delost's banner

delost

@thedelost • 2,660 subscribers

cooking at finance & ai

Shorts

the model in that clip has no good signal in it. it still put up +17% against the index's +5% the formula is doing the work R(t) = (Rmax / 7) · Σ s_i(t) seven separate signals, each scored, averaged into one number that's the entire model. no genius indicator anywhere in it and that's the part retail keeps missing retail hunts for the one signal that works a desk assumes every individual signal is weak and builds around that assumption here's why that assumption wins take N signals, each with sharpe s, and average them if they're uncorrelated, the combined sharpe is: s · √N seven weak signals at sharpe 0.3 each 0.3 × √7 = 0.79 nothing in that stack survives a backtest alone. together they clear the bar the noise in each signal is independent, so averaging cancels it the edge in each points the same way, so averaging keeps it that asymmetry is the whole mechanism but there's a catch, and it's the one that kills retail attempts correlation. the real formula is: s · √( N / (1 + (N−1)ρ) ) at ρ = 0.5 those same seven signals give: 0.3 × √(7 / 4) = 0.40 half the benefit, gone seven versions of momentum with different lookbacks aren't seven signals. they're one signal, repeated so the search isn't for better signals it's for signals that are wrong at different times grinold formalized this in 1989. the fundamental law of active management: IR = IC × √breadth skill per bet times the square root of how many independent bets you take you can be barely right, as long as you're barely right about many uncorrelated things renaissance doesn't run one model. it runs thousands of weak ones that's not a compromise. that's the design retail asks "is this signal good enough to trade" a desk asks "what does this add that i don't already have" the math is public. grinold's paper, every portfolio theory textbook the correlation matrix that tells you whether your signals are actually distinct is three lines of python they weren't finding better signals they were finding signals that disagree full breakdown in the article below

the model in that clip has no good signal in it. it still put up +17% against the index's +5% the formula is doing the work R(t) = (Rmax / 7) · Σ s_i(t) seven separate signals, each scored, averaged into one number that's the entire model. no genius indicator anywhere in it and that's the part retail keeps missing retail hunts for the one signal that works a desk assumes every individual signal is weak and builds around that assumption here's why that assumption wins take N signals, each with sharpe s, and average them if they're uncorrelated, the combined sharpe is: s · √N seven weak signals at sharpe 0.3 each 0.3 × √7 = 0.79 nothing in that stack survives a backtest alone. together they clear the bar the noise in each signal is independent, so averaging cancels it the edge in each points the same way, so averaging keeps it that asymmetry is the whole mechanism but there's a catch, and it's the one that kills retail attempts correlation. the real formula is: s · √( N / (1 + (N−1)ρ) ) at ρ = 0.5 those same seven signals give: 0.3 × √(7 / 4) = 0.40 half the benefit, gone seven versions of momentum with different lookbacks aren't seven signals. they're one signal, repeated so the search isn't for better signals it's for signals that are wrong at different times grinold formalized this in 1989. the fundamental law of active management: IR = IC × √breadth skill per bet times the square root of how many independent bets you take you can be barely right, as long as you're barely right about many uncorrelated things renaissance doesn't run one model. it runs thousands of weak ones that's not a compromise. that's the design retail asks "is this signal good enough to trade" a desk asks "what does this add that i don't already have" the math is public. grinold's paper, every portfolio theory textbook the correlation matrix that tells you whether your signals are actually distinct is three lines of python they weren't finding better signals they were finding signals that disagree full breakdown in the article below

