Synthesis — Quant Trading

The evolving thesis. Spun out of the hub _inbox financial-ml cluster on 2026-06-29, once a third coherent source landed: a financial-market foundation model (kronos-financial-foundation-model, parked first), a reinforcement-learning trading framework (tensortrade, parked second), and an event-driven trading bot (banbot, the trigger that completed the cluster).

Current thesis

A trading system is a stack of three layers, and the founding trio lands one source on each:

1. Predict — Kronos forecasts future OHLCV bars from price history, borrowing the text-LLM tokenize-then-autoregress recipe and retargeting it to the noisy financial modality.
2. Decide (strategy) — turn information into a position, either as a learned policy (TensorTrade’s RL agents) or a hand-coded rule set (the strategies banbot runs).
3. Execute — banbot‘s event-driven engine places and manages the orders against live exchanges, multi-symbol and multi-timeframe.

The through-line binding all three is trustworthy backtesting: a strategy is only as good as the simulation that validated it, and the cardinal failure is lookahead bias — letting a decision use information it couldn’t have had in real time. The shared structural answer is the event-driven design — a forward-only event stream makes peeking at the future impossible — which is exactly why it also enables backtest/live parity (one codebase, two modes). Kronos’s “observe only the past,” TensorTrade’s observer discipline, and banbot’s event loop are three faces of the same constraint.

The shape of the field (early)

• Prediction is not a strategy. A forecaster outputs where the market goes; it still needs a decision layer to size positions and an execution layer to place orders. The three layers are genuinely distinct jobs, not redundant.

• Learned vs. coded strategy — now a three-way split (2026-07-26). RL discovers a policy from reward; a bot runs a policy a human wrote; and supervised-learning-trading fits a policy to labels someone manufactured, which Koshtenko’s XGBoost robot brings into the corpus. The third is the most common form in practice and the least examined, because its hard part hides in plain sight: markets have no labels, so the label is invented (his is “would a 300-point-stop / 800-point-target bracket have won?”). Pick a different bracket and the model learns a different game off identical features. RL’s difficulty is credit assignment and it’s visible; supervised learning’s difficulty is the label definition and it looks like preprocessing.

• Search dominates compute. Tuning a strategy means running the backtest hundreds of times inside a hyperparameter-optimization loop (banbot ships Bayesian / TPE / CMA-ES), so backtest speed caps strategy search — hence banbot’s “1 year of klines in seconds” pitch. Backtest speed even picks the optimizer regime: cheap backtests favor model-free population methods (cma-es), expensive ones favor sample-efficient model-based search (bayesian-optimization).

• Evaluation isn’t only automated — the human-discretionary edge (2026-07-15, smolinski-h1-2026-aar). The founding corpus evaluates a strategy programmatically (a backtest, TradeMaster’s eval toolkit, a tuning loop). A discretionary options trader running a quarterly after-action-review does the same job — validate what’s working, decide what to pivot — by hand, on a live book. This is the spoke’s first source where the evaluation layer is a human narrative review rather than code, and it drags in market-analysis vocabulary the systems never surfaced: market-regime-analysis (adapt to the environment, not a running trend) and a benchmark-panel comparison (SPY/QQQ/IWM/TLT/GLD). It’s the concrete first step of the domain’s stated growth edge — general quantitative-finance practice beyond automated systems — even though it pages no trading system. The AAR’s honesty risk (over-reading one episode) is the discretionary cousin of lookahead bias; Smolinski flags it himself on the dip-buying slide.

• Backtest speed is a design axis, not just a benchmark (2026-07-27). The field splits its backtest engines two ways — event-driven versus vector-based awesome-systematic-trading — and the split is the compute point above turned into an architecture choice. Vectorizing over whole arrays buys thousands of variants per second and gives back the anti-lookahead guarantee and backtest/live parity that the event stream provides structurally. So the corpus’s “search dominates compute” bullet has a second answer beside “make each backtest fast”: change the programming model and accept a weaker simulation for the sweep. The field’s weight still sits with event-driven — roughly nineteen catalog entries to three.

• The corpus finally has a “who,” and it arrives as a hiring filter (2026-08-01, quant-bible-mit-sloan). Every other source here is a system someone built in the open. The firms that actually run capital publish nothing, so the only substantial thing they emit is their quant-interview — which means this wiki’s first view of the institutions is refracted through what they ask twenty-year-olds. Two things show through the refraction. First, the hiring idiom is statistics, not machine learning: regression, econometrics, causal inference and stochastic processes, with Elements of Statistical Learning as the canon and nothing asking for a neural network. That is a real distance from the corpus’s software edge (RL agents, boosted trees, transformer forecasters) — partly the source’s 2022 vintage, partly the difference between what survives a whiteboard and what runs in production. Second, the skill being screened for is pricing under ignorance — commit to a number, state the assumptions, revise as evidence arrives — which is the discretionary form of the same loop backtesting automates and after-action-review runs by hand.

