YT YouTube2Text

YouTube2Text → Transcripts → I Re-Created A Quant Trading Strategy With Claude Code (Nobel Prize Method)

I Re-Created A Quant Trading Strategy With Claude Code (Nobel Prize Method) — Transcript

by Miles Deutscher Finance · 4,633 words · 689 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00There's one calculation that every
0:01single quant desk on earth runs every
0:03single morning, and it's not which way
0:05the market's going. Most traders don't
0:07even know the real question that it
0:09answers. At the center of it sits an
0:11equation that won a Nobel Prize in
0:13economics. And in this video, I'm going
0:15to rebuild the way a quant actually
0:17thinks from the ground up. Here's
0:18exactly what you're going to walk away
0:19with. The full method explained properly
0:21so you don't need any sort of math or
0:22technical background. A cloud code skill
0:25which runs it for you on any asset that
0:27you trade. By the end of this video,
0:28you'll have it fully installed on your
0:30computer so you can have quant-like
0:31intelligence at your fingertips. And a
0:33custom TradingView indicator that puts
0:35the entire framework in a quantifiable
0:38data set on your chart. And here's the
0:40part that a lot of videos like this
0:41don't include. We're actually going to
0:43go backtest the data for 15 years and
0:45look at how this strategy actually
0:47performs with real market data. Oh, and
0:50all of this is for free. I'm going to
0:51show you how to install it at the end
0:52using the link in the description below.
0:54You can claim all the assets and prompts
0:56from today's video. So, let me give you
0:57some quick context as to why I'm
0:59covering this. I've been deep in the
1:00markets for the last 7 years, and I
1:02noticed a few years ago that there was a
1:04massive shift happening with AI. I was
1:06actually one of the first people to put
1:07out AI trading bot videos on YouTube.
1:10And since then, we've had an absolute
1:11explosion with the intelligence of these
1:13models, which in my opinion make them
1:15absolutely essential if you're a trader.
1:17So, the fact we have this information at
1:19our fingertips, the fact that you can
1:20now compress strategies into a single
1:23skill, into a single file that you can
1:25use over and over again, opens up a
1:27plethora of opportunities. So, in
1:29today's video, I wanted to take a
1:30concept that real quant traders use and
1:33condense it down into something that you
1:34can tangibly interact with. And that's
1:36the power of AI. You now have the power
1:38of strategies and indicators and data
1:41that quant firms used to have access to.
1:42And as a retail trader, you can now have
1:44access to the exact same stuff. So, I
1:47personally believe there's a massive
1:48edge for anyone who augments their
1:50trading with AI. But in order for you to
1:51understand what we're actually building
1:52here, let's talk about the theory behind
1:54today's video. I've basically reduced a
1:56quant's thinking down to eight ideas,
1:58and I'll walk you through in order how a
2:00quant would reason with a real trade,
2:02with each idea building on the last. I
2:04actually got inspiration for this
2:05concept from Louis Jackson, who did a
2:07great video on the Markov method. And
2:09that method, if you've seen it,
2:10basically models which direction the
2:12market is moving in. And it's a
2:13reasonable way to think about regimes.
2:15But today, I'm going to present the
2:17other half of the picture, the
2:18calculation that actually determines the
2:20stakes behind real positioning, aka how
2:23much money actually goes on the table
2:25per trade. And on a real quant desk,
2:27that decision is treated with just as
2:29much rigor as the overall direction of
2:31the market, because ultimately that's
2:32what determines profitability and
2:35survival of a trader. So, before we get
2:36into anything else, let's actually
2:37define what the word quant really means,
2:40because it gets thrown around
2:41constantly, but I rarely see it
2:43explained. A quant [music] stands for
2:44quantitative analyst. These people often
2:46study advanced mathematics, physics, or
2:49computer science, and funds will recruit
2:51top talent often on some of the highest
2:53salaries in finance. Their job is
2:55essentially to replace opinion with a
2:57real measurement. So, discretionary
2:59traders, they'll build a view on the
3:01market. Quants will build a model to
3:04represent that view. They'll test it on
3:06history, they'll model it on different
3:07variables, and they come up with a real,
3:09tangible system based on real
3:11constraints. They don't usually work as
3:13lone geniuses, although there are a few.