27,591 次观看

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

a 25-delta put trades at 22% implied vol. the call at the same distance trades at 15% same index, same expiry, same distance from the money. 7 points apart black-scholes says that gap shouldn't exist the model assumes one volatility number for every strike σ constant across strikes and expiries. one distribution, lognormal returns, flat surface plot the real chain and you don't get a flat line you get a surface. tilted, curved, repricing every second that tilt has a name: skew and it's the closest thing markets have to a live fear gauge the reason it exists is structural, not a pricing error crashes are faster and deeper than rallies. returns have fat left tails so protection below the market costs more than the model says, because the model's normal distribution never priced the tail correctly this wasn't always true before october 1987 the surface was roughly flat. the crash rewrote it permanently one event taught the entire options market that the left tail is real, and the skew has never gone away since the measurement is simple: skew = IV(25-delta put) − IV(25-delta call) on the example above that's 22 − 15 = 7 points when that spread widens, demand for downside protection is rising. someone is paying up to hedge when it flattens, the bid for protection is fading same index level, same price on your chart, two completely different states of institutional fear the chart shows where price is. the surface shows what people are paying to be wrong and the shape carries more than direction steepness tells you how much tail risk is priced term structure tells you whether the fear is about this week or this quarter curvature tells you how much the market disagrees with its own base case retail sees one implied vol number on the option they're about to buy a desk sees a surface and trades the difference between its shape and what that shape usually looks like every input is public. the chain lists IV at every strike and expiry strike on one axis, expiry on another, IV on the third. the surface builds itself black-scholes won a nobel for a formula the market has been visibly disagreeing with since 1987 the disagreement is the signal full breakdown in the article below

a 25-delta put trades at 22% implied vol. the call at the same distance trades at 15% same index, same expiry, same distance from the money. 7 points apart black-scholes says that gap shouldn't exist the model assumes one volatility number for every strike σ constant across strikes and expiries. one distribution, lognormal returns, flat surface plot the real chain and you don't get a flat line you get a surface. tilted, curved, repricing every second that tilt has a name: skew and it's the closest thing markets have to a live fear gauge the reason it exists is structural, not a pricing error crashes are faster and deeper than rallies. returns have fat left tails so protection below the market costs more than the model says, because the model's normal distribution never priced the tail correctly this wasn't always true before october 1987 the surface was roughly flat. the crash rewrote it permanently one event taught the entire options market that the left tail is real, and the skew has never gone away since the measurement is simple: skew = IV(25-delta put) − IV(25-delta call) on the example above that's 22 − 15 = 7 points when that spread widens, demand for downside protection is rising. someone is paying up to hedge when it flattens, the bid for protection is fading same index level, same price on your chart, two completely different states of institutional fear the chart shows where price is. the surface shows what people are paying to be wrong and the shape carries more than direction steepness tells you how much tail risk is priced term structure tells you whether the fear is about this week or this quarter curvature tells you how much the market disagrees with its own base case retail sees one implied vol number on the option they're about to buy a desk sees a surface and trades the difference between its shape and what that shape usually looks like every input is public. the chain lists IV at every strike and expiry strike on one axis, expiry on another, IV on the third. the surface builds itself black-scholes won a nobel for a formula the market has been visibly disagreeing with since 1987 the disagreement is the signal full breakdown in the article below

89,999 次观看 • 17 天前

in 2019 my grandfather sold every stock he owned not because he saw a crash coming because he did the math on something Wall Street hides from you on purpose a 10-year Treasury pays 4.3% right now guaranteed. by the US government. twice a year. for a decade no earnings misses. no CEO scandal. no rug pull just a fixed schedule of cash arriving in your account every 6 months until 2034 $100k in Treasuries pays $4,300 a year in coupons alone $500k pays $21,500 $1M pays $43,000 that's not a return you have to trade for it's a return the Treasury owes you by law your broker doesn't mention it because they make nothing when you buy one no commission, no advisory fee, no reason to call meanwhile the wealthiest families in America have quietly built Treasury ladders for 200 years stagger maturities across 1, 2, 5, 10, 20 years coupons arrive every month liquidity every year it's not sexy it's not a 10x it's the thing that pays you while you sleep and never asks anything back my grandfather figured this out at 68 retired two years later watch the video. the professor explains the whole ladder in 82 seconds

in 2019 my grandfather sold every stock he owned not because he saw a crash coming because he did the math on something Wall Street hides from you on purpose a 10-year Treasury pays 4.3% right now guaranteed. by the US government. twice a year. for a decade no earnings misses. no CEO scandal. no rug pull just a fixed schedule of cash arriving in your account every 6 months until 2034 $100k in Treasuries pays $4,300 a year in coupons alone $500k pays $21,500 $1M pays $43,000 that's not a return you have to trade for it's a return the Treasury owes you by law your broker doesn't mention it because they make nothing when you buy one no commission, no advisory fee, no reason to call meanwhile the wealthiest families in America have quietly built Treasury ladders for 200 years stagger maturities across 1, 2, 5, 10, 20 years coupons arrive every month liquidity every year it's not sexy it's not a 10x it's the thing that pays you while you sleep and never asks anything back my grandfather figured this out at 68 retired two years later watch the video. the professor explains the whole ladder in 82 seconds

