MQ"> MRKT QUANT | Quantitative Trading Intelligence
ENGINE MRKT FIELD THEORY Σ.1
PRICE ACTION AS EVOLVING GEOMETRY
FIELD EXCITATION
Φ-SPACE_1 × Φ-SPACE_2 → MRKT FIELD
STRUCTURE FORMS · EXCITES · COLLAPSES
PEAK FIELD ENERGY
QUANTITATIVE TRADING INTELLIGENCE
MQ

We engineer alpha. Proprietary models built from advanced mathematics — stress-tested, systematized, and deployed where the edge actually lives: in structure the market hasn't priced yet.

// DIVISIONS

WHAT WE OFFER

Four divisions, one standard: serious mathematics only. Theory-driven models, autonomous execution, professional terminals, and the research behind them.

Σ
DIVISION 01

MRKT Theory

Five market models built from pure mathematics — rough path signatures, Hawkes cascades, topology, spectral theory, and fractional calculus — decoding structure price-based tools cannot see.

  • Path signature analysis
  • Cascade & criticality detection
  • Topological crash geometry
  • Spectral & long-memory engines
DIVISION 02

HFT Algos

Three autonomous execution systems: the flagship MRKT Theory alpha engine, a compression-breakout hunter, and a precision retracement sniper — each systematic, each risk-managed.

  • VANTA — MRKT Theory alpha engine
  • BREACH — range breakout hunter
  • RECOIL — retracement precision entries
  • Systematic risk management
DIVISION 03

Terminals

Professional-grade workstations: DELPHI aggregates every prediction market into one screen, and MQT-1 is our Bloomberg-class command center with the full theory stack built in.

  • DELPHI — prediction market terminal
  • MQT-1 — MRKT Quant terminal
  • Cross-venue mispricing scanner
  • Theory-stack chart overlays
🔬
DIVISION 04

Research & Development

The theoretical backbone — curated academic research mapped to every model we ship, plus the pipeline of what's coming next out of the lab.

  • Foundational & frontier papers
  • Mapped to each model
  • Free open-access sources
  • Live development pipeline
// DIVISION 01

MRKT THEORY

Market models built from pure mathematics. These tools are streamlined interpretations of our underlying research models — intended for analytical insight, not direct signal replication.

ACTIVE

SIGIL Ψ.1

Path Signature Engine. Converts price paths into signature tensors and scores how strongly the current path rhymes with paths that preceded major moves.

Rough PathsSignature TensorsLead-Lag Structure
ACTIVE

SWARM χ.2

Self-Exciting Flow Model. Models buy/sell pressure as mutually exciting event cascades and tracks the market's live distance from criticality.

Hawkes ProcessesBranching RatioCascade Risk
ACTIVE

ATLAS Ξ.1

Topological Regime Mapper. Tracks the birth and death of topological features in market data — structure forming and collapsing before crashes.

Persistent HomologyCrash GeometryEarly Warning
ACTIVE

SPECTRA Λ.3

Random Matrix Correlation Filter. Separates real cross-asset structure from statistical noise and flags when one macro factor seizes control.

Eigenvalue SpectrumNoise FilteringFactor Dominance
ACTIVE

RELIC η.2

Long-Memory Fractal Engine. Estimates the live Hurst exponent and classifies whether the tape is trending, mean-reverting, or random at each timescale.

Fractional CalculusHurst ExponentMemory Depth
NEW

PYTHIA σ.2

Conditional Quantile Oracle. Forecasts today's session high, low, and close as conditional quantiles — and votes on direction with a five-signal, edge-weighted ensemble.

HAR VolatilityConditional QuantilesSignal Ensemble
Ψ SERIES

SIGIL Ψ.1

Path Signature Engine

Built on rough path theory. SIGIL converts recent price paths into signature tensors — a mathematically complete summary of a path's shape, ordering, and lead-lag structure — then scores how strongly the current path rhymes with historical paths that preceded major moves. Signatures capture nonlinear ordering information that returns-based indicators mathematically cannot see.