3:15They usually split up into different
3:17departments which handle different
3:18segments of the market. One team will
3:20typically research signals, one on risk
3:22and sizing, because every single element
3:25that comprises a trade requires a
3:26different skill set and different
3:28parameters. Now, of course, I'm
3:29generalizing here, because every single
3:31fund operates differently, but at a
3:33large scale, this is typically how you
3:34see them operate. Now, if you're a
3:36trader, here's the fundamental
3:37difference between how you might see a
3:39chart and how a professional quant
3:41trader might see a chart. Most traders
3:43just see price, so they'll look at where
3:45it's been, they'll look at where it's
3:46going, and essentially the entire read
3:48on the market is directional. A quant
3:50will often analyze violence or
3:52volatility first. They'll look at how
3:54hard the market is moving independent of
3:57direction. So, a retail trader might
3:59think, "This feels dangerous." A quant
4:01would essentially hear an incomplete
4:03sentence. They'd be thinking, "Well,
4:04dangerous is a measurable quantity. So,
4:07what's the number?" So, instead of every
4:09morning looking at a price target like a
4:11retail trader, quants will look at
4:13volatility, value at risk, position
4:15sizing, Greeks, exposure, expected
4:17moves, and they'll apply stress test to
4:19market movement and essentially try and
4:21combine all of that into a single score.
4:23In today's video, I'm going to refer to
4:25that as the violence score, but I'm
4:27aware that it's a lot more nuanced than
4:28that in practice. And underneath that
4:31system is essentially a single mental
4:33model. And it's the most valuable idea
4:35that I want to present to you in this
4:37entire video. For every trade that you
4:38ever take, you'll essentially just ask
4:41two questions. Which way, so in which
4:43direction do I think price will move,
4:45and how much? So, how much am I willing
4:48to stake on this position? And the
4:49second question from the start of the
4:50video is something traders often get
4:53wrong, even though it's the most
4:55important in terms of profitability and
4:58survival in the market. And that's how
4:59you can have traders with a really high
5:01win rate, but they're actually not
5:02profitable. And how you can have traders
5:04with a lower win rate, but they are
5:06profitable over time because they
5:07understand when to size and how to size.
5:10Retail treats which way will price move
5:13as essentially the entire game. Position
5:15size is usually an afterthought. Whereas
5:17a quant will rank those two questions by
5:19how answerable they are. Which way is
5:22barely answerable. It's actually the
5:23weakest signal in trading. Whereas how
5:25much you actually stake is fully
5:27controllable. It's the one variable in
5:30trading that's actually at your command.
5:32So, often a quant will invert energy
5:34allocation, meaning they'll put minimum
5:35effort on a question that can't be
5:37answered reliably, and then they'll put
5:39maximum rigor on the variable that can
5:41be controlled completely. So, the
5:43industry, the entire trading industry,
5:45sells you one question, and it's a sexy
5:47sell. Price is going to move up, price
5:49is going to move down, but the reliable
5:51funds over time have actually compounded
5:53their wealth answering the boring
5:55question and compiling it into a
5:57framework that they can use. And today,
5:58the really powerful thing I'm going to
5:59teach you is how to put a framework into
6:02a quant skill, and I'm going to give it
6:04to you today at the end of the video
6:05that you can repeatedly use. This, in my
6:07opinion, is the future of trading. So,
6:09let's go back to what I was discussing
6:10before, the violence score. Because in
6:12order to develop a quant-like model, we
6:14need to discuss why violence can
6:16actually be predictable. It has a
6:17property that market direction doesn't
6:19have, and you can see it clearly on the
6:21chart. Market chaos typically doesn't
6:23arrive in isolation, but instead in a
6:25series of clusters. So, for example, we
6:27can look at March 2020 in the S&P, which
6:29was absolute carnage, and then we can
6:31compare that to the FTX collapse on a
6:34different asset in Bitcoin in 2022, and
6:36it paints a similar picture, but on a
6:38completely different asset. A quant
6:40doesn't settle for just eyeballing that
6:41pattern. They want to define an exact
6:43number. So, for example, I'll walk you
6:44through the thought process. On a
6:46typical day, there's roughly a 10%
6:47chance of a violent move. On a top
6:49decile day, I actually ran this based on
6:5115 years of back-tested Bitcoin data, on
6:54a day after a violent move, that
6:56probability triples to 30%. That's the
6:59predictability I'm talking about. It's
7:01not 100%, but it's more predictable, and
7:03it's a number you can actually trade
7:05off. Then I went and checked the monthly
7:07data on the S&P 500. And if you go back
7:09to 1871, violent moves are twice as
7:11likely to follow violent months. So,
7:14that's 150 years of a completely
7:16different asset class, but it follows a
7:17similar law. So, violence isn't the one
7:19metric a quant would look at, but it is
7:21a metric [music] which can be
7:22quantified, assigned a probability, and
7:25thus developing a framework which you
7:27can trade off. What I just referenced is
7:28something called volatility clustering.