80,689 次观看 • 29 天前

a quant who started as a technical trader explained what changed everything for him he stopped trying to predict where price goes he started measuring the probability of each market state and betting only when the math was asymmetric that shift took him from drawing trendlines to writing models at 22 the realization was simple but brutal: technical analysis gives you a narrative quantitative analysis gives you a number one feels right the other is testable, repeatable, and either works or doesn't across 10,000 trades the setup on his desk tells the whole story no TradingView. no candlestick charts. just code, data, and a terminal he didn't go to MIT he didn't intern at Goldman he learned Python, statistics, and probability theory on his own and built something that actually worked > the math: free in any stats textbook > the data: free on Yahoo Finance, FRED, exchange APIs > the code: Python, 200 lines, running on a laptop > the barrier: not intelligence. just knowing this path exists most people spend years staring at indicators from the 1970s wondering why they can't find consistency the answer was never a better indicator it was a completely different framework one that treats trading as a math problem, not a prediction game full breakdown in the video below

a quant who started as a technical trader explained what changed everything for him he stopped trying to predict where price goes he started measuring the probability of each market state and betting only when the math was asymmetric that shift took him from drawing trendlines to writing models at 22 the realization was simple but brutal: technical analysis gives you a narrative quantitative analysis gives you a number one feels right the other is testable, repeatable, and either works or doesn't across 10,000 trades the setup on his desk tells the whole story no TradingView. no candlestick charts. just code, data, and a terminal he didn't go to MIT he didn't intern at Goldman he learned Python, statistics, and probability theory on his own and built something that actually worked > the math: free in any stats textbook > the data: free on Yahoo Finance, FRED, exchange APIs > the code: Python, 200 lines, running on a laptop > the barrier: not intelligence. just knowing this path exists most people spend years staring at indicators from the 1970s wondering why they can't find consistency the answer was never a better indicator it was a completely different framework one that treats trading as a math problem, not a prediction game full breakdown in the video below

112,271 次观看 • 1 个月前

this model doesn't predict where the S&P 500 will go it predicts when the market's own velocity is about to collapse Z(x,y) = F(β, α, τ, ∇τ; x, y, t) one equation. four parameters. a 3D surface that maps how price momentum evolves across time and space β is drift velocity: μ/σ, the ratio of expected return to volatility when β is high the market is moving with conviction when β decays toward zero the trend is losing energy before the chart shows any sign of it α measures how fast that velocity is changing τ is the time structure of the regime ∇τ is the gradient, the rate at which the regime itself is shifting the 3D surface on screen is not decoration the red peaks are where momentum is concentrated and unstable the blue basin is where the system is calm and mean-reverting the yellow marker is where the S&P sat at the moment of the snapshot the model's output: position sizing optimization not buy or sell. how much exposure to hold given the current position on that surface the bottom chart shows it: blue line is the model, white line is buy & hold same asset, same period, different sizing at every point based on where the dynamics equation said the market was > time dynamics models: rooted in physics, applied to finance since the 1990s > drift-to-volatility ratio: standard risk metric at Two Sigma, AQR, Man Group > this exact framework: free, public, 60 seconds in this video > what retail uses instead: lagging indicators that measure the past, not the velocity of the present retail asks "is the market going up or down?" this model asks "how fast is the market's own energy decaying and where on the surface are we right now?" completely different question. completely different result full breakdown in the video below

