Kolmogorov-Arnold Networks (KANs) for Time Series Analysis - paper posits that KANs are reasonable alternative to MLPs with several advantages…<a href="https://t.co/ofNs3aHRx9">https://t.co/ofNs3aHRx9</a> <a href="https://t.co/r5SO3t7QYA">pic.twitter.com/r5SO3t7QYA</a>
https://t.co/ofNs3aHRx9
https://twitter.com/carlcarrie/status/1791640297287368836Liquid Factor Models (yes, is an Alpha Model)<br><br>Paper:<a href="https://t.co/APV3eQV3ly">https://t.co/APV3eQV3ly</a> <a href="https://t.co/dMFiQU0MaO">pic.twitter.com/dMFiQU0MaO</a>
https://t.co/APV3eQV3ly
https://twitter.com/carlcarrie/status/1790882619401404699Technical Analysis and Currency Trading: False Discoveries and Informative Covariates<br>The paper below has recently been revised. " We apply the new method to a large universe of currency technical trading rules, and construct a dynamic 30-currency portfolio that generates a… <a href="https://t.co/Dr46C2xqvi">https://t.co/Dr46C2xqvi</a>
https://t.co/Dr46C2xqvi
https://twitter.com/quantseeker/status/1790843024223699084Ode to New Diffusion Applications<a href="https://t.co/k1ax8z49iy">https://t.co/k1ax8z49iy</a> <a href="https://t.co/qhupnQU81Z">pic.twitter.com/qhupnQU81Z</a>
https://t.co/k1ax8z49iy
https://twitter.com/carlcarrie/status/1790539993913467021Seven Sins of Quantitative Investing<br>Nice discussion by Deutsche Bank on common pitfalls and mistakes when backtesting. Several illustrative examples are provided. See link in tweet below. <a href="https://t.co/yiGhVySvXC">https://t.co/yiGhVySvXC</a> <a href="https://t.co/qYJQpeiRkZ">pic.twitter.com/qYJQpeiRkZ</a>
https://t.co/yiGhVySvXC
https://twitter.com/quantseeker/status/1790324812184743994Great model! Here's an analysis I wrote on MOMENT: <a href="https://t.co/Is99JQV1dR">https://t.co/Is99JQV1dR</a>
https://t.co/Is99JQV1dR
https://twitter.com/nikos_kafritsas/status/1789820133021462755Moment, a Time Series - <br>family of open-source foundation models for general-purpose time-series analysis (~LLM)<br><br>Python GitHub:<a href="https://t.co/Zj8Ppy0hAq">https://t.co/Zj8Ppy0hAq</a><br><br>Paper:<a href="https://t.co/FomYbtRhPc">https://t.co/FomYbtRhPc</a><br><br>Model Card:<a href="https://t.co/SESKZuxFeb">https://t.co/SESKZuxFeb</a> <a href="https://t.co/dZvHqJBN6X">pic.twitter.com/dZvHqJBN6X</a>
https://t.co/Zj8Ppy0hAq
https://twitter.com/carlcarrie/status/1789815228449820741Forecasting: theory and practice<br>"We provide an overview of a wide range of theoretical, state-of-the-art models, methods, principles, and approaches to prepare, produce, organise, and evaluate forecasts."<a href="https://t.co/j6MhsZuG8H">https://t.co/j6MhsZuG8H</a> <a href="https://t.co/8I69z8A2yD">https://t.co/8I69z8A2yD</a>
https://t.co/j6MhsZuG8H
https://twitter.com/quantseeker/status/1789755048647385538"I construct strategies based on exchange rate jumps only, which perform better than pure momentum or reversal strategies with Sharpe ratios exceeding one." <a href="https://t.co/nUqAEX7lwP">https://t.co/nUqAEX7lwP</a>
https://t.co/nUqAEX7lwP
https://twitter.com/quantseeker/status/1789634405918531924Perspectives on AI Agent Ops - report<a href="https://t.co/yHqauYMajV">https://t.co/yHqauYMajV</a> <a href="https://t.co/uzhlluPlM4">pic.twitter.com/uzhlluPlM4</a>
https://t.co/yHqauYMajV
https://twitter.com/carlcarrie/status/1789150304170033517AQR Paper: Can Machines Time Markets? The Virtue of Complexity in Return Prediction<a href="https://t.co/PIxMUqdctb">https://t.co/PIxMUqdctb</a> <a href="https://t.co/5dbwrXqowF">pic.twitter.com/5dbwrXqowF</a>
https://t.co/PIxMUqdctb