7:30It's one of the most reliable patterns
7:32in financial data, and this is
7:33essentially the quant's worldview.
7:35Direction is close to random, but
7:36violence carries genuine information
7:38about what might happen tomorrow. If
7:40you're confused, don't worry, I'll make
7:42it really simple. The cleanest way to
7:43hold this in your head is actually
7:45seismology, or earthquakes. A major
7:47earthquake tells you nothing about the
7:50specific direction a singular building
7:52might fall, but it tells you with higher
7:54confidence that more shaking might be
7:56coming. That is exactly how a quant
7:58reads a violent trading day. Zero
7:59information about direction, but
8:01substantial information about tomorrow's
8:03volatility. Stated as a rule, direction
8:05is unpredictable, whereas magnitude
8:07typically is predictable. Okay, here's
8:08where we get into the fun stuff, the
8:10Nobel Prize equation, which is an
8:12essential part of the framework that
8:13we're building today. Because we just
8:15confirmed that violence is predictable.
8:16So, the question now becomes, how do you
8:18predict it? And this is exactly where
8:20the Nobel Prize enters. In 2003, the
8:23economist Robert Engle won a Nobel Prize
8:26for solving this exact problem. His
8:28original model was called ARCH. His
8:30student Tim Bollerslev extended it to a
8:32system that trading desks still use
8:35today, GARCH. The Nobel Prize was for
8:37the discovery, but the GARCH extension
8:39is the tool that is still used in
8:41quantitative trading. GARCH stands for
8:43generalized autoregressive conditional
8:46heteroscedasticity. All right, I know
8:48that's an absolute tongue twister, but
8:49essentially it's a model that's used to
8:51predict market volatility, rather than,
8:53as we've been discussing, predicting the
8:54exact direction of asset prices. A quant
8:56models tomorrow's violence as a sum of
8:58three components: the base level the
9:00market always carries, which is the
9:02asset's personality, plus a fraction of
9:04today's shock, the aftershock component,
9:06and a fraction of today's violence
9:07level, which is the memory component. As
9:09an example, I fitted this to Bitcoin's
9:11entire 15-year history, and the weights
9:13it estimated tell you the whole story.
9:15The shock carries 15% of the weight, and
9:17the memory carries 85%. In practical
9:20terms, volatility is almost as
9:22persistent as it's mathematically
9:24possible to be. And that persistence is
9:26why the forecast works, because you're
9:28forecasting a quantity that hardly
9:30changes overnight. Note, this model does
9:32not deliberately claim directionality of
9:34an asset. It forecasts how violent
9:36tomorrow may be, but not in which
9:38direction. Any model which claims to
9:40accurately predict directionality needs
9:43to be met with a grain of salt because
9:45this model earned a Nobel Prize because
9:47it deliberately restricts its claims to
9:49what the data can actually support. And
9:51the restraint itself is a lesson in how
9:52quants actually think. Okay, so let's
9:54say you're a quant and you now have this
9:57violence forecast. Here's how it becomes
9:59deterministic of an actual position
10:01size. The reasoning runs as follows. My
10:03risk on any trade equals my size
10:05multiplied by the market's violence. I
10:07can't control the violence, so instead I
10:10control the size of my position. So step
10:12one is setting an actual limit for your
10:13account. The actual annual swing that
10:15you're prepared to tolerate.
10:16Institutional desks typically run around
10:1815%. Step two is reading the current
10:21market's level. So let me actually show
10:22you. This is the NASDAQ as I'm
10:24recording, which is moving at roughly
10:2730% annualized if you look at the storm
10:30gauge and the market violence percentage
10:32on the storm gauge. This, which I'm
10:34going to explain very shortly, is
10:36custom-built based on the entire
10:38framework. And I'm going to give it to
10:39you at the end, but let me explain. This
10:4130% number here means the market is more
10:43violent than 96% of the past year, which
10:47you could consider a storm by any
10:49reasonable definition. And notably, it
10:51began whilst the index was at all-time
10:53highs. As the next step, step three, you
10:55want to divide this number, so 15 over
10:5730, which gives you 1/2, which is this
10:59number here, this sizing number, which
11:02essentially means if a trade signal
11:04fires today, it receives half of your
11:07normal size allocation. Let's say on a
11:08$10,000 baseline, this would be around
11:11$5,000. The best analogy is like cruise
11:14control. The road's gradient changes
11:16constantly, but you can control your
11:18speed to hold steady. So in this case,
11:20the market's violence changes daily, but
11:22your account risk holds steady. And note
11:25the conditional. If a signal actually
11:26fires, the reasoning never generates
11:28entries. Entries belong to a different
11:31department, which is where you can use
11:32things like the Markov method to
11:34determine which directional environment
11:36we're in. But, you might think that
11:38[music] a strategy like this merely
11:40smooths the ride. So, it makes highs
11:42less high, and it makes lows less low.