this model doesn't predict where the S&P 500 will go it predicts when the market's own velocity is about to collapse Z(x,y) = F(β, α, τ, ∇τ; x, y, t) one equation. four parameters. a 3D surface that maps how price momentum evolves across time and space β is drift velocity: μ/σ, the ratio of expected return to volatility when β is high the market is moving with conviction when β decays toward zero the trend is losing energy before the chart shows any sign of it α measures how fast that velocity is changing τ is the time structure of the regime ∇τ is the gradient, the rate at which the regime itself is shifting the 3D surface on screen is not decoration the red peaks are where momentum is concentrated and unstable the blue basin is where the system is calm and mean-reverting the yellow marker is where the S&P sat at the moment of the snapshot the model's output: position sizing optimization not buy or sell. how much exposure to hold given the current position on that surface the bottom chart shows it: blue line is the model, white line is buy & hold same asset, same period, different sizing at every point based on where the dynamics equation said the market was > time dynamics models: rooted in physics, applied to finance since the 1990s > drift-to-volatility ratio: standard risk metric at Two Sigma, AQR, Man Group > this exact framework: free, public, 60 seconds in this video > what retail uses instead: lagging indicators that measure the past, not the velocity of the present retail asks "is the market going up or down?" this model asks "how fast is the market's own energy decaying and where on the surface are we right now?" completely different question. completely different result full breakdown in the video below

72,275 次观看 • 1 个月前

Bridgewater doesn't have 500 positions they have 5 exposures that explain 78% of everything those 500 positions do this video shows what that looks like in 3D PCA Risk-Mode Trajectory. three axes: PC1, PC2, PC3 each one is a principal component extracted from the return matrix of the entire market PC1 is broad market direction. it alone explains roughly 40% of all stock movement PC2 separates growth from value. another 15% PC3 captures rate sensitivity. another 10% three invisible forces. 65% of the variance in every stock you've ever traded the dot moving through that 3D space is where the market is right now when the dot drifts toward a corner, the market is concentrating into one regime when it sits in the center, forces are balanced and directional trades are noise quant desks don't watch that dot for fun. they use its position to size every trade in the book if PC1 is dominating, stock picking is pointless because everything moves together if PC2 is leading, the value vs growth rotation is the only trade that matters if PC3 spikes, interest rate sensitivity is driving everything and your "stock pick" is really a rate bet you didn't know you made > PCA: Karl Pearson, 1901 > used at Bridgewater, AQR, Two Sigma for factor decomposition since the 1990s > this visualization: free, public, 29 seconds > scipy builds PCA in 3 lines of Python retail analyzes 500 tickers and thinks they're making 500 decisions quant desks decompose the same 500 into 5 forces and make 5 decisions one is complexity. the other is the math that simplifies it full breakdown in the video below

Bridgewater doesn't have 500 positions they have 5 exposures that explain 78% of everything those 500 positions do this video shows what that looks like in 3D PCA Risk-Mode Trajectory. three axes: PC1, PC2, PC3 each one is a principal component extracted from the return matrix of the entire market PC1 is broad market direction. it alone explains roughly 40% of all stock movement PC2 separates growth from value. another 15% PC3 captures rate sensitivity. another 10% three invisible forces. 65% of the variance in every stock you've ever traded the dot moving through that 3D space is where the market is right now when the dot drifts toward a corner, the market is concentrating into one regime when it sits in the center, forces are balanced and directional trades are noise quant desks don't watch that dot for fun. they use its position to size every trade in the book if PC1 is dominating, stock picking is pointless because everything moves together if PC2 is leading, the value vs growth rotation is the only trade that matters if PC3 spikes, interest rate sensitivity is driving everything and your "stock pick" is really a rate bet you didn't know you made > PCA: Karl Pearson, 1901 > used at Bridgewater, AQR, Two Sigma for factor decomposition since the 1990s > this visualization: free, public, 29 seconds > scipy builds PCA in 3 lines of Python retail analyzes 500 tickers and thinks they're making 500 decisions quant desks decompose the same 500 into 5 forces and make 5 decisions one is complexity. the other is the math that simplifies it full breakdown in the video below