https://twitter.com/carlcarrie/status/1789148926341197935Short Interest and Aggregate Stock Returns: International Evidence<br><br>Short interest predicts stock returns negatively in most countries. <a href="https://t.co/Pt2ycyuOTM">https://t.co/Pt2ycyuOTM</a>
https://t.co/Pt2ycyuOTM
https://twitter.com/quantseeker/status/1788856495804694771Commodity Prices and Currencies<br><br>"...changes in a country's commodity export prices predict its exchange rate." <a href="https://t.co/uRy4TrINHD">https://t.co/uRy4TrINHD</a>
https://t.co/uRy4TrINHD
https://twitter.com/quantseeker/status/1788690795207283123New Erb & Harvey study on the drivers of <a href="https://twitter.com/hashtag/gold?src=hash&ref_src=twsrc%5Etfw">#gold</a> prices. On ETFs, China, Costco buyers, and difficulties with finding reliable data. Correlation or causation? <a href="https://twitter.com/jsblokland?ref_src=twsrc%5Etfw">@jsblokland</a> <a href="https://twitter.com/camharvey?ref_src=twsrc%5Etfw">@camharvey</a> <a href="https://t.co/zsioCcNr5S">https://t.co/zsioCcNr5S</a> <a href="https://t.co/twyFuhRTF6">pic.twitter.com/twyFuhRTF6</a>
https://twitter.com/hashtag/gold?src=hash&ref_src=twsrc%5Etfw
https://twitter.com/paradoxinvestor/status/1788101866091315317Interesting white paper from the AQR crew of <a href="https://twitter.com/CliffordAsness?ref_src=twsrc%5Etfw">@cliffordasness</a> on the virtue of complexity in return prediction.<br><br>Complex models (number of predictors > number of observations) better capture nonlinearities, leading to improved return predictions and “evolutionary, not… <a href="https://t.co/2w9WOpLkP3">pic.twitter.com/2w9WOpLkP3</a>
https://twitter.com/CliffordAsness?ref_src=twsrc%5Etfw
https://twitter.com/quantseeker/status/1787941524438663194Lasse Pedersen at Copenhagen Business School offers a great course in "Big Data Asset Pricing" with really useful lecture notes.<br><br>Topics include: Empirical asset pricing, working with asset-pricing data, multiple testing issues and the factor zoo, machine learning, and more.… <a href="https://t.co/4BXMdGCvfl">pic.twitter.com/4BXMdGCvfl</a>
https://t.co/4BXMdGCvfl
https://twitter.com/quantseeker/status/1787793433026818450Building Energy Prediction is a time-series problem. SARIMA, XGBoost & firefly metaheuristic algorithms (MetaFA-LSSVR) were benchmarks confirming the superior prediction ability of memory-based models (Seq2Seq & RNN)<br><br>Paper:<a href="https://t.co/W22azqxpbV">https://t.co/W22azqxpbV</a><br><br>Code:<a href="https://t.co/Z4VGIKvL5U">https://t.co/Z4VGIKvL5U</a> <a href="https://t.co/2jLgoTxfoz">pic.twitter.com/2jLgoTxfoz</a>
https://t.co/W22azqxpbV
https://twitter.com/carlcarrie/status/1787433749426098263Earnings Conference Call Analyzer - LLM paper <a href="https://t.co/0gJwfZk9Up">https://t.co/0gJwfZk9Up</a><br><br>FinX StockX <a href="https://t.co/rKbCtXn8Jg">pic.twitter.com/rKbCtXn8Jg</a>
https://t.co/0gJwfZk9Up
https://twitter.com/carlcarrie/status/1786946034523889841Systemic Risk detection using intraday 1 minute data / realized eigenvalues - paper<a href="https://t.co/WLtud3AIBc">https://t.co/WLtud3AIBc</a> <a href="https://t.co/v74KQpbrej">pic.twitter.com/v74KQpbrej</a>
https://t.co/WLtud3AIBc
https://twitter.com/carlcarrie/status/1786941082418225322New paper on carry strategies and their potential diversification benefits <a href="https://t.co/hTkSFenCuo">https://t.co/hTkSFenCuo</a>
https://t.co/hTkSFenCuo
https://twitter.com/quantseeker/status/1786495820805652971Corporate bond factors: "...integrated multifactor portfolios outperform mixed multifactor portfolios." <a href="https://t.co/I6pfuzTcbb">https://t.co/I6pfuzTcbb</a>
https://t.co/I6pfuzTcbb
