normal. It lost all the gains from overnight and then some. Even as I'm writing this the punches keep coming.
View attachment 293760
So you'd think the easiest solution is to stop the program before 9am, and that may be correct. The problem is that the NASUSD usually farts around a bit at the start then takes off in one direction for a nice profit, like I saw yesterday. I'm going to put some thought into it and see if there's a good risk management solution for the 9am to 4pm trading day that can help with this. Maybe manual trades or smaller stop losses, not sure. Or maybe this was just a crappy trading session and it happens. On the plus side, the bot worked very well but I did bring the sleep timer from 45 seconds 50.

Why you stopped journaling here? have you stopped trading?

 

[francisirving]

Havent stopped yet, just decided to concentrate on finding some solutions to various problems.
The 1 minute timeframe just wasn't working out for me so after doing a lot of back testing the other week I decided to do a soft reset:

Week of Nov 6 to Nov 10
- start a new demo account with $200. this would mimic what I would start with testing a live account
- trade NASDAQ, XAUUSD, and USDJPY/AUDCAD/CADJPY/GBPAUD/EURUSD (just some random forex pairs for testing)
- 1 hour timeframe. this allowed me to slow down the bot and get a baseline of a higher timeframe to see how it goes
- no stop loss or take profit, also would manually close trades to take profits during the day depending on what my gut told me
- minimal lot sized for NASDAQ and XAUUSD since a $200 account could get blown rather quickly if there are not stop losses and the program crashed/things went sideways

long story short, did pretty well for the week, 9+% increase in account size (pic is obviously edited to relevant information)


pretty nice week. I figured if this was an average week all I'd have to do is double the lot sizes and things should move along nicely.
But I didn't, I'm making modifications to the chart timer, lot sizes and stop loss and so far this week has been insane, I'll put all those details in my next post.