49,014 次观看 • 1 个月前

a $40/month server beat a room full of analysts to the same trade by five and a half hours market opens at 9:30. his position was already in at 4am the system is a neural net trained on 11 years of tick data. it flagged the setup before the candle that "confirmed" it had even started forming this is the part retail misunderstands about ML in markets it isn't prediction in the mystical sense. it's pattern classification at a speed and scale human eyes physically cannot match the mechanics: 847,000 labeled historical setups as training data 4,200 data points per second ingested live each new state scored against every pattern the net has ever seen, in milliseconds the model isn't asking "where is price going" it's asking "how closely does the current microstructure match the conditions that preceded a move in my training set" that's a classification problem, and classification is what neural nets do better than anything else output: 3-4 candidate trades a day. he takes the top 2 by confidence score last 90 days: 71% win rate at 2.3 average risk-reward the edge isn't the architecture. the architecture is public pytorch is free, the papers are on arxiv, the network is a few hundred lines the edge is the labeling. what you feed it and how you tag the setups is the entire game retail feeds a model price and time and gets noise a desk feeds it order flow, volatility state, cross-asset context, each example hand-labeled by outcome same network. different training data. that's the whole difference retail watches the news at the open and reacts this system scored every pattern before sunrise and already decided you're not losing because your analysis is wrong you're losing to something that doesn't sleep, doesn't panic, and doesn't second-guess a probability it already computed the dataset was free. the framework was free. the compute was $40 a month the edge was never behind a paywall. it was sitting in a format almost nobody bothered to train on full breakdown in the article below

a $40/month server beat a room full of analysts to the same trade by five and a half hours market opens at 9:30. his position was already in at 4am the system is a neural net trained on 11 years of tick data. it flagged the setup before the candle that "confirmed" it had even started forming this is the part retail misunderstands about ML in markets it isn't prediction in the mystical sense. it's pattern classification at a speed and scale human eyes physically cannot match the mechanics: 847,000 labeled historical setups as training data 4,200 data points per second ingested live each new state scored against every pattern the net has ever seen, in milliseconds the model isn't asking "where is price going" it's asking "how closely does the current microstructure match the conditions that preceded a move in my training set" that's a classification problem, and classification is what neural nets do better than anything else output: 3-4 candidate trades a day. he takes the top 2 by confidence score last 90 days: 71% win rate at 2.3 average risk-reward the edge isn't the architecture. the architecture is public pytorch is free, the papers are on arxiv, the network is a few hundred lines the edge is the labeling. what you feed it and how you tag the setups is the entire game retail feeds a model price and time and gets noise a desk feeds it order flow, volatility state, cross-asset context, each example hand-labeled by outcome same network. different training data. that's the whole difference retail watches the news at the open and reacts this system scored every pattern before sunrise and already decided you're not losing because your analysis is wrong you're losing to something that doesn't sleep, doesn't panic, and doesn't second-guess a probability it already computed the dataset was free. the framework was free. the compute was $40 a month the edge was never behind a paywall. it was sitting in a format almost nobody bothered to train on full breakdown in the article below

20,790 次观看 • 21 天前

implied vol was 11.42% on Feb 19, 2020 realized vol over the next 30 days: 95.33% the market was pricing a normal week. what actually happened was one of the most violent moves in financial history that gap between implied and realized volatility is the single most important number quant desks watch implied vol is what options are pricing in for the future realized vol is what actually happened when implied is far below realized, options are cheap relative to what the market will actually do when implied is far above realized, options are expensive relative to what the market will actually do that spread has a name: the volatility risk premium Citadel, Susquehanna, Optiver built entire options desks around one question: what's the expected value of shorting overpriced vol vs buying underpriced vol? the scatter plot in the video shows every data point over years of history regression line: y = -0.73x + 18.5% translation: when implied vol is low, realized vol tends to explode. when implied vol is high, realized vol tends to collapse that's mean reversion in volatility itself. not price. vol retail buys calls hoping the stock goes up quant desks measure whether the option itself is mispriced relative to statistical expectation, and trade the premium one is a directional bet the other is arbitraging the gap between what people expect and what actually happens > volatility risk premium: documented in academic literature since the 1980s > average VRP on S&P 500: implied vol overprices realized vol by 3-5% historically > data: free from CBOE, Yahoo Finance, options chains > implementation: 20 lines of Python Feb 19, 2020 was a warning the math had been printing for weeks nobody in retail was reading it full breakdown in the video below