https://twitter.com/quantseeker/status/1785678034651025730Great article on Jane Street <a href="https://t.co/TbzndHHMwW">https://t.co/TbzndHHMwW</a>
https://t.co/TbzndHHMwW
https://twitter.com/quantseeker/status/1785283237754847548Hmmm<br><br>New Paper: CRISPR-GPT: An LLM Agent for Automated Design of Gene-Editing Experiments<br><br>How do you validate the LLM CRISPR design creation? <br><br>What could go wrong?<br><br> <a href="https://t.co/HpdhRoTNEG">https://t.co/HpdhRoTNEG</a>
https://t.co/HpdhRoTNEG
https://twitter.com/carlcarrie/status/1785269976930116051really good article on Jane Street<br><br>Jane Street is big. Like, really, really big<br><br>An article only someone like <a href="https://twitter.com/RobinWigg?ref_src=twsrc%5Etfw">@RobinWigg</a> could write. Nice job!<a href="https://t.co/2UGRibm9KS">https://t.co/2UGRibm9KS</a>
https://twitter.com/RobinWigg?ref_src=twsrc%5Etfw
https://twitter.com/ltabb/status/1784966978492334326The often called emergent ChatGPT mystery has apparently been solved:<br><br>To 'delve' or not to delve - that is the question. <br><br>The answer has been found... <br><br>in Nigeria...<br> <a href="https://t.co/PVzYRaB9Ix">https://t.co/PVzYRaB9Ix</a>
https://t.co/PVzYRaB9Ix
https://twitter.com/carlcarrie/status/1783474253402194195"Cryptocurrencies. If You're Gonna Do It, Do It Right"<br><br>Note from <a href="https://twitter.com/ManGroup?ref_src=twsrc%5Etfw">@ManGroup</a> on allocating to crypto.<a href="https://t.co/tjbzPm8Old">https://t.co/tjbzPm8Old</a> <a href="https://t.co/0jfZZV6GcV">pic.twitter.com/0jfZZV6GcV</a>
https://twitter.com/ManGroup?ref_src=twsrc%5Etfw
https://twitter.com/quantseeker/status/1783455023902372095A new paper by Goyenko et al. uses machine learning to predict trading volume and find volume to be “highly predictable”. Predictors include technical signals, firm characteristics, calendar information, and earnings schedules. <br><br>Volume predictability is incorporated into a… <a href="https://t.co/aRwBS4ZRGY">pic.twitter.com/aRwBS4ZRGY</a>
https://t.co/aRwBS4ZRGY
https://twitter.com/quantseeker/status/1782706064670048536Interesting illustration of the finding that analyst forecast dispersion tends to predict stock returns negatively in the cross-section. <a href="https://t.co/tA2EkElhis">https://t.co/tA2EkElhis</a>
https://t.co/tA2EkElhis
https://twitter.com/quantseeker/status/1782501063133683727The risk of rare disasters is one of many explanations put forward to explain why equities should be subject to a risk premium. However, estimating such risks is challenging due to the scarcity of macroeconomic disasters. <br><br>This paper pools data from over 40 countries and… <a href="https://t.co/lxYnJVKs0q">pic.twitter.com/lxYnJVKs0q</a>
https://t.co/lxYnJVKs0q
https://twitter.com/quantseeker/status/1782492026857988556The finance sector is realizing that over the past several decades, it has lost millions, even billions, of dollars due to uncalibrated models.<br><br>What's the most effective technology for calibrating these models without reengineering the whole pipelines?<br><br>Undoubtedly, <a href="https://twitter.com/hashtag/conformal?src=hash&ref_src=twsrc%5Etfw">#conformal</a>… <a href="https://t.co/laUjuqJO93">pic.twitter.com/laUjuqJO93</a>
https://twitter.com/hashtag/conformal?src=hash&ref_src=twsrc%5Etfw
https://twitter.com/predict_addict/status/1782363233480843687Hmmm<br><br>OpenAI allegedly used illegally scraped data to train Whisper<br> <a href="https://t.co/GlccIPGcFM">https://t.co/GlccIPGcFM</a>
https://t.co/GlccIPGcFM
https://twitter.com/carlcarrie/status/1782095619877716419Six great books on trading, investing, and portfolio management. <a href="https://t.co/pkyW4trnOX">pic.twitter.com/pkyW4trnOX</a>
