YT YouTube2Text

YouTube2Text → Transcripts → CODING TRADING AUTOMATIONS | July 31 | Michele Rossi

CODING TRADING AUTOMATIONS | July 31 | Michele Rossi — Transcript

by FunderPro · 5,565 words · 930 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:02Code code is wrong.
0:16[music]
0:56What is going on guys? Welcome back.
1:01Test test test 1212. Here we are again.
1:04Welcome back beautiful people. Before we
1:06begin, I want to thank Thunderro for
1:09making this educational series possible.
1:11Thunderpro is in my opinion one of the
1:14best prop firms for traders who want to
1:16use algorithm trading. If you want to
1:19check it out, you will find the link in
1:21the description or on screen. Uh you can
1:24also use the code live to receive the
1:26best discount currently available. Now a
1:30very quick recap. In the previous video,
1:33we started with an idea, the opening
1:36range breakout. We transformed that idea
1:40into precise trading rules. We gave the
1:44specification to codeex. We generated a
1:46pine script strategy. We installed it on
1:49trading view. We fixed couple of errors
1:53very quickly and we created the first
1:56working version of our orb bot. It is
2:01called Orbinder Pro. Right.
2:05Um, if you missed the lesson, you can
2:07find it up here or here or in the link
2:10down below on YouTube. So, today we are
2:14going to take the first working version
2:17and improve it. But we need to be
2:20careful. There is a major difference
2:23between optimization
2:25and curve fitting. Optimization means
2:30testing a small number of reasonable
2:34alternatives to understand how the
2:37strategy behaves.
2:39Curve fitting means repeatedly changing
2:44parameters until the historical equity
2:48curve looks perfect. Today I will show
2:51you the different tests I perform to
2:55reach this result.
2:57Then we will inspect the settings and
3:00the results you can see right here.
3:02Okay. And then we will inspect the
3:06settings to simulate more realistic
3:08market conditions. Export the trades and
3:11use codeex shad GPT cla I will use
3:15codeex. You can use whatever you wish to
3:17validate the strategy. So by the end of
3:20the lesson we will not simply have a
3:22profitable back test. we will have a
3:25much clearer idea of whether the
3:29strategy deserves to move to the next
3:32stage. Right? So once again, we're going
3:36to use AI. It's free. It's available for
3:40everyone. Is it the best tool that you
3:44can use? Of course not. There are
3:47professional tools to do this in a much
3:50better profound precise detailed way and
3:54it will gives us also more precise
3:57results. But
4:00I mean it's free. You can use it. It's
4:03going to give you an overall overview.
4:06It it is going to teach you a clear
4:08process. It will guide you through uh
4:12how to create a bot, back test,
4:14optimization, validation, and deploy.
4:17So, we're going to use this.
4:20Um I'm now showing you the results I
4:24obtained
4:25um on the E mining NASDAQ futures using
4:30a five minute chart. As you can see
4:32right here, the test period show here
4:36runs from April 19 to July 31st. So I
4:40tested the last three months. The better
4:43things to do is go even way back in time
4:46like testing at 365 days or even more.
4:51As you can see, the strategy is still
4:53profitable even though in the beginning
4:55give us gives us a a bit of a draw down
4:59and it stays in the red for quite a
5:01while. So this might suggest is
5:06a regime part which is giving us this
5:09beautiful uh curve in the last 90 days
5:13but we're going to use that as for
5:15educational purposes only. Okay. So the
5:18strategy produced a total of 17,152
5:23which is a 17%
5:2617% with a maximum draw down of 9.15
5:32it gives us a 61.66
5:36um win rate with a profit factor of 1.3.
5:42Right? So the equity curve finishes at a
5:47new high and the last section is
5:50particularly strong. Do I like the
5:53results? Yes, I like it as a starting
5:56point. It is profitable after adding
6:00execution costs. It has a reasonable
6:03number of trades for an initial
6:06investigation and the equity curve does
6:09not depend on one single trade. But I'm
6:12not saying the strategy is ready. A
6:14profit factor of 1.3 is positive, but it
6:17does not gives us an enormous safety
6:20margin. The maximum draw down is also
6:23important for a prop. We made
6:26approximately 17% with the um
6:32with the one point the strategy was down
6:35across approximately 9%. That gives us a
6:39return to draw down ratio of around two.
