Real, usable edge for systematic traders — strategies, backtesting and code, no fluff
54 articles
Multiarmed bandits formalize the explorevsexploit problem: you have K strategies (arms), each with unknown reward distributions, and you must allocate capital over time to maximize cumulative reward. Unlike a static [portfolio optimizer](/post.php?slug=portfoliooptimizationmarkowitz) that assumes known means and covariances, bandits learn…
Read →Causal inference asks whether changing X would change Y — not whether X and Y move together. In quantitative trading, most "signals" are correlational: momentum predicts returns insample, but is it because momentum causes price movement, or because both respond…
Read →Onchain metrics are quantitative measures derived from blockchain transaction data — wallet balances, exchange inflows/outflows, miner behavior, stablecoin mints, and network activity. Unlike priceonly signals, they claim to measure holder behavior and capital positioning directly. Some onchain signals have genuine…
Read →A regime model is useful only if it reacts to a distribution change before the old model causes unacceptable losses. Changepoint detection seeks timestamps where the parameters of a process—mean, variance, correlation, or microstructure behavior—shift. Unlike a fixedstate classifier, it…
Read →Alternative data is any nontraditional dataset — satellite imagery, cardpanel spend, web scrapes, geolocation, onchain flows, text sentiment — used to estimate an economic quantity before it shows up in prices or official statistics. The thesis is clean: measure a…
Read →Sentiment analysis converts unstructured text — news, social posts, filings, earningscall transcripts — into pointintime numeric signals. The premise is narrow but real: prices move on information, much of it arrives first as text, and a desk that reads and…
Read →Reinforcement learning (RL) learns a policy that maximizes cumulative reward through interaction — a structural fit for trading, where you make sequential, statealtering decisions under uncertainty. The promise is real for execution; for alpha generation it remains a research frontier…
Read →Random forests and gradient boosting (XGBoost, LightGBM, scikitlearn's HistGradientBoosting) are the workhorses of practical [machinelearning trading](/post.php?slug=machinelearningtrading). They model nonlinear, regimedependent interactions, tolerate messy heterogeneous features, train fast, and — critically for a lowsignal, smallsample domain — overfit less than deep…
Read →In [machinelearning trading](/post.php?slug=machinelearningtrading), feature engineering is where most of the realized edge — and almost all of the realized risk — actually lives. With a nearzero signaltonoise ratio, the marginal return on a better algorithm is small while the marginal…
Read →Machine learning in trading is an exercise in estimating a tiny conditional expectation, E[r{t+h} | Xt], from a short, autocorrelated, nonstationary sample where the irreducible noise dwarfs any signal. That framing — not the choice of architecture — determines whether…
Read →Markets evolve while a batch model waits for its next retraining run. Online learning updates parameters one observation at a time or in small batches, making it attractive for intraday forecasting, execution, and rapidly changing crosssectional signals. Its promise is…
Read →A firm’s revenue shock rarely stays inside one ticker. It moves through suppliers, customers, logistics providers, commodities, and competitors. A graph makes those relations explicit: nodes are entities, edges are economically defined links, and attributes are timestamped observations. The opportunity…
Read →Synthetic market data is generated rather than observed: simulated return paths, order books, trade sequences, or alternativedata records. It can expand scarce training samples, support stress testing, and let researchers share data without exposing raw records. But a generator that…
Read →Principal Component Analysis (PCA) diagonalizes the covariance (or correlation) matrix of a return panel, rotating correlated assets into orthogonal factors ranked by variance explained. In quant trading, PCA does four jobs: extracting the dominant risk factors of a universe, constructing…
Read →Market regime detection is the problem of inferring an unobserved, persistent market state — calmbull, volatilebear, crisis — from noisy returns, so you can run the right strategy and size risk for the environment you are actually in. Hidden Markov…
Read →Alpha combination is where good quant research turns into a deployable strategy. A single predictor is almost always weak — an information coefficient (IC) of 0.03–0.05 is a good signal in equities. The edge comes from combining many such weak,…
Read →Fractional differentiation resolves a tension at the heart of quant feature engineering: models need [stationary](/post.php?slug=stationaritytimeseries) inputs, but the standard way of achieving stationarity — taking integer differences (returns) — erases almost all of the memory that made the series predictive…
Read →The Kalman filter is exact only when transitions and observations are linear with Gaussian noise. Financial latentstate problems routinely violate both assumptions: volatility is positive and persistent, returns have jumps and heavy tails, and option quotes are nonlinear functions of…
Read →Labeling is the most underappreciated step in [machine learning for trading](/post.php?slug=machinelearningtrading). Everyone obsesses over features and models, but how you define "what the model should predict" silently determines whether the whole exercise is even sensible. The naive fixedtimehorizon label —…
Read →LSTM networks are the deeplearning model traders most want to work, and the one most likely to humble them. A long shortterm memory network is a recurrent neural network designed to carry information across long sequences — which sounds tailormade…
