Latest / Deep Values: Value Investing Deep Dives / AI Algorithms Hack Traditional Value Investing
Transcript
- 0:00You know that classic image of Warren Buffett?
- 0:03Oh, absolutely. The Oracle of Omaha.
- 0:05Yeah, exactly. He's like sitting in this quiet,
- 0:08dimly lit office out in Omaha, far away from all the noise of
- 0:12Wall Street. Right, just surrounded by these
- 0:15towering stacks of paper annual reports.
- 0:18Yeah, reading through corporate filings for, well, for hours on
- 0:21end. And he uses decades of sheer
- 0:23intuition and pattern recognition to just find a
- 0:27hidden gem in the stock market. It's incredible it.
- 0:30Is we treat value investing almost like this, this romantic
- 0:34art form, You know it's like the ultimate human endeavor, right?
- 0:37Because it relies entirely on a financial 6th sense that takes a
- 0:41lifetime to cultivate. Totally.
- 0:43I mean, the traditional philosophy going all the way
- 0:45back to Benjamin Graham and David Dodd in the 1930s is
- 0:48really about finding an imperfect relationship between a
- 0:51company's historical financial metrics and its future stock
- 0:54price. You want to buy a dollar for
- 0:56$0.50? Exactly, you are always looking
- 0:59for a margin of safety. So, OK, imagine taking that
- 1:02deeply human, intuitive art form, ripping it out of the
- 1:05hands of legendary stock pickers, and just feeding it
- 1:09into a machine learning algorithm.
- 1:11Which sounds kind of. Crazy it does, but that is
- 1:14exactly what we are diving into today.
- 1:16Data scientists are essentially hacking traditional value
- 1:20investing. They really are.
- 1:21So today, for you listening, our mission is to explore how modern
- 1:26algorithms are quantifying intuition, and why the financial
- 1:31numbers you rely on every day to evaluate a company might be
- 1:35completely lying to you. We are really compelling
- 1:37collision of old school philosophy and cutting edge data
- 1:41science to look at today. Yeah, we've got a great stack of
- 1:43research. We do.
- 1:44We are drawing from an academic paper by RJ Sanic on predictive
- 1:49modeling in systematic value investing.
- 1:51Which is fascinating. It is.
- 1:53We're also looking at a specialized real estate journal
- 1:55that analyzes how financial leverage artificially
- 1:58manipulates profitability metrics, right.
- 2:00And then to tie it all together, we have a forensic accounting
- 2:02brief from new constructs. OK, let's unpack this, because
- 2:06the first major hurdle in classic value investing is that
- 2:10it is just infuriatingly subjective.
- 2:13Extremely subjective. Right.
- 2:15Like if you read a book like Buffettology, which is Mary
- 2:18Buffett's breakdown of Warren Buffett's methods, you are hit
- 2:22with all these qualitative metrics.
- 2:23A soft metrics, yeah. Buffett looks for a company with
- 2:26an an identifiable consumer monopoly or a business that
- 2:31demonstrates strong earnings. Well, strong is not a
- 2:35mathematical formula. Exactly.
- 2:36And monopoly is not a standard accounting metric.
- 2:39You could just, you know, scrape from a database.
- 2:42That is the core problem, and it brings us right to RJ Sack's
- 2:45research, because Sack actually takes Buffett and Clark's highly
- 2:48subjective value investing metrics and defines them in
- 2:51purely quantitative terms. It translates them.
- 2:54Yeah, but I struggle with this premise a bit.
- 2:57Oh really? Why is that?
- 2:58Well, I mean, can an algorithm truly capture something as
- 3:01abstract as a consumer monopoly or franchise power?
- 3:05It's a fair question. It feels like trying to
- 3:07translate a beautiful piece of poetry into cold algebra.
- 3:10You know? Like are we just stripping the
- 3:12soul out of the strategy and pretending we still have the
- 3:15same thing? I get that you might lose some
- 3:17of the poetry, sure, but Sack argues the trade off is
- 3:21absolutely worth it. OK, how so?
- 3:23Let's look at how he actually does this.
- 3:25He doesn't just assign a random number to a feeling, right?
- 3:28He translates a concept like consistently upward earnings
- 3:32trends into specific geometric mean calculations over like an 8
- 3:37year period. Oh wow.
- 3:38OK. Yeah, he takes the abstract idea
- 3:40of a company having a durable competitive advantage and turns
- 3:44it into a strict threshold of historical gross margin
- 3:47stability. So he is forcing the algorithm
- 3:49to mathematically prove the things Buffett just intuitively
- 3:52senses by looking at a balance sheet.