• Market making is an obligation, not a prediction (2026-08-01). market-making was previously one entry in a strategy taxonomy; the guide makes it a job — stand ready to buy or sell at your own quoted prices, and be paid for it in spread. That reframes the strategy layer of the founding thesis. The other strategy families take a view and wait; a market maker takes whatever inventory the flow hands it, which makes its central problem adverse selection (the counterparty trades when they know more) and its central control the skew of the quote against its own position. It’s the first source here where the quantity being modelled is the price formation process rather than the price. Note the width tension it introduces, which has no analogue elsewhere in the corpus: a wider quote is more likely to be right and less likely to trade, so correctness and competitiveness pull against each other continuously.

• Data scarcity is the constraint under the ML layer. There is one price history, it arrives one bar at a time, and the regime that produced it keeps moving — so the corpus’s ML sources all manufacture more (synthetic-market-data): Koshtenko augments 1,000 hours into 5,000 examples and wants GANs for 25 years he doesn’t have; trademaster ships Market-GAN. The catch is that a generator trained on your history produces more of your history — it widens the sample, not the distribution, and distribution shift is what kills strategies live.

Open questions

• Do they compose? The trio implies a pipeline — Kronos-style forecasts as features for a TensorTrade policy executed by a banbot-style engine — but no founding source wires the layers together. Partial evidence (2026-06-30, trademaster): TradeMaster (NTU) is a single ecosystem spanning data → market simulation → RL policy → evaluation, so within one platform the layers do compose; it even folds in a foundation-agent (FinAgent) and synthetic data (Market-GAN). Still an integrated RL platform rather than a Kronos+TensorTrade+banbot mash-up, but it shows the stack is buildable end-to-end.

• Where’s the evidence? The founding trio are author-published (T3) with self-reported performance and no independent benchmark — a neutral results source is the top gap. Moved, not closed (2026-06-30, trademaster): TradeMaster adds a systematic evaluation toolkit (six axes, 17 measures, PRIDE-Star) and published-paper backing for its algorithms — structured, reproducible evaluation rather than a single number. It’s still first-party (T3), not a neutral third-party benchmark, so the gap narrows from “no rigor” to “no independent referee.” Sharpened from inside the corpus (2026-07-26): Koshtenko reports 64–65% accuracy and then discounts his own number — “I do not really trust homemade Python testers” — preferring a vendor tester on real ticks, and ships no live validation at all. It’s still a T3 self-report, but it’s the first one where the author names the evidence problem this question is about instead of reporting past it. The pattern across the corpus is now explicit: every performance figure here comes from a simulator its own author wrote, and the one author who says so out loud is the most credible of them. A different evidence habit (2026-07-27, awesome-systematic-trading): its strategy tables carry a paper, a Sharpe ratio, a volatility, a rebalancing frequency and a runnable QuantConnect implementation per entry, sorted by Sharpe — and they keep the losers (Short Term Reversal with Futures at -0.05). That’s not the independent referee this question wants; the Sharpes are the papers’ own backtests, and the catalog’s owner paperswithbacktest sells access to the strategy database. But reporting a negative result rather than curating winners is the first thing in this corpus that behaves like evidence instead of a pitch. Answered from outside the corpus (2026-08-03, pseudo-mathematics-financial-charlatanism — the spoke’s first T1): the question was framed as needing a referee — someone neutral to re-run the numbers. The paper says the referee is not the missing piece. Selecting a strategy by searching N of its configurations inflates the best in-sample Sharpe by itself: seven independent trials reach an expected Sharpe of 1 on two years of data whose true out-of-sample Sharpe is zero. So a self-reported Sharpe is not weak evidence pending verification — without N it is uninterpretable, because the identical number means “edge” at N=1 and “nothing” at N=45. That reframes every T3 figure in this wiki. The gap is no longer “no independent referee” but no disclosed trial count, and it is the cheaper of the two to close: the authors ask only that a backtest report N, which every source here could do and none does. strategy-optimization is where this corpus manufactures N — a hyperparameter loop over the backtest — so the spoke’s own optimization page and its evidence problem turn out to be one subject. See backtest-overfitting.