6:42That is interesting. But a 9% draw down
6:44can already be too large for many prop
6:46firm accounts, especially when the
6:48strategy is risking 1% of equity per
6:51trade. But hey, you can reduce it to 0.5
6:55just as an idea. So
6:58this uh scenario is not the conclusion.
7:02it is the candidate that we are going to
7:05investigate again for educational
7:08purposes. I I I would go and back test
7:11even more. I would try to find better
7:13settings etc etc but
7:16we have a restricted amount of time to
7:18be live with you guys today. So we're
7:20going to take this last three months I
7:22went through just before the call I
7:24found this last three month. It's okay.
7:26We can back test it. We can optimize it.
7:28So let's open the strategy inputs right
7:31here. And these are the settings that
7:35produced the results. So from here to
7:39here, let me let me get my pen.
7:47Where is it? Where is it? Where is it?
7:49Where is it? Where is it? Where is it?
7:51Screen brush.
7:53Screen brush. Show screen brush. So what
7:56you want to look is from here to here,
7:59right? Um
8:03the first trade risk-to-reward is set to
8:060.8. The reverse trade option is
8:10enabled. The risk-to-reward ratio for
8:12the reverse trade is set to 1.0. The
8:15strategy is allowed to trade in both
8:18direction, long and short. And the
8:21active days are Monday, Tuesday,
8:22Wednesday, and Friday. Thursday is
8:25disabled. So, the opening range candle
8:28begins at 10:30.
8:31Um,
8:33be careful with the time zone. The
8:35numbers shown in the input.
8:39Uh,
8:42I mean, we got to be careful. We got to
8:44be careful with this opening time and
8:48and opening range session. Okay. I I was
8:52forgotten. I set this on Asia time zone,
8:55right? So, always be careful to find the
8:58right time align with the time zone,
9:01right? Um
9:04because a strategy can produce of course
9:07a very different
9:10results if you change the time zone.
9:12It's 10:30 in Italy. It's 10:30 in Asia.
9:16It's 10:30 in Spain. Okay. Um
9:20so the chart is the last final visual
9:23check. Now let's understand how I
9:26arrived to at these values. I did not
9:29ask codeex to test thousands of random
9:32combinations. So I divided the problem
9:35into small experiments.
9:39The main variable I tested
9:42was
9:45let me give a little bit more of room
9:47right here. Okay. So, the main variables
9:51were the risk to reward.
9:58There it goes.
10:01Um,
10:03of the first breakout
10:05uh whether or not allow
10:09reverse trade. So, that means we have
10:13our range right here. We got a breakout.
10:16We take a long position buy. We got a
10:19reversal and a breakdown. And we take
10:22another position, a sell like you can
10:26see it happened right here. First
10:29breakout, sell, stop-loss, second
10:31breakout, take profit, right? Um, so
10:36what I tested, what else I tested? I
10:39tested uh of course the risk to reward
10:45into the reverse trade. Uh I tested long
10:50only
10:52short only and both directions
10:57and I tested the day of the week. So by
11:04picking which day to exclude like for
11:08example I saw that Thursday was the one
11:12day where the strategy was performing
11:14the worst. So in this period of time so
11:17I decided it for this um educational
11:21content to remove it. But we can do it.
11:24It needs to have a reason. Maybe the
11:26strategy performed not so good when we
11:29have major news releases such as we have
11:33job data on Thursday might be the case.
11:36I don't know. So I removed it but it is
11:38logical. So every time I changed one
11:41concept I uh recorded the results. So I
11:47screenshot it. I screenshotted the
11:49settings. I created a a folder on my
11:52desktop or you can simply sign up on a
11:55piece of paper. Um this is very
11:59important. If we change five inputs
12:01together, we may improve the results but
12:03we will not know which change was
12:06responsible for that. So um before we
12:11running the tests, I will ask codeex
12:15to create a simple optimization log.
12:20Okay,
12:23then I will use this prompt. Let me open
12:26up codeex.