11:45But, that's actually not true. In
11:46reality, it can produce more money.
11:48[music] And the reason is an empirical
11:50fact about markets. The worst losses in
11:52history do not occur on randomly
11:54distributed days. They cluster inside
11:57the violent periods. 2008, March 2020,
12:00the major crypto collapses. In each
12:02case, volatility was already elevated
12:05before maximum damage arrived. This is
12:08the aftershock logic. The shaking was
12:10already underway before the building
12:12eventually came down. So, by reducing
12:14size in a storm, you're refusing to
12:16carry full exposure during a period
12:18where there is a higher likelihood of an
12:20extreme event which could threaten your
12:22account. All right, we're about to
12:23build, but one final thing. I want to
12:25discuss a final piece of mindset,
12:27because I believe any serious
12:28practitioner, especially before you're
12:30about to use a model like this, should
12:31explain what a model can't do. This will
12:33never call a top or a bottom. It
12:35contains, in fact, no directional
12:37information whatsoever. You probably
12:39have many indicators, many other
12:40methods, and even your own discretionary
12:42biases to come up with those conclusions
12:44yourself. If there's a crash that erupts
12:47from a full period of calm, then even if
12:49you use this, you may still catch it
12:50with full size. That's where you need
12:52stop losses and drawdown protection.
12:54There are still shock events which can't
12:56always be predicted. And you'll see in a
12:58minute why this isn't hypothetical, it's
13:00actually in the data. If you're a trader
13:02that specifically trades breakouts and
13:04you've developed an edge during these
13:05expansionary periods, then naive
13:07volatility sizing could actually work
13:09against you. If you guys like this
13:10video, I'll make it a series, and that
13:12could be a future component. So, I can't
13:14claim that this model eliminates risk or
13:16will maximize everyone's trading
13:18strategy. But, what I can say is it's a
13:20damn good place to start if you're
13:21interested in getting more into
13:23quantitative trading, and the risk
13:25frameworks that real quantitative
13:27traders will use. I believe you can
13:29expand upon what I'm about to give you
13:31to make it even better, and in fact,
13:33that's something I'm going to bring to
13:34the channel in the future. So, of
13:36course, make sure you subscribe, and
13:37we're going to continue diving into this
13:39stuff. As AI gets better, so does our
13:41capabilities to make better tools for
13:43ourselves. All right, here we go.
13:44Everything we just walked through, the
13:46complete chain of reasoning I've encoded
13:49into a Claude skill that you can
13:51download. So, what you want to do is you
13:52want to use the link in the description
13:54below, follow the steps, and you'll get
13:55access to a GitHub. This GitHub has the
13:59entire GARCH method. So, it has the
14:01installation path, which you put into
14:04Claude code. It explains what the skill
14:06actually does. It explains the
14:08onboarding prompt, and then it has a
14:10Pine Script bonus, which I'll show you
14:12how to install on TradingView in a
14:14minute for you to get the exact overlay
14:17that I showed you before. So, what you
14:19want to do to install it is go into the
14:21installation section, click copy, then
14:23you want to go into Claude code. All
14:25right, so you want to paste it in. If
14:27you're on the terminal, {forward slash}
14:28plugin should work. If you have any
14:30issues, so if you're on the app, you can
14:32replace the {forward slash} with the
14:33word Claude if you have any errors. So,
14:35that's the route that I'm going to take.
14:37Now, it's running the plugin commands.
14:38It's installing the plugin, and the
14:40commands have successfully installed, so
14:42you now have the skill on your Claude.
14:45The other option, and this is the way
14:46that you can see the skill being written
14:48line by line, is clicking on
14:50garchmethod.md,
14:52and you can copy and paste this entire
14:54file, and you can put it into Claude
14:56code. And now, it's going to fetch the
14:58skill files from the repo. You don't
14:59need to do both methods. I'm just
15:01showing you both ways that you can
15:02install it in case you actually want to
15:04see things line by line. You can see
15:05it's installed the strategy and verified
15:07the strategy. So, what do you do now?