implied vol was 11.42% on Feb 19, 2020 realized vol over the next 30 days: 95.33% the market was pricing a normal week. what actually happened was one of the most violent moves in financial history that gap between implied and realized volatility is the single most important number quant desks watch implied vol is what options are pricing in for the future realized vol is what actually happened when implied is far below realized, options are cheap relative to what the market will actually do when implied is far above realized, options are expensive relative to what the market will actually do that spread has a name: the volatility risk premium Citadel, Susquehanna, Optiver built entire options desks around one question: what's the expected value of shorting overpriced vol vs buying underpriced vol? the scatter plot in the video shows every data point over years of history regression line: y = -0.73x + 18.5% translation: when implied vol is low, realized vol tends to explode. when implied vol is high, realized vol tends to collapse that's mean reversion in volatility itself. not price. vol retail buys calls hoping the stock goes up quant desks measure whether the option itself is mispriced relative to statistical expectation, and trade the premium one is a directional bet the other is arbitraging the gap between what people expect and what actually happens > volatility risk premium: documented in academic literature since the 1980s > average VRP on S&P 500: implied vol overprices realized vol by 3-5% historically > data: free from CBOE, Yahoo Finance, options chains > implementation: 20 lines of Python Feb 19, 2020 was a warning the math had been printing for weeks nobody in retail was reading it full breakdown in the video below

28,202 次观看 • 1 个月前

MIT defines an algorithm in one sentence that changes how you think about trading "a computational procedure that takes an input and produces an output through a well-defined sequence of steps" that's it. not AI. not machine learning. not a black box a set of rules that takes data in and spits a decision out every quant strategy ever built is just an algorithm Citadel's execution system that routes 40% of US equity volume is an algorithm Renaissance's Medallion Fund running millions of trades per year is an algorithm Jane Street's market making engine processing $26 trillion annually is an algorithm input: market data rules: mathematical conditions output: trade or no trade the difference between a quant desk and a retail trader is not the data it's that one side wrote down their rules precisely enough for a machine to execute them retail says "if RSI is low and the chart looks good, i'll probably buy" a quant desk says "if RSI 1.5, buy 0.3% of NAV" same logic. one is a feeling. the other is an algorithm the feeling can't be tested, can't be repeated, can't be measured the algorithm can be backtested across 10,000 trades and you know exactly when it works and when it doesn't > this lecture: MIT, free, 70 seconds > algorithmic trading volume: 60-75% of all US equity trades > Jane Street, Citadel, Two Sigma: every trade is algorithmically executed > tools to build your own: Python, free data, a laptop you don't need a faster computer or better data you need to write your strategy down precisely enough that a machine could run it without you that's the whole leap. from intuition to algorithm full breakdown in the video below

MIT defines an algorithm in one sentence that changes how you think about trading "a computational procedure that takes an input and produces an output through a well-defined sequence of steps" that's it. not AI. not machine learning. not a black box a set of rules that takes data in and spits a decision out every quant strategy ever built is just an algorithm Citadel's execution system that routes 40% of US equity volume is an algorithm Renaissance's Medallion Fund running millions of trades per year is an algorithm Jane Street's market making engine processing $26 trillion annually is an algorithm input: market data rules: mathematical conditions output: trade or no trade the difference between a quant desk and a retail trader is not the data it's that one side wrote down their rules precisely enough for a machine to execute them retail says "if RSI is low and the chart looks good, i'll probably buy" a quant desk says "if RSI 1.5, buy 0.3% of NAV" same logic. one is a feeling. the other is an algorithm the feeling can't be tested, can't be repeated, can't be measured the algorithm can be backtested across 10,000 trades and you know exactly when it works and when it doesn't > this lecture: MIT, free, 70 seconds > algorithmic trading volume: 60-75% of all US equity trades > Jane Street, Citadel, Two Sigma: every trade is algorithmically executed > tools to build your own: Python, free data, a laptop you don't need a faster computer or better data you need to write your strategy down precisely enough that a machine could run it without you that's the whole leap. from intuition to algorithm full breakdown in the video below

23,905 次观看 • 1 个月前

an MIT professor who taught both physics and finance told his class something none of them expected "finance is harder than physics" not as a joke. as a mathematical statement in physics, the laws don't change. gravity works the same today as it did a billion years ago you can run an experiment, get a result, and repeat it forever in finance, the moment you discover a law, the participants learn it too and their behavior changes the system you just measured in physics, electrons don't read your paper and start moving differently in finance, traders do. every published edge gets arbitraged away by the people who read it this is why quant models have a half-life and physics equations don't Newton's laws: 300+ years and counting Long-Term Capital Management's model: worked perfectly until it didn't, lost $4.6 billion in 4 months the system you're modeling is aware of you modeling it that's not a solvable problem. it's a permanent condition and the quants who survive are the ones who build for it instead of pretending it doesn't exist > this lecture: MIT finance series, free, public, 53 seconds > LTCM collapse: 1998, Nobel Prize winners, $4.6B loss > Renaissance's solution: never stop researching, replace signals before they decay > average lifespan of a quant signal: 2-5 years before it's crowded out retail builds one strategy and trades it until it breaks quant desks build a research engine that produces new strategies faster than old ones die that's not a difference in skill. it's a difference in understanding what game you're actually playing full breakdown in the video below