• Beyond crypto. Partly answered (2026-06-30, algorithmic-trading-wikipedia). Algorithmic execution is already the norm in established markets — ~50% of US equity volume and ~80%+ of FX orders are algorithmic — so the crypto-centric founding bots are a recent entrant, not the center of the field. The strategy taxonomy (arbitrage, pairs/statistical arb, mean reversion, market making, momentum, event arb) is now named on algorithmic-trading. Residual gap: equities/FX market microstructure, risk, and portfolio construction as their own subjects are still unsourced. The strategy families get their first depth (2026-08-03, chan-algorithmic-trading): mean-reversion arrives with its statistical apparatus — ADF, Hurst, variance ratio, half-life for stationarity; CADF and Johansen for cointegration; linear, Bollinger and Kalman for trading it — across equities, ETF pairs and triplets, currency pairs and futures spreads. Until now the taxonomy on algorithmic-trading was a list of names. Momentum stays a name: two chapters to mean reversion’s four, and the excerpt cuts off on its main theme. Mapped, still unsourced (2026-07-27, awesome-systematic-trading): the catalog gives that gap an inventory — portfolio optimization (PyPortfolioOpt, Riskfolio-Lib, Deepdow), pricing (tf-quant-finance, FinancePy, QuantLib), risk analytics (pyfolio) — plus the incumbent engines the spoke never paged (zipline, backtrader, QuantConnect Lean, vnpy, nautilus_trader, vectorbt, Freqtrade). Names and a reason to page them, not yet pages. The failure catalogue arrives (2026-08-03, chan-algorithmic-trading): the question above asks for an independent referee and keeps getting self-reports. Chan supplies the other half — a named inventory of why a correct backtest still lies: data-snooping, survivorship, primary vs consolidated quotes, venue-dependent FX quotes, short-sale constraints, futures continuous-contract construction, closing vs settlement prices, and regime shift (backtesting). Six of the seven are properties of the data, so the spoke’s engine-side sources address none of them. This doesn’t produce a referee; it explains what the referee would have to check. Microstructure opened (2026-08-01, quant-bible-mit-sloan): market-making brings spread, inventory, order flow and adverse selection in as subjects rather than frictions — the first real dent in this gap. It’s pedagogy, though: the guide teaches quoting on baseball wins and national populations, not on equities. Still unsourced: a production quoting system, exchange market-maker programs, latency, and risk/portfolio construction.

• Does simple beat complex? (added 2026-08-03) chan-algorithmic-trading argues linear strategies are the antidote to overfitting; the rest of the corpus builds transformers, RL policies and boosted ensembles. Neither side offers a comparison. What would settle it is the same thing question 2 asks for — out-of-sample results on the same instruments, scored by someone who didn’t build either. Note the source is 2013, which is a real handicap for its side of the argument and not a disqualification: the pitfalls it names are structural, not vintage.

Growth edges

Ranked; each names the kind of source that would close it (see ../QUALITY.md → Growth edges).

1. A source that reports its N. The evidence question is now specific: no strategy or engine here discloses how many configurations were searched. — needs: any T1/T2 backtest or engine doc that states its trial count. Half closed 2026-08-08 (research pass): deflated-sharpe-ratio (Bailey & López de Prado, JPM 40(5), T1) supplies the correction the edge named, and sharpens why the omission matters — the trial count is an input to the statistic, so a backtest published without N cannot be deflated at all and is uninterpretable rather than merely thin. The half still open is the original ask: a published strategy or engine that states its own N.
2. The execution layer has no independent measurement. Slippage, fill quality and latency are asserted by the engines themselves. — needs: T1/T2 market-microstructure or execution-quality study.
3. Nothing on risk or portfolio construction. The corpus stops at “does the strategy make money” and never reaches position sizing, drawdown control or capital allocation. — needs: T1 text or paper on portfolio risk management.

Coverage edges (added 2026-08-08, at the curator’s request for a wider backlog). These widen what the spoke covers instead of answering an open question above; one ordinary solid source closes any.

4. Derivatives. Half closed 2026-08-11. The original paper is in: black-scholes (T1, JPE 81(3), 1973) with option-pricing as the concept node — the hedge argument, the seven ideal conditions, the formula, delta as hedge ratio, and implied volatility as the inversion the market actually quotes. The half still open is the Greeks beyond delta. Gamma, vega, theta and rho are named nowhere in the corpus except as a feature in staskh-trading-skills. — needs: a derivatives text chapter, or a source measuring a hedging practice rather than listing the quantities.
5. Market microstructure. market-making describes a role inside a venue the corpus never describes: the limit order book, order types, queue position, Reg NMS. — needs: a microstructure text or an exchange’s own matching-engine documentation.
6. The anomaly literature. backtest-overfitting and pseudo-mathematics-financial-charlatanism are critiques of a body of published factor research — Fama-French, momentum — that the spoke does not hold. — needs: the original factor papers, T1.
7. Data and its defects. Survivorship bias, point-in-time fundamentals and corporate-action adjustment silently decide every backtest here. — needs: a paper quantifying one of these biases, or a data vendor’s methodology.

The marketing layer arrives, as a specimen

ai-trading-bot-platforms-2026 is the first source here about the retail platform market — products that sell automated trading to people who will never write an event loop — and it is T4: a ranked roundup whose byline is “Marketing/IT Dept.”, in which the top pick carries every link in the piece with a tracking parameter attached and the other four platforms carry none, with no disclosure and nothing tested.

Ingested rather than refused, because two things in it are worth keeping.