15:09Well, now you can just talk to it. It's
15:10a skill that you can talk to to verify
15:13against the exact method. So, let me
15:14show you an example. You could say,
15:16"What's the volatility forecast on
15:18Bitcoin?" And it's going to run directly
15:20from the skill, which is trained on all
15:22of this in the GitHub repo. And you can
15:24see that it's given me the Bitcoin
15:26volume forecast, 40.9% annualized. It's
15:29given me a calm regime. It's got the one
15:31day ahead forecast. The good thing about
15:33this is you can actually just query it.
15:34So, you could say, "Could you explain in
15:36simple terms what this means and how
15:38this should impact my Bitcoin position
15:40sizing?" So, you can actually speak to a
15:43skill that is trained on a real quant
15:44method. And you can actually just build
15:46on this over time. So, if you have
15:48refinements you want to make or
15:49improvements or if new models come out
15:51and we can refine it together even
15:53further, you can just train the skill so
15:55it gets better over time. And right now
15:57I'm creating skills for a bunch of
15:58things. I'm creating skills for
15:59different methods. I'm creating skills
16:01to help me size positions. I'm creating
16:03skills to help me execute positions,
16:05just like a real quant firm would do.
16:07Taking skills for varying parts of
16:09comprising a trade. That is really how
16:11you can get the most out of AI to be a
16:13real live trading assistant. So, it came
16:16up with a plain English version. Think
16:17about it like a weather forecast. The
16:19model says a typical day tomorrow
16:20Bitcoin moves around 2.1% which
16:22annualizes to 41%. Over the past year
16:25Bitcoin's rougher than this about 2/3 of
16:26the time. So, right now it's on the calm
16:28end. That's the calm regime. This is why
16:30it matters for sizing. Or let's say
16:31you're about to enter a position on the
16:33Nasdaq. You could say, "Help me size my
16:35Nasdaq position." And it can actually
16:37walk you through how to think of the
16:38trade like a quant. So, you can see now
16:40it's running the GARCH forecast on the
16:42Nasdaq. So, it said for the Nasdaq the
16:44model says to cut your position down and
16:47not grow it. And you can actually go
16:48back and forth and have a discussion
16:50with the model based on positions that
16:52you might want to size into. So, if you
16:54have a directional read like we
16:55discussed before, you can use this for
16:57the volatility or the risk component of
17:00sizing the trade. But let's go a step
17:01further. Let's actually test if this can
17:04make money. So, I'll test it with a very
17:06popular strategy, the same strategy,
17:07which is an EMA cross on Bitcoin, which
17:10is trained on 15 years of data, once at
17:12a fixed size and once the quant way,
17:15using the skill. So, I just asked it to
17:17download Bitcoin price history in case
17:18it didn't have access, Right run the
17:20sizing comparison on it, and then make
17:21sure the target risk matches the EMA
17:23strategy to make sure the comparison is
17:25apples to apples. All right, so you can
17:28see the growth of $1 from 2012 with the
17:32fixed size strategy on Bitcoin, which is
17:35the gray line, and then the volatility
17:37target GARCH size line, which you can
17:40see actually ends up marginally beating
17:42the fixed strategy on Bitcoin. At the
17:45same time, the volatility targeted
17:46strategy from GARCH actually took less
17:49drawdowns than Bitcoin in the meantime
17:50as well. So, whenever you run a model
17:52like this, you can actually get
17:54volatility targeted graphs generated
17:57like this, which I think are really
17:58cool, so you can map out potential
18:00trades, which I think is valuable, and
18:02it's what a quant firm would do to
18:03visually map out a trade before they
18:05enter. And obviously, Claude does this
18:07for you if you download the skill. I
18:08think the crazy part is actually the
18:10Sharpe ratio. This is essentially the
18:12risk-adjusted return of an investment,
18:14because investments aren't just about,
18:16you know, whether you made money, it's
18:18also about how much risk you took on to
18:20make that amount of money, and you can
18:22see that the volume targeted approach
18:24had a significantly better Sharpe ratio,
18:26meaning it had better risk-adjusted
18:28returns. So, even if you make the same
18:29amount of money, if you can do so taking
18:31on less risk, then that is the better
18:33approach, and that was reflected in the
18:34drawdown charts. It simply took on less
18:37drawdown. So, the max drawdown with the
18:38volatility targeted approach that used
18:40our strategy was 63%. The max drawdown
18:43with the fixed size strategy was 81%.