12:28This is last week uh
12:32chat where we coded the strategy.
12:35I have the prompt already done. So, let
12:38me pick it up. I don't want to waste
12:40time.
12:46Okay, this is it. And I'm going to read
12:49it for you guys.
13:09There it goes.
13:11So the Cody the the the prompt is I'm
13:14developing an opening range breakout
13:16strategy in Trading View. I have
13:18attached a screenshot of the candidate
13:21strategy results, a screenshot of the
13:22strategy inputs, and the screenshot of
13:24the trading view strategy properties.
13:27Read the visible settings and create a
13:29optimization experiment log. Use the
13:32screenshot only to record the visible
13:34configuration. Do not calculate missing
13:36statistics from the image and do not
13:38invent values that are not visible. The
13:41candidate configuration currently uses
13:44trade to risk to reward 0.8. So right
13:46here I have all the settings that are
13:49being used right now and I want you to
13:51create a CSV which is a text file named
13:55orb optimization log CSV with this
13:58column test ID test purpose the range
14:01instrument time frame first trade risk
14:04reward reverse trade enable blah blah
14:06blah all the things that we want to test
14:09and the visible candidate results as the
14:11first row mark any unavailable values as
14:15not provided. Do not recommend a final
14:19uh configuration
14:22yet. So this is the this is not useful,
14:27right? So what are we going to do? We're
14:29going to screenshot this
14:37and put it into codeex.
14:41We're going to scroll down.
14:44Screenshot this,
14:52put it into codeex.
14:56We're going to go to the
14:59returns,
15:01screenshot this,
15:06put it into codeex.
15:09And we can also give
15:13the list of the trades. We can download
15:16it. A CSV file going to be downloaded
15:19and we can put it into our
15:27right here. Codeex. We can screenshot
15:31the settings.
15:40put it into codeex
15:43and we're going to screenshot also the
15:45properties. This is a very important
15:48part because right here we're going to
15:50talk about it a little bit later. I gave
15:54um trading view stats and let it run. I
15:59gave trading view strats um initial
16:03capital default order sides bar
16:06detailization
16:07high 40 ticks per bar. As I said this is
16:10not full depth data.
16:13Okay. So this is a
16:18something that we can use as a reference
16:20as a starting point. These are not real
16:24full indepth tick data millisecond per
16:28millisecond. Okay, we're going to add a
16:30little bit of commission. We're going to
16:32add a little bit of slipage couple of
16:34ticks. We're going to add if we're using
16:36limit orders an extra layer of one tick
16:40beyond of sleeps and order execution
16:42delay. If we're using uh market orders
16:45is going to be delayed by one tick. So
16:47we're trying to simulate the the the at
16:50the best possible uh not best possible
16:54but worst possible conditions how the
16:57strategy would really execute on live
17:00market conditions. Right?
17:04Stop.
17:13So I'm using the spreadsheet skill
17:15because the requested deletable is the
17:17instructor CBS. I record only values
17:20explicitly visible blah blah blah. The
17:21spreadsheet instruction should point to
17:23local reference file skill fer and
17:25correcting the path. So codex is doing
17:28its own thing.
17:30Um what is going to give us is going to
17:33give us a good path to
17:38um
17:41to do what to open the settings and it
17:45gives us suggestion on how to
17:50change these parameters in a intelligent
17:53way. So to have multiple
17:57um results
17:59then we're going to give it to Codex
18:02again and he's going to tell us look
18:05this is what I believe on a mathematical
18:10and statistical
18:12um reference
18:15the best possible results that I
18:19believe. Okay. So, it's going to compare
18:21the net profit, the draw down, the
18:23profit factor, the win rate, the average
18:27um trade, and the trade count. Okay. Um
18:33I don't know how long it's going to
18:34take. Okay, good. Let's open it.
18:39Let's take a look.
18:43Okay, gives us is in a very bad
18:47situation to read. Let's try this way
18:54now. Way better.
18:57Okay. It's a It's a very long It's a
18:59very long thing. It's I I believe it's
19:01not created properly.