The category map. Five products named and described (3Commas, Pionex, Cryptohopper, Trade Ideas, SaintQuant), all unverified, plus a control axis that is genuinely the right way to cut the category: managed → configurable → exchange-native → signal-only (trading-bot-platform). It maps onto this wiki’s own predict/decide/execute stack — a managed platform sells all three as one product, a configurable one sells execution plus a toolkit, a signal-only one sells prediction and leaves the rest with you. banbot and tensortrade are the open-source form of the middle.

The vocabulary gap, which is the actual finding. Nothing in the article mentions Sharpe ratios, out-of-sample testing, transaction costs, slippage or capacity. Not hedged, not glossed — absent. Set that beside pseudo-mathematics-financial-charlatanism, which shows that seven configurations over two years manufacture an in-sample Sharpe of 1 from a strategy with no edge, and the shape of the retail market becomes legible: the things that decide whether an automated strategy works are not the things it is sold on, and a managed platform’s risk rules are precisely the part a buyer cannot inspect.

That extends the corpus’s standing thread rather than opening a new one. The tension section below already holds “nothing in this corpus can tell you whether a reported edge is real” — this source is the case where nobody even reports one, and it still ranks five products. Its most accurate sentence is its disclaimer: “No bot guarantees profit.”

The corpus’s oldest source is the only one that reports where it fails

black-scholes arrives as coverage — the spoke could not describe most of what professional desks trade — and what it actually supplies is a standard of conduct the rest of the corpus does not meet.

It prices without a backtest. Everything else here settles “does this work” by replaying history in a simulator the author wrote (backtesting). Black and Scholes derive a value from a no-arbitrage condition: a continuously adjusted hedged position carries no risk, a riskless position must earn the short rate, and one formula satisfies that constraint. No history is consulted. Options are, so far, the only instrument in this wiki where a claim about value can be checked against theory instead of against a simulation, which is what option-pricing adds to the predict→decide→execute stack rather than merely beside it.

It states its assumptions as a list. Seven ideal conditions, numbered, at the point of use — constant rate, lognormal prices with constant variance, no dividends, European exercise, no transaction costs, unlimited borrowing, free short selling. Every one is false somewhere, and saying so is what lets a reader decide whether the formula applies to their instrument. No backtest in this corpus ships that list. fincept-terminal ships factor discovery, ML and RL with no statement of conditions at all.

And it reports a negative result about itself. The authors’ own empirical test found market prices deviating systematically from their formula — buyers consistently overpaying, the gap widest on low-risk stocks — and then closed the loop: given transaction costs, the deviation “does not imply profit opportunities for a speculator.” They found their model wrong in a specific direction, and they published that the error was not tradeable.

Fifty-three years later this wiki holds superior-skills validating a strategy on six trades at a 100% win rate, and ai-trading-bot-platforms-2026 ranking five products with no mention of Sharpe, out-of-sample testing, costs or capacity. The corpus’s standing worry has been that performance numbers come from simulators their authors wrote. The sharper version, visible only now that there is something to compare against: not one modern source here reports a negative finding about its own method. deflated-sharpe-ratio supplies the statistic that would expose the problem and chan-algorithmic-trading argues for the model class that resists it; this is the older and simpler discipline both are reaching for — say what you assumed, then say where it broke.

One honest caveat about the comparison. A closed-form pricing result and a machine-learning trading strategy are not the same kind of object, and a paper in the JPE answers to reviewers that a GitHub README does not. The point is not that the repositories should have derived a theorem. It is that stating conditions and publishing a failure are cheap, and nobody does them.

Contradictions / tensions

Three of the four below turned out to be the same argument seen from different layers, so read them as one thread: nothing in this corpus can tell you whether a reported edge is real. The first names the methodological fork (does simplicity or validation protect you?), the second shows what the fork costs once an LLM can place the order, and the third shows the most prominent artifact in the spoke bundling every layer and reporting discipline for none.

Complexity as the disease, not the cure

The corpus’s first genuine internal disagreement, and it is about method rather than a fact.

Everything in this spoke until now bets on more model: kronos-financial-foundation-model is a pretrained transformer over OHLCV bars, tensortrade and trademaster learn policies with deep RL, supervised-learning-trading fits gradient-boosted classifiers to manufactured labels, multithreaded-ml-trading-robot runs an ML ensemble, and strategy-optimization exists to search parameter spaces hard. The shared assumption is that capacity plus a validation protocol beats simplicity.

chan-algorithmic-trading states the opposite as its organizing principle: simple, linear strategies as an antidote to overfitting and data-snooping, with explicit scorn for “the usual ‘data-mining’ approach” of throwing indicators at a series to see which ones profit. On this view the guard against overfitting is the model class, not the holdout — you cannot fit a two-parameter spread rule to noise the way you can fit a boosted ensemble to it.

Nothing in the corpus adjudicates this, and the two sides are not evenly matched on evidence: Chan is a practitioner asserting a preference in a preface, and the ML sources are self-reported backtests. But it is a real fork, and the spoke’s own evidence problem is what makes it live — if every performance number here comes from a simulator its author wrote, then the argument “complex models overfit and you won’t be able to tell” is not a stylistic objection, it is a prediction about exactly this corpus. Note also the date asymmetry: Chan is 2013, before the deep-learning wave the rest of the corpus rides.