- 3:53Precisely by quantifying these metrics, AC enables them to be
- 3:57used in systematic algorithmic stock icking.
- 4:01Which changes the game entirely. It really does, because when you
- 4:03do that, you remove human bias. You completely eliminate the
- 4:07emotional error of falling in love with a stock story.
- 4:10Or panicking during a market downturn, which we all do.
- 4:14Exactly. You turn the legendary Oracle
- 4:16into a cold, repeatable process. I mean, I understand sacrificing
- 4:20the poetry for discipline, but people have tried systematic
- 4:24value investing before. They have, yes.
- 4:26Like it isn't a new concept to just write a list of
- 4:29mathematical rules and let a computer pick the stocks.
- 4:32They have tried it and traditional systematic value
- 4:34investing has a massive flaw. OK, what is it?
- 4:38Well, look at Wesley Gray and Tobias Carlisle's well known
- 4:42quantitative value strategy. It's a brilliant strategy, but
- 4:46it uses static variable coefficients.
- 4:48Meaning the rules are completely rigid and they never change.
- 4:52Right, A static model assumes that the relationship between a
- 4:55specific financial metrics, say like the price to earnings ratio
- 4:59and the stocks future market outcome never alters.
- 5:03It assumes the rule works the exact same way across every
- 5:05single business cycle. Which is wildly unrealistic if
- 5:08you have spent like any time at all watching the market because
- 5:12the metrics that predicted a winning stockduringthe.com boom
- 5:15of the late 90s would get you absolutely crushed in the 2008
- 5:19financial crisis. The economy is a living,
- 5:22breathing ecosystem. You know it changes.
- 5:26And that rigidness is the exact problem SAC introduces
- 5:30predictive modelling to solve. OK, instead of static rules,
- 5:34predictive modelling uses machine learning algorithms to
- 5:37dynamically update these relationships every single time
- 5:41new financial statement data becomes available.
- 5:43As wild. And out of the several models
- 5:45SAC tests, the one that showed the most predictive power was a
- 5:48random forest algorithm. OK, so for anyone who isn't a
- 5:52data scientist, a random forest algorithm sounds incredibly
- 5:55intimidating. It does sound like a sci-fi
- 5:56concept, but think of it this way.
- 5:59A single decision tree is just a flow chart.
- 6:01Like a yes or no thing. Right.
- 6:02If earnings are up, go left. If debt is high, go right.
- 6:05But a single tree can be easily fooled by weird, noisy data.
- 6:09So what does the forest do? A random forest algorithm builds
- 6:13thousands of these distinct decision trees, gives them
- 6:16slightly different slices of the data, and then merges them all
- 6:19together to take a collective vote.
- 6:21Oh, I see. Yeah, it provides a much more
- 6:23accurate and stable prediction because it just averages out the
- 6:26individual errors. That makes a lot of sense.
- 6:29So Sacks set this random forest to a 50% cut off threshold,
- 6:33meaning it had to be highly confident to make a pick.
- 6:36And he didn't just feed it standard financial data.
- 6:40He combined the classic quantitative value variables
- 6:43with the quantified buffetology variables we just discussed.
- 6:46He fed the machine both the standard math and the translated
- 6:49poetry. I love that and the results of
- 6:51his back test are really what make this research stand out.
- 6:54They. Are phenomenal.
- 6:55He ran the model over a 26 year period from 1990 to 20, 2016.
- 7:00He trained specific variations, Model 11 and Model 12, to
- 7:04predict whether a stock would generate alpha over a five year
- 7:07and 10 year horizon. And they drastically
- 7:10outperformed traditional systematic strategies.
- 7:12But I mean in the finance world anyone can over optimize a back
- 7:16test to make it look like they beat the S&P 500.
- 7:18Oh sure, which is exactly why we have to look at the academic
- 7:21benchmarks, right? The models generated
- 7:24statistically significant returns above both the Fama
- 7:28fringe 3 factor and Carhartt 4 factor models.
- 7:31OK, let's break those down, because throwing around Foma
- 7:34French is a great way to lose someone's attention.
- 7:37Fair enough. Why do those benchmarks actually
- 7:40matter? Well, in finance we have to
- 7:42prove an algorithm isn't just getting lucky, right?