19:05Uh
19:10again, I I'm going to say to do it again
19:13and be sure to have the full list. Okay.
19:19It happens. So, but what is what it's
19:22going to give us is a list of
19:26do this this and that. Try these
19:29parameters. So, what you're going to do,
19:30you're going to change those parameters
19:33and you're going to write down and
19:36collect all the data, all the
19:38screenshot, all the results as I said I
19:40did before. Okay. So with these results
19:44you are going to compare all all the the
19:49results the draw down the profit factor
19:51etc. But let's say you have a set of 10
19:54different one two three four five six
20:00etc etc. You have a set of 10 different
20:04um [clears throat]
20:05outcomes.
20:07Do not automatically select the value
20:09with the highest net profit.
20:12Try to identify whether there is a
20:15stable region across
20:18neighboring values. Okay. Which values
20:22produce excessive draw down? Whether the
20:27apparent improvement
20:30comes mainly from a higher win rate or a
20:34larger average winners. It is very
20:39simple to spot because you have the full
20:41list of trades. But with this particular
20:43strategy is not going to happen because
20:46again it's a very fixed riskreward,
20:49right? Um
20:52whether any results appear isolated and
20:57potentially overfeeded.
21:00Okay. Um, so you're going to
21:07get these
21:10values that Codex is going to give you
21:14and give it back to him and say, "Look,
21:17I picked this one and this one and this
21:21one for this particular reason."
21:25Okay. What do you believe is the one
21:28that is the best for my strategy?
21:32Okay. Um let's say my candidate
21:35configuration the first trade target is
21:370.8R.
21:39That means that if the initial stop
21:41distance represents one unit of risk,
21:44the target is 0.8 times that distance.
21:46That target is smaller than the stop.
21:49But that is not automatically
21:52uh means that is a problem. So the
21:55complete results depend on the win rate,
21:58the reverse trade, the transaction costs
22:01and the combined expectancy of both
22:05entry type. Right? So um when we review
22:12all of these,
22:14the next test you want to do is um the
22:21the candidate settings. So let's let's
22:24let's
22:27pick these settings. So I just explained
22:30you
22:32how I get to these results, right? So I
22:34got to these results. I use a codeex in
22:36the way I showed you. Now we have these
22:38results.
22:42Given these results,
22:45right?
22:48I save this condition.
22:52Right? So, we're going to go to
22:57the cost stress test.
23:01So, I do not want to use only one
23:05execution assumption.
23:08So, I will repeat the test under
23:11different cost scenarios,
23:14right? And we're going to have let's say
23:17four four scenarios. One which is a
23:24um
23:26[music]
23:29optimistic
23:32right.
23:34Scenario B
23:37baseline
23:40right? Scenario C is going to be a
23:45conservative
23:49and scenario scenario four which is
23:51scenario B
23:55is a stress condition right let's say
23:58for scenario H A we have commission 0.5
24:02per contract slipage one tick for the
24:05baseline still 025 per contract two
24:08ticks of delay one tick For the delay,
24:12one tick limit verification. For the
24:14conservative, we're going to even raise
24:17the commission. We can do it three
24:19ticks. One tick, one tick. And for the
24:23stress, we're going to give $1 per
24:25contract, which is unreasonable, but
24:27we're going to do it again anyway. Four
24:30ticks of sleepage and more conservative
24:32field assumptions. Okay. And the purpose
24:36of this test is not to predict the exact
24:39cost of every future order. The purpose
24:43is to determine whether a small
24:46execution difference destroys the
24:49strategy.
24:51If the strategy
24:53is profitable only with perfect fields,
24:58again it is what you find into the
25:01properties.
25:03um it is not ready for live deployment.
25:06Right? So, we're going to export all the
25:11evidence and we're going to start to um
25:18to give um to sorry to start with um
25:23with this list of candidates, right? The
25:27list of candidates. This is our
25:29baseline, right? We have all of these
25:32trades. We downloaded all of these and
25:37what we're going to do is doing an in
25:40sample and out of sample validation. Let
25:43me write it down.
25:46So the next step
25:48is in sample
25:52and out of sample validation.