The agent became the deployment surface

Not a contradiction — a new axis, and one that changes what the evidence problem costs.

superior-skills packages strategy authoring, backtesting, funding, live deployment, sub-account orchestration, withdrawal and a kill switch as Agent Skills, so the whole loop from idea to filled order runs through a conversation with an LLM (agent-deployed-trading). Every other artifact here is a library a programmer drives. A “managed wallet” option is described as “no-key trading, custodial-style UX.”

The spoke’s standing worry has been that performance numbers come from simulators their authors wrote. Agent deployment adds a second: the last human check is an approval prompt, and the sibling ../agentic-tooling-wiki has spent two months collecting evidence about what such prompts are worth — including a shipping harness conceding that an approval means a human accepted the action as displayed, not that the behaviour is safe. Those findings concern bad commits. Here a mistake is a filled order: a trade has no diff to inspect, and exit-all closes open positions rather than unwinding closed ones.

The commercial shape is the part to keep. The skills are MIT and installable from five entry points; they are inert without an API key to a paid platform. So an open skills pack works as a customer-acquisition channel — publish the interface, keep the backtester, the data and the execution — and the same party then writes the strategy, runs the backtest validating it, and earns on the deployment. Which is visible in the artifact: the repo’s headline “validated” strategy is six trades, 100% win rate, 0% drawdown (backtesting).

That number arrived hours after chan-algorithmic-trading gave this spoke data-snooping as a named worry, and it is the cleanest specimen the corpus is likely to get. It does not settle the complexity-versus-simplicity fork above — the Donchian rule is exactly the simple, linear kind Chan argues for — which is itself the useful correction: a simple model is not a defence against a sample of six. Chan’s antidote addresses the model class; nothing addresses the evidence.

The terminal bundles every layer, and reports discipline for none

fincept-terminal is the first source that is not one layer of the stack but all of them at once — a Bloomberg-class terminal wrapping data connectors, a QuantLib pricing/risk suite, factor discovery, an ML/HFT/RL “Quant Lab,” a paper-trading engine, and live execution through 16 brokers, behind one UI. Its size (29k★) makes it the most prominent single artifact the spoke holds.

It does not adjudicate the complexity-vs-simplicity fork above; it stacks the deck for one side and stays silent on the cost. Everything Chan warned against — searching many indicators, fitting high-capacity models, mining factors — is a menu item here, and the README describes capabilities without a word on validation: no MinBTL, no held-out protocol, no reported trial count N. So it extends the spoke’s central evidence problem rather than easing it. The backtest-overfitting worry was that a searched-over Sharpe hides N; a terminal makes the search a few clicks and still does not surface N. It is superior-skills‘s “6 trades, 100% win” failure shape on a far larger surface, and the useful synthesis is narrow: tooling that lowers the cost of searching configurations, without lowering the cost of reporting how many, moves the field the wrong way on its own hardest problem.

The other thing it marks is the agent-as-terminal turn. agent-deployed-trading arrived through superior-skills as a strategy library run through an LLM; here it is 37 investor-persona agents (Buffett, Graham, Lynch…) with MCP-tool node workflows sitting inside a full terminal. The deployment surface is no longer a skill pack but an application, which raises the same “the last human check is an approval prompt” concern at product scale. The agent machinery itself is agentic-tooling-wiki’s subject (logged runner-up); what stays here is that a trade authored by a persona agent still has no diff to inspect.

Two skillpacks, and the incentive question that separates them (added 2026-08-06)

staskh-trading-skills is the second agent-skills pack for trading to reach this spoke, and it arrived precisely as a control on what the first one let us conclude. agent-deployed-trading had read superior-skills as an open pack fronting a paid platform — publish the interface, keep the backtester, the data and the execution. staskh is MIT too, and needs no platform: Yahoo Finance for data, the user’s own Interactive Brokers account for the portfolio, one optional paid feed the author does not sell. Nobody earns on the deployment.

The correction is small and worth making anyway: openness says nothing about incentives; what the pack needs a key for does. One case looked like a business model, and it was — its own, not the category’s.

The more interesting result is that the two packs fail on opposite sides of the spoke’s central problem. Superior ships strategies and validates one on six trades. staskh ships no strategies and no backtester at all — 25 skills that pull quotes, compute Greeks and screen for PMCC setups, with nothing between a scanner hit and a position but the trader’s judgement. The first is backtest-overfitting in miniature. The second is not an improvement on it: an evidence problem you cannot have because you never make a claim is the same gap with the paperwork removed, and the agent’s fluency about RSI and delta does the persuading that a backtest at least pretended to.

The control boundary moved too, and this is the concrete part. staskh defaults Interactive Brokers to read-only, requiring an explicit switch to read-write before orders can be placed. Better than Superior’s custodial “no-key trading,” and still not review: it is one decision, made once in advance, that covers every order afterwards. The two properties agent-deployed-trading named survive both designs — a trade has no diff, and the kill switch is downstream of the mistake.