- 7:45Like if the overall market goes up 20% in a year, a monkey
- 7:49throwing darts at a newspaper could make 15%.
- 7:51Exactly, everyone looks like a genius in a bull market.
- 7:54Right. So Foma, French, and Carhart are
- 7:57models that strip away obvious market trends.
- 8:00They adjust for things like general market risk, the
- 8:02tendency for small cap stocks to outperform large cap stocks, the
- 8:06tendency for cheap value stocks to outperform expensive growth
- 8:10stocks. No, I get it.
- 8:11It's simple price momentum, right?
- 8:13Exactly. So if your algorithm beats those
- 8:15models, it means you aren't just riding a general market waste.
- 8:19You're actually generating unique standalone value, yes.
- 8:23The machine is genuinely finding hidden gems.
- 8:26That is incredible. It is, but what's fascinating
- 8:30here is SAC ran into a highly nuanced statistical hurdle while
- 8:35proving this. And it is a fascinating problem
- 8:38that plagues quantitative finance.
- 8:40What's the problem? Well, when you run massive
- 8:43financial back test with thousands of variables across
- 8:46decades of data, you run into the problem of false positives.
- 8:50Wait, if it's scanning decades of historical data, isn't it
- 8:53basically cheating? Sort of, yeah.
- 8:55Like of course it is going to find a pattern looking backward.
- 8:58How do we know the pattern is real and not just a coincidence?
- 9:00That is the exact danger of data mining.
- 9:03Imagine you test 10,000 completely random variables to
- 9:06see if they predict the S&P 500. OK.
- 9:09You might test whether the amount of butter produced in
- 9:11Bangladesh correlates with the tech sector in California.
- 9:13Just totally unrelated things. Right.
- 9:16And just by sheer mathematical luck, you will find a few
- 9:19patterns that look incredibly predictive on paper, but they
- 9:23are completely fake. So how do you fix that?
- 9:26In statistics, you normally use something called the Benjamin E
- 9:29Hochberg procedure to control that false discovery rate.
- 9:32OK, so it acts as a like a mathematical penalty?
- 9:35Exactly. It punishes the model for taking
- 9:38too many wild guesses, ensuring the patterns it finds are
- 9:41actually rooted in reality, not just random noise.
- 9:44You've got it, and Sack actually proposed a brand new statistic
- 9:48in this paper termed the SAC ratio.
- 9:51Oh, he named it after himself. Class.
- 9:54Naturally, but it extends that Benjamin E Hochberg procedure
- 9:58specifically to help inferentially identify false
- 10:01positive observations in these massive financial back tests.
- 10:04So it acts as a built in lie detector for the algorithm just
- 10:08to ensure it isn't just finding shapes in the clouds.
- 10:10That's a perfect way to put. It but OK, here's where it gets
- 10:12really interesting. We have built this perfect
- 10:15emotionless lie detecting stock picking we have.
- 10:19But here is the fatal flaw in this entire setup.
- 10:22A predictive model, no matter how advanced, is only as good as
- 10:26the raw ingredients you feed into it.
- 10:29The data, right? What happens to our brilliant
- 10:32random forest algorithm if the standard accounting numbers it
- 10:35relies on are actually manipulating the truth?
- 10:38Well that is the Achilles heel of quantitative strategy.
- 10:41Yeah, if the underlying corporate data is legally
- 10:45manipulated or misleading, the models predictions will be
- 10:49dangerously wrong. So let's look at how a company
- 10:51can legally manipulate its profitability on paper, which
- 10:54brings in the study from the Indonesia Auditing Research
- 10:57Journal. Right, the real estate study.
- 10:58Yeah, they looked at the property and real estate sector.
- 11:02It is a highly cyclical, extremely capital intensive
- 11:05industry. Very much so.
- 11:06And they focused on the relationship between financial
- 11:09leverage, which is taking on debt and return on equity or
- 11:12Roe. Which is a huge metric for
- 11:14investors. The research shows a robust
- 11:17positive correlation. When real estate companies take
- 11:19on more debt, their Roe shoots up.
- 11:22To understand the mechanics of how that trick works, we really
- 11:25have to look at the formula for Roe.
- 11:27OK, let's do it. Return on equity is calculated
- 11:30by taking net income and dividing it by shareholders
- 11:33equity. Let's walk through a real world
- 11:36scenario so we don't get lost in the fractions.
- 11:38Let's say you and I start a real estate company.
- 11:40Together, we put in $1 million of our own money.
- 11:43OK. So that is our equity.