26:01Okay. Right. Um
26:06so we must now separate optimization
26:09data from validation data. The period
26:14used to select the 0.8R
26:17enable the reverse trade and remove
26:19Thursday's development data. This is it.
26:22We cannot use the same period as
26:25independent proof that those settings
26:27work. For a proper chronological test, I
26:30will define an earlier period of for
26:34optimization and the later untouched
26:37period for validation. So the later
26:39period must retain remain untouched
26:43until the candidate settings are fixed.
26:46So I'm I'm not going to reoptimize the
26:49strategy after seeing the validation
26:52results. So here's the prompt.
26:56Let me copy and paste because I prepared
26:58it in advance of course or it's going to
27:00take a long time.
27:24And we're going to give it the codeex.
27:36[music]
27:43Right. So I copied and paste. Oh, sorry.
27:46You're you're not seeing it. I copied
27:47and pasted it everything right here. So
27:51it is uh we must now separate
27:53optimization data from validation data.
27:57The period used to select 0.8 are enable
28:00the reverse shade and remove Thursday's
28:02development data. We cannot use the same
28:05period as independent proof that those
28:08settings work for a proper chronological
28:10test that will define an early period of
28:13optimization and the later untouched
28:16period for validation. Right. uh audit
28:19the file and confirm chronological
28:22separation, matching strategy
28:24configuration, matching instrument and
28:26time frame, matching execution customs
28:29and and assumptions. Compare the two
28:32periods using number of trades,
28:34expectancy, profit factor, win rate,
28:36average win, average loss, maximum draw
28:39down, longest losing streak, long versus
28:42short result, first versus reverse trade
28:46results, and return to drawdown ratio.
28:48Because the sample have different
28:52lengths, focus on per trade and
28:55percentage matrix rather than total
28:59profit alone. Identify which behaviors
29:02remain stable and which deteriorated
29:05materially uh is not working. Do not
29:08optimize the out of sample period and
29:12save the results as orb out of sample
29:15validation.
29:19Okay, we can have two scenarios. We can
29:22have a weaker out of sample results or a
29:27good out of sample results. A weaker out
29:30of sample results does not automatically
29:33mean failure.
29:35Some deterioration is normal. But if the
29:39strategy moves from positive expectancy
29:42to a clear and persistence persistent
29:45negative expectancy, the optimization
29:49probably captured noise. Right?
29:53Let's let it run. What we're going to do
29:56next? Next, we're going to do a walk
30:00forward test. So, a single in sample and
30:04out of sample split can still be uh
30:07influenced by the selected dates, right?
30:11So, we can perform a workforward test.
30:16For example, we can optimize on six
30:19month, we can test on the following two
30:21months. Uh we can move the window
30:24forward and repeat. Right? Let me say
30:28that again. This is very important. Let
30:30let me write it down. Um,
30:34optimize
30:37on six months,
30:40right?
30:42Sorry, months.
30:46One. Uh, we're going to test on the
30:48following month.
30:51[music]
30:54Let's say the following
30:58two months.
31:02We're going to move the window
31:07[music]
31:08forward
31:09and we're going to repeat,
31:13right?
31:15The strategy must always be developed
31:19using past data and tested using later
31:23data and codeex chat GPT uh Gemini
31:28whatever uh clo can help organize
31:33results but trading view must generate
31:36the trades for each window.
31:39Okay, is that clear? you have any
31:41question
31:49right? Yes. Yes. That that is a good
31:52question. Um
31:56these are
31:59the the candidate settings. So our
32:02candidate uses 0.8R for the first trade
32:05and one R for the reverse trade.
32:09I tested and you have to do so. So I'm
32:11telling you to test neighboring values
32:15for example first trade 0.6 0.81 and 1.2
32:19reverse trade 0.81 and 1.2 I've already
32:23done that and already gave that to
32:26codeex otherwise he cannot produce what
32:29he's doing right now but it it take
32:32quite a bit of time and we cannot stay
32:34here all day long. So I prepared some
32:36stuff before and
32:39in this way we're going to have
32:43and use
32:46a small parameter matrix let's say. So
32:50we're not searching for a perfect
32:52combination. We are checking whether the
32:54candidate sits inside a stable region.