18:45The worst month with our approach was
18:4719%. The worst month normally was 40%.
18:50The final equity on our approach was
18:5221,000, and the final equity on the
18:54fixed size approach was 17.9 thousand.
18:56Now, I'm going to ask it, "Run the same
18:58comparison on my Nasdaq position over
19:01the last 50 years." Now, for Nasdaq,
19:03this is actually extremely interesting.
19:05It improved the Sharpe ratio. It
19:07improved every single risk number, so
19:09you took on a lot less risk, but it did
19:10cost a full point of compounded annual
19:12growth rate. So, over 50 years this
19:14would have made you 99 less dollars. So,
19:17Bitcoin won risk and absolute return.
19:19Equities didn't. Equities [music]
19:21won the risk component, which may have
19:22enabled you to size up more, which may
19:24have enabled you to, you know, end up
19:26out-returning, but that is a variable
19:28test we need to run on new numbers. But
19:29nonetheless,
19:30it did cost in this experiment less
19:33returns. Because the size of throttle
19:35mostly trend exposure during violent
19:37rallies off the lows. However, it did
19:39earn its keep in some capacity because
19:41the goal of a fund is not to blow up.
19:43So, in the dot-com bust, where fixed
19:45sizing hit its yearly worst at minus
19:4728.8%
19:49volume targeting kept it to roughly
19:51minus 13%. Same story in 2008, 28%
19:54versus 13% and 2022, 23% versus 13%. And
19:59its worst month was cut from 10.9% to
20:02only 7.7%. So, the takeaway with the
20:04Nasdaq is that volume targeting is
20:06buying insurance, not extra return. So,
20:09it helps the pain, but it costs you 1%
20:12compounded annual growth rate. And
20:14whether that trade is worth it, as Fable
20:15points out, depends on whether you'd
20:17actually sit through a minus 29% without
20:19capitulating. Most people wouldn't and
20:21capitulation costs more than the
20:22insurance premium. Very interesting
20:24thought process. Once again, this skill
20:26is not meant to be the only thing you
20:28use to trade. This is a tool for you to
20:30use to understand risk and to help
20:32understand what environment you're in
20:34and how to target. I do think a lot of
20:36people though, they struggle with this
20:38so much that if they had a tool like
20:40this, they would lose significantly less
20:42amounts of money cuz they'd be able to
20:44think in a different capacity when
20:45opening trades. Even if it is for
20:47riskier asset types, it can help a lot.
20:49And it even helps when charting. So, if
20:50you go into the GitHub and you go into
20:53Pine Script and go Storm Gauge Pine,
20:55what you can do is you can copy the
20:58entire file here. You can just either
21:00copy the raw file in the top right-hand
21:02corner or simply copy and paste. Then go
21:05into TradingView. Then click on the
21:07little pyramid logo here, then I've
21:09already got one in, but I'll clear it
21:10out and show you what it would look
21:11like. Paste it in, click enter, click
21:13refresh, and this is going to load the
21:16indicator on your screen, which is this
21:18indicator here, which goes through the
21:20metrics we discussed before. So, the
21:21market violence metric, the current
21:23percentage versus the limit, the
21:24category [music] it's in versus the last
21:26year, so we could see it's heavily in
21:27storm territory, 96% more volatile
21:31versus the past year. It has a sizing
21:33suggestion and an example on $10,000 and
21:36a suggested action based on the number
21:38above. And then you can also see the
21:40volatility bands in here as well to help
21:42you understand the movements of the
21:44market. So, this is a very helpful tool
21:46that you can use, and of course you can
21:47build upon it. The best thing about this
21:48being in cloud is that you can make
21:50tweaks and edit things over time and
21:52create your own parameters. So, if you
21:53want all of this, use the link in the
21:54description below. If you enjoyed this
21:56video, make sure to subscribe because
21:58I'm going to be doing a lot more AI
21:59trading content like this, and we can
22:01even build skills together for other
22:03areas of trading and potentially other
22:05quant strategies because there are
22:06multiple approaches here, but I wanted
22:08to touch on one of the most relevant
22:09ones today. And yeah, thank you for your
22:11time today. I'll see you in the next
22:13video. Have a lovely rest of your day.
22:14Peace out.

About this transcript

This page contains the full transcript of I Re-Created A Quant Trading Strategy With Claude Code (Nobel Prize Method) by Miles Deutscher Finance, generated from the public captions YouTube serves with the video. The transcript has 4,633 words across 689 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.