https://t.co/pkyW4trnOX
https://twitter.com/quantseeker/status/1781703780788052213Prompt Principles Paper when interacting with LLMs: Handy rules of thumb for optimizing output quality / prompt engineering:<a href="https://t.co/CRoTVmZeVH">https://t.co/CRoTVmZeVH</a> <a href="https://t.co/etKjwsMGJc">pic.twitter.com/etKjwsMGJc</a>
https://t.co/CRoTVmZeVH
https://twitter.com/carlcarrie/status/1781299831299833977<blockquote class="twitter-tweet" data-width="550"><p lang="en" dir="ltr">As Johnny von Neumann
used to say, "with four parameters I can fit
an elephant, and with five I can make him
wiggle his trunk."<br><br>2014 Paper: Pseudo-Mathematics and Financial
Charlatanism: The Effects of<br>Backtest Overfitting on
Out-of-Sample Performance <a href="https://t.co/eHGq2tc7z4">pic.twitter.com/eHGq2tc7z4</a></p>— Carl Carrie (@) (@carlcarrie) <a href="https://twitter.com/carlcarrie/status/1781295623590801551?ref_src=twsrc%5Etfw">April 19, 2024</a></blockquote>
<script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script> https://t.co/eHGq2tc7z4
https://twitter.com/carlcarrie/status/1781295623590801551Most forecasting tools - Prophet, TIDE, XGBoost underperform when challenged with financial data sets spanning multiple business cycles and reaching a new turning point <br><br>Dunning-Kruger effect mitigation approaches - eg detecting the onset of recession:<a href="https://t.co/oj7fCopZZP">https://t.co/oj7fCopZZP</a> <a href="https://t.co/GmZbk6tcUs">pic.twitter.com/GmZbk6tcUs</a>
https://t.co/oj7fCopZZP
https://twitter.com/carlcarrie/status/1781289934164316187Dynamic Factor Modeling using auto regressive techniques as a Python library: Metran (use cases beyond just hydro timeseries)<a href="https://t.co/U6HPakG1Sd">https://t.co/U6HPakG1Sd</a><br><br>Tutorial:<a href="https://t.co/Vso0EuTVWK">https://t.co/Vso0EuTVWK</a><br><br>Classic example:<a href="https://t.co/U6HPakG1Sd">https://t.co/U6HPakG1Sd</a><br><br>Dependencies on scipy and pastas
https://t.co/U6HPakG1Sd
https://twitter.com/carlcarrie/status/1780921056435241311"Democrats are good for the stock market, Republicans are good for the bond market. Trend and Cross-Sectional Momentum are both much stronger under the GOP." from <a href="https://twitter.com/ManGroup?ref_src=twsrc%5Etfw">@ManGroup</a> <br><br>Link: <a href="https://t.co/aQftAru9GW">https://t.co/aQftAru9GW</a> <a href="https://t.co/zWpaO32qtN">pic.twitter.com/zWpaO32qtN</a>
https://twitter.com/ManGroup?ref_src=twsrc%5Etfw
https://twitter.com/quantseeker/status/1780894912675758092Gradient Boost Regression Tree (GBRT) for Limit Order Book <a href="https://twitter.com/hashtag/LoB?src=hash&ref_src=twsrc%5Etfw">#LoB</a> modeling - paper<a href="https://t.co/g6PYDt36V2">https://t.co/g6PYDt36V2</a> <a href="https://t.co/bjqIW9of2a">pic.twitter.com/bjqIW9of2a</a>
https://twitter.com/hashtag/LoB?src=hash&ref_src=twsrc%5Etfw
https://twitter.com/carlcarrie/status/1780769703335510468How long will it be before a large language model with a whisper like voice model is embedded in an android?<a href="https://t.co/rsaov1mP2d">https://t.co/rsaov1mP2d</a> <a href="https://t.co/f7FAgAnMRq">pic.twitter.com/f7FAgAnMRq</a>
https://t.co/rsaov1mP2d
https://twitter.com/carlcarrie/status/1780654397086629982Tiny Time Mixers (TTMs): Fast Pre-trained Models for Zero/Few-Shot Forecasting of Multivariate Time Series<br><br>Python GitHub:<a href="https://t.co/MOfMbbiURC">https://t.co/MOfMbbiURC</a><br><br>Pretraining TTMs is very easy and fast, taking 3-6 hours using 6 A100 GPUs<br><br>Model Card:<a href="https://t.co/ZlqEcRFcSn">https://t.co/ZlqEcRFcSn</a><br><br>IBM LLM Research