32:59Right?
33:01Okay,
33:05let's see what point we're at.
33:08Beautiful.
33:10Okay, let's see.
33:14Validation design. The experiment.
33:22The experiment user chronological load
33:24out in sample and out of sample
33:27configuration audit for this educational
33:29example. board export assumed to contain
33:31the following identical configuration 5
33:33minutes 5 minutes 1030 1030 100% of
33:37equities so it gives us a lot of data
33:42136 in sample 68 out of sample
33:46uh winning trades 8240 5428 60% win rate
33:5158%
33:53average win 1 average win 925
33:58and we got of
33:59different very different results even if
34:04still positive 33% and 5.64%
34:08that's a huge difference but at the same
34:11time maximum draw down nine maximum draw
34:13down 6.4 four, longest losing streak,
34:17five trades. Longest
34:20losing streak, four trades. And we have
34:23all the results that we talked about,
34:26right? So, we have our in sample and out
34:30of sample validation.
34:33Um,
34:36the next the next thing that we have to
34:40do is with this file right here. Yeah.
34:44still with the same file. Uh we're going
34:47to give it to Codex and ask him to run a
34:54Monte Carlo test. Right.
35:00Let's give it a Monte Carlo test.
35:07So perform random reshuffleling of the
35:12existing trades without replacement.
35:15Bootstrap resampling with replacement
35:18block bootstrap resampling of
35:21consecutive trades using several
35:23reasonable block sizes to preserve some
35:27short-term dependence. Use the fixed
35:30random C2 to so the analysis is repro
35:34reproducible
35:36for each method. Calculate the
35:38distribution of final profit, maximum
35:40draw down, longest losing streak,
35:42minimum equity, and return to draw down
35:45ratio. report medium 75th percentile
35:5090th percentile 95%
35:54percent percentage sorry for my English
35:56and 90 99%
35:58draw down also report
36:01probability of finishing below zero
36:05probability of exceeding the historical
36:089% draw down probability of exceeding a
36:1110% draw down because
36:14we're we're already forward thinking
36:17about where we're going to use this
36:20inside of a prop account and 90 and
36:2595th percentile losing tree clearly
36:29explain that this analysis does not
36:32model future market regime change. So
36:34this is a protection that [laughter]
36:37codex is giving but it's it's totally
36:39fine. So we took the file we take the
36:41file once again. This is it. We're going
36:44to give it and we're going to ask one
36:48extra question which is
36:52give me an
36:56HTML file with a visual representation
37:03of the Monte Carlo test. Okay, let's go.
37:09So
37:11um
37:16what do we know? What do we know? Is it
37:18working? Right? Is it working? Uh we
37:21know that the historical draw down is
37:279%. Right? But after Monte Carlo
37:30simulation, we may discover a 12 or 15
37:33or larger draw down. It's plausible.
37:37This is particularly important because
37:39the strategy currently uses a 1% chance
37:431% of equity per trade. So for many prop
37:47firm accounts that might be too
37:49aggressive. If we reduce the risk from
37:511% to 0.5 the approximate percentage
37:55impact of every trade and draw down will
37:58also be reduced. So we we we should
38:01choose the risk based on unfavorable but
38:05plausible scenarios not the historical
38:08results. Is that clear? Do you guys have
38:12questions about this?
38:14This is very important.
38:19Right. So let's see what it gives us.
38:24It's going to take a little bit of time.
38:26So if you have question this is a good
38:28moment to ask
39:19>> [snorts]
39:34>> So, let's do this while we wait. It's
39:36going to take quite a bit of a time. Um,
39:40let's
39:41let's recap what we just did, right?
39:48Um,
39:59today we begin with the first working
40:01version of our opening range breakout
40:05strategy that we coded
40:08last week.
40:10This is the code, right?
40:18And we tested the first trade target
40:24reverse trade logic direction and day
40:26filters. We reached a candidate
40:29configuration with a 0.8 R
40:35uh first target 0.1
40:38second target reverse both directions
40:41enabled and Thursday disabled. And on
40:44the period show we got a 17% of total
40:48profit and loss. Max draw down on 9.15%.