- 11:45Right. We buy a building, we rent it
- 11:47out and at the end of the year our net income is $100,000.
- 11:52So our Roe is 10%. Yeah, a 10% return is solid, but
- 11:56let's say you want to aggressively expand.
- 11:58As one does, right? Instead of finding more
- 12:01investors and issuing more stock, which would increase that
- 12:04equity denominator, we go to the bank.
- 12:06We take out a loan. Exactly.
- 12:08We borrow $9 million, we buy 10 more buildings and assuming the
- 12:12market is good and the properties perform well, our net
- 12:15income might jump to $500,000 after paying the interest on the
- 12:19loan. Wow.
- 12:20OK. But our equity, the money you
- 12:23and I actually put in is still just $1 million.
- 12:26Yes. So now you divide that $500,000
- 12:29net income by our 1,000,000 in equity?
- 12:32And our Roe just skyrocketed to 50.
- 12:35Percent exactly. We didn't necessarily become
- 12:37better operators. We just use the bank's money to
- 12:40artificially pump up the return on our own money.
- 12:43That is wild and if you are someone who just looks up a
- 12:47stock on a free finance website, sorts by highest Roe and hits by
- 12:52this research is a massive red flag.
- 12:54Oh, a huge red flag. You might be looking at a
- 12:57company that is just drowning in debt.
- 12:59The study explicitly points out that while leverage amplifies
- 13:03gains on paper, it exposes the company to extreme downside
- 13:07risks during market fluctuations.
- 13:09Right, because debt doesn't care about a recession.
- 13:11Exactly. If property values drop or
- 13:14interest rates rise, that massive debt load still has to
- 13:17be serviced. The equity can be wiped out
- 13:19rapidly. The study also contrasts this
- 13:21with earnings per share, or EPS, right.
- 13:24They found the impact of leverage on EPS is far more
- 13:26multifaceted and nuanced. EPS is heavily influenced by
- 13:30operational efficiency and market cycles, not just, you
- 13:32know, capital structure trickery.
- 13:34So if a machine learning algorithm is blindly scanning
- 13:37the market for a high Roe, it might just be loading up on
- 13:40highly indebted real estate companies right before a massive
- 13:44housing crash. Completely convinced it is
- 13:46buying great value stocks. If we connect this to the bigger
- 13:50picture, the real estate debt problem perfectly illustrates
- 13:54the core warning from our final source.
- 13:56The new constructs brief. Right, The forensic counting
- 13:59brief from new constructs flat out says that because leverage
- 14:02artificially inflates Roe, it is an incredibly dangerous metric
- 14:06for both human investors and machine learning models to rely
- 14:09on. It's just bad data.
- 14:11Tandard counting data is fundamentally flawed if you were
- 14:13trying to find true intrinsic value.
- 14:16And the new contracts research argues that the alternative,
- 14:19like the true driver of a company's valuation is return on
- 14:22invested capital or ROIC. So how does ROIC actually solve
- 14:28the debt manipulation problem we just talked about?
- 14:30Well, ROIC looks at the return a company makes on all the capital
- 14:33invested in the business, both debt and equity.
- 14:35So it removes the illusion entirely.
- 14:37It does go back to your real estate example.
- 14:40ROIC would look at the full $10 million deployed.
- 14:43The 1,000,000 in equity plus the 9 million in debt.
- 14:46Exactly. It measures how effectively
- 14:48management is allocating the total pool of capital,
- 14:51completely neutralizing the accounting trick of keeping the
- 14:55equity denominator small. That sounds like the perfect
- 14:57metric, but I'm guessing there's a catch.
- 15:00There is always a catch. You cannot just look up a
- 15:02pristine unmanipulated ROIC number on a standard stock
- 15:07Screener, can you? No, you cannot.
- 15:10Calculating a clean ROIC requires true economic earnings.
- 15:14Yeah, OK. And to get those, you have to
- 15:16dig incredibly deep into the financial footnotes and the
- 15:19MDNA. The management's discussion and
- 15:21analysis section exactly which is where companies bury the
- 15:24inconvenient truths. Oh absolutely.
- 15:26They hide off balance sheet debt.
- 15:27They obscure operating leases. They bury unusual write offs.
- 15:31And they mask changes in their accounting assumptions deep in
- 15:34like page 140 of a dense legal text.