https://t.co/MOfMbbiURC
https://twitter.com/carlcarrie/status/1780553403057258502If you are interested in financial history, this paper could be of interest. Using long-term data from <a href="https://twitter.com/GlobalFinData?ref_src=twsrc%5Etfw">@GlobalFinData</a> , the author describes movements in stocks and bonds over the past eight(!) centuries. The paper provides a fascinating review of historical eras and events and… <a href="https://t.co/LK81LQ4WTO">pic.twitter.com/LK81LQ4WTO</a>
https://twitter.com/GlobalFinData?ref_src=twsrc%5Etfw
https://twitter.com/quantseeker/status/17803709188977709702024 AI Index - Stanford University Report<a href="https://t.co/wHmzrqAhPe">https://t.co/wHmzrqAhPe</a> <a href="https://t.co/6oQ9kIg6F5">pic.twitter.com/6oQ9kIg6F5</a>
https://t.co/wHmzrqAhPe
https://twitter.com/carlcarrie/status/1780310491069587517This paper evaluates the ability of a range of machine learning models to predict corporate bond returns. The authors find significant return predictability, yielding meaningful Sharpe ratios net of transaction costs. Features related to both bond characteristics and aggregate… <a href="https://t.co/kItTyjzwV9">pic.twitter.com/kItTyjzwV9</a>
https://t.co/kItTyjzwV9
https://twitter.com/quantseeker/status/1780293968518922580The S&P500 yields now less than bonds, as shown in <a href="https://twitter.com/johnauthers?ref_src=twsrc%5Etfw">@johnauthers</a> newsletter today. However, not all stocks are created the same! US value stocks (definition in comment) still have fwd earnings yields of more than 7%. Internationally, you can even find more yield! <a href="https://t.co/xTubRObhSL">https://t.co/xTubRObhSL</a> <a href="https://t.co/bl2pO3DqPE">pic.twitter.com/bl2pO3DqPE</a> https://twitter.com/johnauthers?ref_src=twsrc%5Etfw
https://twitter.com/HanauerMatthias/status/1780200684115128491Six recommendations from my bookshelf. <a href="https://t.co/xOuZ33pMRO">pic.twitter.com/xOuZ33pMRO</a>
https://t.co/xOuZ33pMRO
https://twitter.com/quantseeker/status/1779949005134233973Algorithmic Trading and Quantitative Strategy Lecture 6 is posted: <a href="https://t.co/raWpVMtVja">https://t.co/raWpVMtVja</a><br>Topics:<br>- idiosyncratic covariance matrix estimation, including <br>* off-diagonal cluster analysis<br>* short-term idiovol updating (actually very effective)<br><br>- portfolio construction. Mostly:…
https://t.co/raWpVMtVja
https://twitter.com/__paleologo/status/1779934331068142029Hundreds of equity factors have been documented. Do they reflect mispricing or risk? Frey estimates that at least 40% of factors represent mispricing and most factors “reflect the convergences of prices back to fundamental values (resolution of mispricing)” as opposed to a… <a href="https://t.co/cvCE8K9Obl">pic.twitter.com/cvCE8K9Obl</a>
https://t.co/cvCE8K9Obl
https://twitter.com/quantseeker/status/1762852981844660532Sequoia's new top 50 GenAI companies list and their impact on price structures and productivity <a href="https://t.co/jVeYYbW4JS">https://t.co/jVeYYbW4JS</a> <a href="https://t.co/DIOrI7npOh">pic.twitter.com/DIOrI7npOh</a>
https://t.co/jVeYYbW4JS
https://twitter.com/carlcarrie/status/1778751217671278603How Do Stocks Perform During High Inflation? Not Good<br><br>That's bad news if you're looking to protect your investment portfolio... Our recent Financial Analyst Journal study has quite some impact. It uncovers very weak returns for equities and bonds during periods of high… <a href="https://t.co/At0u5EQ9yr">pic.twitter.com/At0u5EQ9yr</a>
https://t.co/At0u5EQ9yr
https://twitter.com/paradoxinvestor/status/1778748471094256127