40:53Win rate 61% profit factor 1.3.
40:58So we like it. We like this as a
41:01starting point. But we did not stop
41:04right here. That would be a a a crazy
41:07error. We added commissions. We added
41:13sleepage, order delay, limit
41:15verification and intraar detail.
41:20We exported
41:22the trades from here
41:25right
41:27into codeex together with the
41:28screenshot. Then we designed an out of
41:31sample test, work forward analysis, a
41:34parameter stability, cost stress test,
41:37multicar simulation. And the last thing
41:41that we're going to do probably
41:45probably next week. Yes, I believe we're
41:48going to do it next week. We're going to
41:51do a prop firm risk analysis. Right?
41:55That is going to be very very important
41:57and very very useful.
41:59And this is the difference between
42:03optimizing
42:04a strategy and believing an
42:07optimization.
42:09So the back test gives us a candidate.
42:13Validation decides
42:16whether the candidate deserves another
42:18step.
42:20Right? So in the next final episode,
42:24we're going to talk about prop firm risk
42:27analysis. that deal with the same
42:29results, same multicar results that
42:32Codex is producing right now. And we're
42:35going to go in depth on that. So we're
42:37going to note
42:39how much
42:41the the probability in percentage that
42:45we have to pass a two-phase challenge.
42:49So we're going to have let's say you
42:50have 63% of passing a the first stage of
42:55the classic challenge of Funpro. you
42:57have 73%
43:00of chance to pass the second phase and
43:04you have let's say I don't know um
43:0872%
43:11just random number just to give you an
43:13idea of
43:17probability to get your first $8,000
43:20payout
43:23right
43:24Um,
43:28and we we will also
43:32do the same. We're going to run the
43:34strategy in forward testing. We're going
43:37to create trading view alerts.
43:41We're going to build a structured web
43:45hook alert,
43:47which is this thing right here, right?
43:52Um,
43:54and we're going to connect the signal to
43:58the execution platform can be any third
44:01party platform. I'm not going to name
44:04any, but I'm going to tell you how it
44:06works, right?
44:09And um yes,
44:13of course, we're going to do couple of
44:15other things to add security like
44:20prevent duplicates or delayed orders. Uh
44:23add the daily loss protection or create
44:27a red button, an emergency stop button,
44:32right? and
44:35and we can prepare the bot for control
44:38use on a fun pro account. And what I'm
44:42going to do, I'm going to actually run
44:44this strategy on a fun pro two-phase
44:48classic challenge. Right?
44:52So this is where the project becomes a
44:55complete trading system rather than only
44:57a back test. I know today uh lesson it's
45:02quite complex. It's not very easy for me
45:06to explain. I hope I did my best to
45:09explain you guys all the steps, but you
45:11can go back in time and rewatch if you
45:13miss some part.
45:15It's not going to take more for you to
45:18do it more than a manual back test. I
45:22see people continuously going back and
45:26forth and back testing manual strategy.
45:29That's to me totally stupid. It's good
45:34at the beginning of your career to
45:38put your hands in the mud and understand
45:41stuff, watch things, create ideas, learn
45:45how to manually execute.
45:48>> [snorts]
45:48>> But when you got a little bit of
45:51experience,
45:53there's no point of
45:56one, trade manually and two,
46:01not to code a strategy, back test it
46:04automatically and deploy it. Right? So
46:09we have our results. Let's go take a
46:11look. There it goes.
46:14complete the completed the reproducible
46:17multiarl analysis uses 68 unique close
46:20trade. Again, this is a very small
46:24amount of trades. We want to go with a
46:26one year,
46:28500 trades, 600 trades, 700, 1,000
46:31trades. That is going to be best, right?
46:35Uh reshuffle
46:38bootstrap. We got the block three, we
46:41got the block five, and we got the block
46:4310. Look how beautiful this thing is and
46:46it's very very easy to use and
46:48understand.
46:50Okay. And you have all the matrix
46:52maximum draw down by by percentile.
46:56Uh with the reshuffle only we have z%
47:00probability to finish below zero. We
47:03have 50% finishing above 10%. And the
47:0890th 95th
47:11it's very difficult for me to pronounce
47:14percentile draw down is 16.3%.