One caveat neither pack’s README treats as a design constraint: staskh’s data is Yahoo’s, ~15 minutes delayed, and it is used to compute Greeks and unusual-options-activity scans — the parts of options trading most sensitive to the current underlying price. Fine for study, not for the execution the same feature list implies.

The standing external tension

The older note still holds: self-reported performance vs. the discipline’s well-known overfitting and distribution-shift risks — now reinforced by smolinski-h1-2026-aar, whose +33% vs 11% S&P is a trader’s own unaudited number reported second-hand (no drawdown/vol context), the same self-report caution the founding bots earn.

Cross-spoke adjacency

• ../optimization-algorithms-wiki — owns the optimizers as the subject; here they’re a means to tune a strategy. strategy-optimization is the bridge node: it maps banbot’s tuner menu onto that spoke’s model-free (cma-es, metaheuristic-optimization) ↔ model-based (bayesian-optimization) axis, and borrows its no-free-lunch-theorem (no single best tuner) and exploration-vs-exploitation framing. That spoke explicitly excludes algorithmic-trading strategies, which routes them here. Cross-link the algorithm, keep the trading substance here.
• ../llm-providers-wiki — owns the text/multimodal-LLM provider market and the open-weight lineage Kronos descends from. The financial OHLCV modality keeps Kronos here; cross-link the foundation-model pattern.
• ../agentic-tooling-wiki — owns LLM agents; TensorTrade’s are RL agents — same word, different method.
• ../machine-learning-wiki — owns classical ML as an engineering discipline. The MIT guide‘s data-science half (least squares vs. kNN, bias/variance, ridge / lasso / elastic-net, stepwise selection) is that spoke’s subject, taught here as a hiring prerequisite; it was routed here whole on its dominant substance rather than split. Its causal-inference material — selection bias, randomization — is unclaimed by either spoke and worth watching.
• ../psychology-wiki — daniel-kahneman lives there; the guide recommends Thinking, Fast and Slow for “trader thinking styles,” the same borrow-the-metaphor pattern that spoke already tracks.

Index — Quant Trading Wiki

Catalog of every page, grouped by schema.org @type. Spine: synthesis (thesis), log.md (history), this file (catalog). Spun out of the hub _inbox financial-ml cluster on 2026-06-29 (3 sources).

DefinedTerm (concepts)

• algorithmic-trading — the umbrella: software trades markets across a predict → decide → execute stack · concept
• mean-reversion — the bet that a price or a constructed portfolio returns to a level; the testable-before-you-build camp (ADF/Hurst/variance-ratio/half-life, CADF/Johansen; linear/Bollinger/Kalman), four of Chan’s eight chapters · strategy-family
• backtesting — replaying a strategy on history; the lookahead-bias hazard and backtest/live parity · concept
• event-driven-trading — forward-only event stream; the structural guard against lookahead · concept
• agent-deployed-trading — the LLM agent as the surface that authors, backtests and puts live a strategy; the last human check becomes an approval prompt, and a trade has no diff to inspect · practice
• trading-bot-platform — the commercial layer above the corpus: consumer products selling automation on a control axis (managed → configurable → exchange-native → signal-only). The vocabulary that decides whether a strategy works — Sharpe, out-of-sample, slippage, capacity — appears nowhere in how they are sold · product-category
• reinforcement-learning-trading — the learned-policy strategy layer (agent / action / reward / observer) · concept
• financial-time-series-foundation-model — pre-trained OHLCV forecasters; the predict layer · concept
• strategy-optimization — tuning a strategy’s params on backtested fitness; the bridge to ../optimization-algorithms-wiki (CMA-ES / Bayesian / TPE) · concept · cross-spoke bridge
• after-action-review — AAR: a structured post-period trading review (what worked / didn’t / change); the evaluation layer done by a human, the discretionary cousin of backtest/eval · concept
• supervised-learning-trading — labelled-bar classifiers as the decide layer (XGBoost/ensembles); the label is manufactured, and manufacturing it is the strategy · concept
• synthetic-market-data — augmentation + GAN-generated history against financial data scarcity; widens the sample, not the distribution · concept
• vectorized-backtesting — array-at-once backtesting: thousands of variants per second, at the cost of the anti-lookahead guarantee and backtest/live parity · concept
• market-making — quoting both sides on demand: the spread as pay for an obligation, inventory skew, and adverse selection; the strategy the others trade against · concept · microstructure
• quant-interview — the field’s hiring filter (brainteasers, pricing games, research cases) and the near-only artifact the firms emit · concept
• quant-trading-firm — the institutions: prop shops / quant funds / bank desks, and the QT / QR / SWE role split · concept
• market-regime-analysis — segmenting history into distinct market environments (vol / correlation / leadership) and adapting to the regime, not a running trend; a market-state concept · concept
• backtest-overfitting — selecting a strategy by searching N of its configurations inflates its in-sample Sharpe with no edge present; the trial count N is what makes a backtest interpretable, and nothing here reports it · theory
• option-pricing — the layer the predict→decide→execute stack does not have: a derivative has a theoretical value derived from the underlying, so a trade can be wrong on price without being wrong on direction. Five inputs, four observable, and the fifth (volatility) is what an options market actually quotes; delta as black-scholes‘s hedge ratio, and the rest of the Greeks recorded as a gap · concept · derivatives