Read →Neural networks are the most overhyped and underdelivering tool in quant trading. They are universal function approximators that can, in principle, learn any relationship between features and returns — and that very flexibility is exactly why they fail so often…
Read →A return over a fixed interval loses the order in which the market travelled. For execution, barriers, drawdowns, and intraday flow, that order matters: a 1% gain after a 3% drawdown is not operationally equivalent to a smooth 1% gain.…
Read →Gaussian processes (GPs) are Bayesian models over functions. Rather than committing to one fitted curve, a GP produces a posterior distribution for the unknown mapping from features to a return, probability, volatility, or cost target. That uncertainty is valuable in…
Read →Correlation screens are effective at finding candidate signals, but they cannot establish whether a feature causes a return, reacts to it, or shares a common driver. Causal discovery algorithms search for graph structures that are compatible with conditional independences, temporal…
Read →Financial labels are scarce, noisy, and regimedependent. Transfer learning tries to improve a target task by reusing a representation, parameters, or data relationship learned on a source task. A model trained on liquid equities may help initialize a lessliquid equity…
Read →Transformers dominate language and vision, so it is natural to ask whether the attention architecture that powers large language models can find structure in price series. The honest answer, which this article defends in detail, is: rarely, and almost never…
Read →Bayesian methods are the natural framework for trading because trading is a smallsample, highnoise problem where you are never certain and you always have prior knowledge. The frequentist habit of plugging in a single point estimate — the sample mean…
Read →Online portfolio selection (OLPS) updates allocation after each return observation without fitting a stationary return model. The sequence is simple: select weights, observe returns, update. Its appeal is that regret—not prediction accuracy—is the native objective. Its limitation is equally important:…
Read →Bayesian online changepoint detection (BOCPD) turns the vague impression that “the regime changed” into a sequential posterior over how long the current regime has lasted. At each new observation it updates a distribution over the run length—the number of observations…
Read →Hierarchical Bayesian models solve a recurring trading problem: individual assets have too little data to estimate reliably, while treating every asset as identical erases useful structure. Partial pooling learns a shared population distribution and shrinks weakly observed assets toward it…
Read →Gaussian mixture models (GMMs) provide a probabilistic way to describe markets as a blend of distinct statistical environments rather than one stationary distribution. A mixture component may correspond to calm positive returns, highvolatility selloffs, or inflationary correlation shocks—but those labels…
Read →Satellite data can observe economic activity before it appears in earnings releases: cars in retail car parks, inventory at ports, flare intensity at oil fields, or construction progress at industrial sites. That observational advantage is real, but it is not…
Read →Credit card transaction data offers a nearrealtime view of consumer spending, often weeks before a company reports revenue. It can nowcast samestore sales, detect demand inflections, and measure category shifts. But the raw feed is a convenience panel, not the…
Read →Earnings calls contain management language, analyst challenges, guidance detail, and vocal cues that are not fully summarized by headline earnings. NLP can quantify that information, but naive positiveminusnegative word counts mostly rediscover whether the company reported good news. A tradable…
Read →Knowledge Graphs Finance Entities is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Temporal Graph Networks is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Graph Attention Supply Chains is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Temporal GANs Caution is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →GAN Mode Collapse Finance is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Variational Autoencoders Returns is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Autoencoder Data Cleaning is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Cost Sensitive Learning Fills is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Quantile Regression Forests is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Learning to Rank Finance is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →FT Transformer Tabular is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Tabular Deep Learning Finance is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Neural Architecture Search Caution is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Bayesian Optimization Tuning is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Conformal Prediction Intervals is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why it…
Read →Neural ODE Market Dynamics is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Meta Learning Trading Tasks is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Shapley Values for Features is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Bayesian Model Averaging Trading is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →Group Lasso Factor Selection is a practical quant topic: specify it, estimate it, cost it, and only then allocate risk. What follows is a nofluff working note aimed at systematic researchers who need something they can implement and falsify. Why…
Read →