- 15:37Right. And no human being, not even a
- 15:40massive team of Wall Street analysts, can efficiently read
- 15:44and decode the footnotes for thousands of publicly traded
- 15:47stocks every single quarter. It is physically impossible to
- 15:50keep up with the sheer volume of text.
- 15:52It really is, which is where the technology circles back to save
- 15:56the strategy. The New Constructs paper
- 15:59explains that to feed accurate, unmanipulated data into the
- 16:03kinds of predictive models RJ SAC built, you require robo
- 16:07analysts. Robo analysts, OK.
- 16:10These are systems utilizing natural language processing to
- 16:12instantly read, parse, and translate the dense text to
- 16:15financial footnotes into clean economic data.
- 16:18So the AI is reading the footnotes for us.
- 16:20Exactly. Natural language processing
- 16:22allows the machine to read context.
- 16:24It isn't just looking for keywords, it actually
- 16:26understands. When a company is trying to hide
- 16:28A liability at a convoluted sentence that.
- 16:30Is amazing it. Extracts the true economic
- 16:33reality of the company. Calculates the clean ROIC and
- 16:36then feeds that pristine data into the dynamic random forest
- 16:40algorithm to pick the stocks. Wow.
- 16:43It is literally a complete end to end evolution of the value
- 16:47investing process. It is.
- 16:48The machine does the reading, finds the truth hidden in the
- 16:52fine print, calculates the real metrics, and dynamically updates
- 16:56its investing rules based on the current economic cycle.
- 16:58It's beautiful in a very nerdy way.
- 17:00It really is now for you listening if you were trying to
- 17:03figure out how to actually track all this to access more advanced
- 17:07researches, applications and our latest published personal
- 17:10intelligent multi agent Deep value intelligence.
- 17:13Welcome to explore and join Deep Values AI community and attract
- 17:18total market temperature. Welcome to Search and download
- 17:20iOS App Buffett Indicator from Apple App Store.
- 17:23And I do have to mention, just as a standard requirement
- 17:26disclaimer, all analysis is based on publicly available
- 17:29sources and may be subject to revisions or differing
- 17:32interpretations. All contents from this channel
- 17:34is for research and entertainment purposes only and
- 17:37does not constitute any investment or financial advice.
- 17:40Conduct your own due diligence before making investment
- 17:42financial decisions. Don't forget to like this video.
- 17:45Much appreciate it. Thanks for that.
- 17:48So what does this all mean? We started today with the image
- 17:51of the romantic, intuitive value investor sitting in a quiet
- 17:55office. The.
- 17:56Classic Buffett image. Right.
- 17:58But we have seen that value investing has dramatically
- 18:00evolved. It moved from subjective human
- 18:04intuition looking for a consumer monopoly to static mathematical
- 18:08rules that couldn't adapt to changing markets.
- 18:11And now it has evolved into dynamic machine learning
- 18:14ecosystems. Exactly.
- 18:15Ecosystems that require pristine robo analyzed data prioritizing
- 18:21true metrics like ROIC over easily manipulated numbers like
- 18:24Roe. The translation of poetry into
- 18:27algebra isn't the loss of the art form you know.
- 18:30It is the ultimate refinement of it.
- 18:32I think you're right. But this evolution raises a
- 18:34critical issue for you, the listener, to consider long after
- 18:38we wrap up today. Oh, what's that?
- 18:40Well, Benjamin Graham's original concept of the margin of safety
- 18:43relied on the fact that other investors in the market were
- 18:45making emotional, irrational mistakes.
- 18:48Right. It relied on the fact that other
- 18:50people simply hadn't read the footnotes.
- 18:52Exactly human error left value on the table for the smart
- 18:56disciplined investor to find. So if predictive algorithms can
- 18:59successfully decode Warren Buffett's intuition, dynamically
- 19:03adapt to market cycles without emotion, and instantly use
- 19:06natural language processing to parse every footnote to avoid
- 19:10accounting traps. What happens to the stock
- 19:13market's margin of safety when every investor has access to
- 19:16this exact same artificial intelligence?
- 19:19Oh wow, does the advantage disappear entirely?
- 19:21Or does the market simply evolve into a new Uncharted game?
- 19:25When the X-ray machine is in everyone's office, the hidden
- 19:28fractures aren't hidden anymore. Precisely.
- 19:30The market might just become a battle of whose AI can read the
- 19:33footnotes A millisecond faster. It's a brave new world.
- 19:36That is a fascinating thought to leave on.
- 19:38Thanks for joining us on this deep dive, and we'll see you
- 19:40next time.