Article (sources)

• ai-trading-bot-platforms-2026 — InnoTech Today roundup of five retail bot platforms (3Commas, Pionex, Cryptohopper, Trade Ideas, SaintQuant): the spoke’s first source on the retail market and its first specimen of the marketing layer — byline “Marketing/IT Dept.”, tracking parameters on the top pick’s links alone, no disclosure, nothing tested. Useful for the category map and the control axis, worthless as a ranking · source · T4 · innotechtoday.com
• algorithmic-trading-wikipedia — encyclopedic overview: scale (≈50% US equity / ≈80%+ FX volume), strategy taxonomy, the backtest→forward→live validation ladder; the spoke’s neutral anchor · source · T2 · en.wikipedia.org
• smolinski-h1-2026-aar — BI profile of a discretionary options trader’s 19-slide H1-2026 AAR deck: regime analysis, broadening leadership, dip-buy speed, a benchmark-panel prescription (SPY/QQQ/IWM/TLT/GLD); self-reported +33% vs 11% S&P; the spoke’s first human-discretionary evaluation source · source · T2 · businessinsider.com

TechArticle (sources)

• multithreaded-ml-trading-robot — MQL5 (Koshtenko, 2026-07-23): Python-threaded 5-pair FX robot over MT5 — XGBoost+Bagging on RFE-selected indicators, GMM regime clustering, augmented training data, ATR sizing against a portfolio risk budget; 64–65% test accuracy, no live validation, and the author distrusts his own tester · source · T3 · mql5.com

SoftwareSourceCode (sources)

• kronos-financial-foundation-model — first open-source financial-market foundation model: OHLCV tokenizer + decoder-only Transformer · source · T3 · github.com/shiyu-coder
• tensortrade — Python RL framework for trading agents (Gym-compatible; Ray RLlib) · source · T3 · github.com/tensortrade-org
• banbot — event-driven multi-symbol/strategy/timeframe/account trading bot; backtest + live, Go core · source · T3 · github.com/banbox
• staskh-trading-skills — 25 SKILL.md skills making Claude an options-trading advisor (MIT, 324★): quotes, chains, Greeks, RSI/MACD/Bollinger, PMCC and whale scanners, Interactive Brokers portfolio reads, PDF reports; Yahoo data ~15 min delayed. The control case for superior-skills — same artifact class, no platform behind it, and no backtester and no strategies, so no performance claim and nothing between a scanner hit and a position · source · T3 · github.com/staskh
• superior-skills — Superior Trade’s Agent-Skills library for Hyperliquid/Lighter/Polymarket (MIT, 225★): 8 strategy templates + primitives (regime gate, ratcheting exit, fee optimizer, pre-trade thesis). Its headline “validated” strategy is 6 trades / 100% win / 0% drawdown — the corpus’s clearest data-snooping specimen; open skills as a channel to a paid API · source · T3 · github.com/Superior-Trade
• trademaster — NTU’s holistic RL-for-quant-trading platform: 6 modules (data/sim/13+ algo zoo/eval), 5 tasks (portfolio→market-making); the heavyweight to TensorTrade’s lightweight · source · T3 · github.com/TradeMaster-NTU
• fincept-terminal — Fincept Corp, ~29.6k★, C++/Qt: an open-source Bloomberg-class terminal bundling data connectors + QuantLib pricing/risk + factor-discovery/ML/HFT/RL Quant Lab + paper-trading + live execution via 16 brokers + 37 investor-persona AI agents. The first whole-terminal source; ships every overfitting tool and reports no validation discipline · source · T3 · github.com

Collection (sources)

• awesome-systematic-trading — the field’s map: 97 libraries (event-driven vs vector-based backtesters, bots, analytics/risk/pricing, data, ML), 40+ paper strategies with Sharpe + QuantConnect code, 55 books; ~8.8k★ but unpushed since Jan 2025 · source · T3 · github.com/paperswithbacktest

LearningResource (sources)

• quant-bible-mit-sloan — 51-page student-written quant-interview study guide (MIT Sloan Business Club, 2022): probability/stats, regression & econometrics, three research cases, market-making theory + worked games, and a per-firm question bank across ten shops; the spoke’s first source on the people and institutions rather than the code · source · T3 · raw/quant-bible-mit-sloan.pdf

Organization (entities)