an MIT professor who taught both physics and finance told his class something none of them expected "finance is harder than physics" not as a joke. as a mathematical statement in physics, the laws don't change. gravity works the same today as it did a billion years ago you can run an experiment, get a result, and repeat it forever in finance, the moment you discover a law, the participants learn it too and their behavior changes the system you just measured in physics, electrons don't read your paper and start moving differently in finance, traders do. every published edge gets arbitraged away by the people who read it this is why quant models have a half-life and physics equations don't Newton's laws: 300+ years and counting Long-Term Capital Management's model: worked perfectly until it didn't, lost $4.6 billion in 4 months the system you're modeling is aware of you modeling it that's not a solvable problem. it's a permanent condition and the quants who survive are the ones who build for it instead of pretending it doesn't exist > this lecture: MIT finance series, free, public, 53 seconds > LTCM collapse: 1998, Nobel Prize winners, $4.6B loss > Renaissance's solution: never stop researching, replace signals before they decay > average lifespan of a quant signal: 2-5 years before it's crowded out retail builds one strategy and trades it until it breaks quant desks build a research engine that produces new strategies faster than old ones die that's not a difference in skill. it's a difference in understanding what game you're actually playing full breakdown in the video below

18,961 次观看 • 1 个月前

an ex-Goldman banker said the key to understanding markets is reading 3 WSJ articles a day a quant at Renaissance would tell you that's the exact wrong input here's why the Wall Street Journal publishes after the move by the time you read "tech stocks fall on rate fears" the move already happened, the positioning already shifted, and every algorithm already repriced news explains what happened. it doesn't predict what's next quant desks don't read news to form trade ideas they measure whether news has a statistically significant effect on price after controlling for everything else the answer, across decades of academic research: individual news articles explain less than 0.5% of daily price variance the other 99.5% is flow, positioning, regime, volatility structure, and cross-asset correlation none of which shows up in a WSJ headline this is the Goldman mindset vs the quant mindset Goldman: read narratives, form opinions, act on conviction Renaissance: measure everything, strip out noise, trade only what survives statistical testing one produced traders who sound smart at dinner parties the other produced $100 billion in profit over 30 years > WSJ subscription: $468/year > SEC EDGAR data: free > FRED economic data: free > exchange-level order flow: free > academic papers on news impact studies: free on ArXiv since the 2000s the banker tells you to read the news the quant tells you to measure whether the news matters those are not the same instruction full breakdown in the video below

an ex-Goldman banker said the key to understanding markets is reading 3 WSJ articles a day a quant at Renaissance would tell you that's the exact wrong input here's why the Wall Street Journal publishes after the move by the time you read "tech stocks fall on rate fears" the move already happened, the positioning already shifted, and every algorithm already repriced news explains what happened. it doesn't predict what's next quant desks don't read news to form trade ideas they measure whether news has a statistically significant effect on price after controlling for everything else the answer, across decades of academic research: individual news articles explain less than 0.5% of daily price variance the other 99.5% is flow, positioning, regime, volatility structure, and cross-asset correlation none of which shows up in a WSJ headline this is the Goldman mindset vs the quant mindset Goldman: read narratives, form opinions, act on conviction Renaissance: measure everything, strip out noise, trade only what survives statistical testing one produced traders who sound smart at dinner parties the other produced $100 billion in profit over 30 years > WSJ subscription: $468/year > SEC EDGAR data: free > FRED economic data: free > exchange-level order flow: free > academic papers on news impact studies: free on ArXiv since the 2000s the banker tells you to read the news the quant tells you to measure whether the news matters those are not the same instruction full breakdown in the video below

15,317 次观看 • 1 个月前

没有更多内容可加载