47:19The selected method profile we got all
47:24all the summarizing right here.
47:29So
47:31what you can tell at the moment about
47:36this strategy
47:41is it going to be this is not a good
47:44prompt but to show you guys what we're
47:46going to do next week to be
47:49useful and potentially
47:53profitable
47:55on a Thunder Pro two steps classic
48:00challenge
48:02account.
48:03Am I going to
48:06um how
48:08many statistical
48:13probabilities
48:18[music]
48:23I would have
48:26to pass
48:29phase one and to pass phase two and to
48:35get a $1,000
48:39payout
48:41if and once funded.
48:45So, next week we're going to have a
48:47quick look, but next week I'm going to
48:49show you guys how to design
48:54stealing codeex or CL or whatever
48:59and give it a proper prompt
49:04to
49:06um
49:08understand
49:10if the strategy is going to be good and
49:15or profitable inside of a strat inside
49:18of a prop firm account. So, we're going
49:21to run a prop firm risk validation,
49:26right? We're going to give a proper
49:30um prompt. This is not a good prompt.
49:33This is a a super basic baseline, right?
49:37I'm going to show you how so you can
49:40copy paste the prompt that you can use.
49:44Right. Okay. Let's see what it what it
49:47says and we're going to finish this
49:50lesson. If you don't have any question,
49:52let me see in the chat if you have any
49:54question. Hi
50:01Roberto. Do you ever trade too or is he
50:04just talking? Roberto
50:08GF YS.
50:11Google it.
50:15I traded for more than one year. Allies
50:18recorded live. We took profits. We took
50:22losses. We took payouts. So this is an
50:25educational content. If you don't like
50:28it, GF YS,
50:33Google it.
51:02There he goes. Short answer. The
51:04strategy shows a potentially real edge,
51:06but I would not run it unchanged on a
51:08paid founder pro challenge. It is
51:10currently sightsee
51:13beneath fun prodown limits and the back
51:15test instrument blah blah blah blah
51:17blah. What the strategy currently
51:18demonstrates blah blah blah blah blah
51:20blah blah blah. Okay. So, what it's
51:22saying is
51:25we have pass phase one, we have a 79 to
51:3181% of chance to pass.
51:36Phase two, we have 85% and
51:40to get funded, we have a 93%
51:43of getting a payout. It says it's not
51:47perfect. We have to make it better. But
51:51this is a good starting point, right?
51:54So, thank you guys for being here with
51:56me today. Thanks again, Founder Pro for
51:58making this possible. I believe this is
52:00one of the best ever educational content
52:06that you can find on a prop firm because
52:10I'm telling you guys how to actually
52:15fool the prop firm. So, Fun Pro is going
52:19to hate me for that.
52:21Maybe they're going to delete this
52:25this series because if you follow this,
52:29hey, I'm not I'm not telling you to do
52:32it, but if you do this, you have a high
52:36chance to
52:38prepare, do nothing, watch the account
52:41growth, pass, and get your payout.
52:48That's it.
52:50Yeah, but Roberto, you have a SG and
52:54you're never going to be profitable.
52:57So, enjoy it.
53:00Um,
53:02so guys, thank you so much for being
53:03here with me today. Thanks for the hate
53:06from Roberto.
53:07Um, I wish he have a great career in his
53:11life. And thank you so much to my life
53:16for all the good questions. If you have
53:19any further question, ask it down below
53:22in the comment section. I'm going to
53:24reply every question. Thank you so much
53:26and I'll see you again on Monday on
53:29Wednesday for live trading and weekly
53:32preparation, midweek preparation. And
53:34next week on Friday, same time, we're
53:36going to finish the setup, have the
53:39final bot, and deploy it live and see
53:41how uh it's going to work. Thank you.
53:44Cheers, guys. Bye bye. Bye bye.

About this transcript

This page contains the full transcript of CODING TRADING AUTOMATIONS | July 31 | Michele Rossi by FunderPro, generated from the public captions YouTube serves with the video. The transcript has 5,565 words across 930 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.