• paperswithbacktest — org/site behind the awesome list; free catalog as funnel, paid strategy database as product · entity
• innovation-and-tech-today — NewsMediaOrganization: consumer tech outlet; publisher of ai-trading-bot-platforms-2026, which it ran under a “Marketing/IT Dept.” byline with undisclosed tracking links · entity · publisher
• fincept-corporation — org behind fincept-terminal; custom NOASSERTION licence on a 29k★ project · entity
• banbox — GitHub org behind banbot; AGPLv3 + commercial dual-license · entity
• nanyang-technological-university — NTU Singapore; its AMI group maintains trademaster · entity · CollegeOrUniversity
• mit-sloan-business-club — MIT student club; publisher of quant-bible-mit-sloan · entity
• massachusetts-institute-of-technology — MIT; a named feeder into the quant industry, with a course-road mapped to the quant-interview · entity · CollegeOrUniversity
• lawrence-berkeley-national-laboratory — US DOE national lab; affiliation of two authors of the spoke’s first T1 · entity
• american-mathematical-society — AMS; publisher of the Notices where that paper appeared · entity

ScholarlyArticle (sources)

• deflated-sharpe-ratio — Bailey & López de Prado, Journal of Portfolio Management 40(5):94–107, 2014: deflate a reported Sharpe by the expected maximum under the null given the trial count, sample length and return skew/kurtosis. Corrects selection bias under multiple testing and non-normality together. The operative consequence for this wiki: N is an input, so a backtest that never states how many configurations it searched cannot be deflated by anyone · source · T1 · ssrn.com
• pseudo-mathematics-financial-charlatanism — Bailey/Borwein/López de Prado/Zhu, Notices of the AMS 61(5) 2014: seven searched configurations reach an expected in-sample Sharpe of 1 on two years of data with zero true edge; derives Minimum Backtest Length and demands backtests report N — the spoke’s first T1 and first peer-reviewed source · source · T1 · ams.org
• black-scholes — Black & Scholes, Journal of Political Economy 81(3):637–654, 1973: the European call priced off a continuously adjusted hedged position that must earn the riskless rate. Seven “ideal conditions” stated up front; the expected return on the stock does not appear in the formula, which is what makes it usable. Their own empirical test reports systematic deviations — buyers overpay, and transaction costs make the deviation untradeable · source · T1 · sfu.ca (JSTOR scan)

Book (sources)

• chan-algorithmic-trading — Ernest Chan, Wiley 2013 (publisher excerpt: front matter + preface, 15pp): mean reversion in four chapters and momentum in two, built on “simple and linear strategies, as an antidote to the overfitting and data-snooping biases” — the corpus’s first practitioner book, its first internal disagreement, and its best inventory of why a correct backtest still lies · source · T2 · assets.thalia.media
• john-wiley-and-sons — Organization: publisher of the Wiley Trading series, whose own blurb — books by traders who “have survived… and have prospered” — states a survivorship filter as editorial policy · publisher

Person (entities)

• yevgeniy-koshtenko — Kazakhstan-based algo trader publishing MQL5 build logs; unusually legible about his own weaknesses · entity
• ernest-chan — quant trader and author (2013); writes for the serious retail trader, prefers statistics to machine learning, and reports from running a fund that strategy performance itself mean-reverts · author
• erik-smolinski — full-time discretionary options trader; practitioner behind the smolinski-h1-2026-aar AAR deck; the spoke’s first human-trader node · entity
• marcos-lopez-de-prado — quant practitioner (Guggenheim) and peer-reviewed critic of practitioner evidence; behind the Deflated Sharpe Ratio · author
• david-h-bailey — LBNL mathematician; co-author of the spoke’s first T1 · author
• jonathan-borwein — experimental mathematician (1951–2016); co-author · author
• qiji-jim-zhu — Western Michigan mathematician working in optimization and mathematical finance; co-author · author

Synthesis

• synthesis — the thesis: a trading system is a predict → decide → execute stack, bound by trustworthy, lookahead-free backtesting

Notes

• The TensorTrade owner-org and Kronos’s author (shiyu-coder) entity nodes are deferred — thin at founding; page them when one recurs or a query needs it. banbox paged as the trigger source’s owner.
• For smolinski-h1-2026-aar, the subject/practitioner erik-smolinski is paged (central); the publisher (Business Insider) and author (Kathleen Elkins) are deferred — thin/evidence-only on a borderline single-source ingest; page them if a second BI/Elkins source lands.
• For quant-bible-mit-sloan, the named student contributors (Evan Vogelbaum, Ravi Raghavan, Guang Cui, Kyri Chen, Brian) are deferred — private individuals, first-name-only in places, and the source gives no evidence about them beyond section bylines. The publisher mit-sloan-business-club and massachusetts-institute-of-technology carry the entity weight. The firms named in it (Jane Street, Citadel, Optiver, SIG, HRT, Two Sigma…) are deliberately not paged individually — the source supports a landscape, not twenty organization nodes; they live on quant-trading-firm until one earns a page.