  • Signature Extraction — iterated-integral features of the live path
  • Path Rhyme Score — similarity to pre-move historical templates
  • Lead-Lag Structure — captures ordering effects invisible to returns
  • Depth Truncation — tunable signature order for noise control
Rough Path TheoryIterated IntegralsTemplate Matching
// HOW IT’S USED — SIGIL Ψ.1 PLAYBOOK
  1. Load it on your working timeframe. SIGIL continuously converts the most recent price path into a signature tensor — no settings required to start; defaults are tuned for liquid majors.
  2. Watch the Path Rhyme Score. Readings above 0.75 mean the live path's shape closely matches historical paths that preceded major moves. Below 0.4, the tape has no recognizable structure — stand down.
  3. Check the matched template's direction. A high rhyme against bullish pre-break templates is a long setup cue; against bearish templates, a warning. Require agreement with your own bias before sizing up.
  4. Tune Depth Truncation to the asset. Lower signature depth for noisy, choppy instruments; higher depth for clean trending ones. More depth = more path detail, but more noise sensitivity.
High rhyme + rising score = the tape is repeating itself. That is your edge window.
SIGIL Ψ.1PATH SIGNATURE ENGINELIVE// ROUGH PATHS · SIGNATURE DEPTH 4WINDOW 256 · TEMPLATE LIB 12,408SIGNATURE TENSORS⁴PATH RHYME0.87MATCH: PRE-BREAK TEMPLATE ▲LIVE PATH → ITERATED INTEGRALS ∫∫ dX⊗dX
LIVE ON YOUR CHARTS

Test SIGIL Ψ.1 on your charting platform. Customize parameters and integrate it into your workflow.

Χ SERIES

SWARM χ.2

Self-Exciting Flow Model

Built on Hawkes processes. SWARM models buying and selling pressure as mutually exciting event cascades, estimating the market's live branching ratio — how close activity is to criticality, where one trade triggers avalanches of others. It flags the transition from mean-reverting calm into reflexive, cascade-prone states before volatility expands.

  • Branching Ratio — live distance from criticality (n → 1)
  • Excitation Kernel — how long each event's influence persists
  • Cascade Asymmetry — buy-side vs sell-side reflexivity imbalance
  • Endogeneity Score — how much activity is self-generated vs news
Hawkes ProcessesPoint ProcessesCriticality Detection
// HOW IT’S USED — SWARM χ.2 PLAYBOOK
  1. Track the live branching ratio n. Below 0.7 the market digests flow calmly — mean-reversion tactics work. As n approaches 1.0, activity becomes self-feeding and moves extend farther than they “should.”
  2. Treat n crossing 0.85 as a regime alarm. Tighten stops, reduce size, and stop fading strength — near criticality, one aggressive order can trigger an avalanche of follow-on trades.
  3. Read Cascade Asymmetry for direction. Buy-side reflexivity dominating = squeeze fuel above; sell-side dominating = flush risk below. Position with the coiled side, not against it.
  4. Use the Endogeneity Score as a truth filter. High endogeneity means the move is feeding on itself, not on news — those moves reverse hard when the cascade exhausts.
Fade extension only when n is falling. Never fade a tape whose reflexivity is still climbing.
SWARM χ.2SELF-EXCITING FLOW MODELLIVE// HAWKES KERNEL φ(t)=αe^-βtBRANCHING n=0.94 · ENDOGENEITY 88%INTENSITY λ(t) — SELF-EXCITED CASCADESn=1n=0n = 0.94→ CRITICALITY⚠ CASCADE RISKORDER-FLOW EVENTS · PARENT → CHILD EXCITATION
LIVE ON YOUR CHARTS

Test SWARM χ.2 on your charting platform. Customize parameters and integrate it into your workflow.

Ξ SERIES

ATLAS Ξ.1

Topological Regime Mapper

Built on persistent homology. ATLAS embeds recent market data into point clouds and tracks the birth and death of topological features — loops, voids, connected components — as structure forms and collapses. Documented in academic literature as an early-warning signal ahead of market crashes, and almost nobody brings it to retail.

  • Persistence Landscapes — quantifies topological structure over time
  • Feature Lifespan — separates durable structure from noise
  • Collapse Detection — structural degradation ahead of drawdowns
  • Norm Tracking — L¹/L² landscape norms as crash-risk gauges
Persistent HomologyPoint CloudsTopological Data Analysis
// HOW IT’S USED — ATLAS Ξ.1 PLAYBOOK
  1. Run ATLAS on a basket, not a single chart — index components, sector ETFs, or your own portfolio. It maps the market's shape, which no single price series can show.
  2. Watch the landscape norms trend. Rising L¹/L² norms = structure building as correlations organize. A sharp norm collapse after a sustained peak is the documented pre-crash signature.
  3. Treat Collapse Detection as a de-risking trigger. When durable loops start dying, cut gross exposure, buy protection, and widen stops — before the drawdown, not during it.
  4. Ignore short-lived features. Noise dies young; real structure persists. The Feature Lifespan filter is what separates a regime change from a random Tuesday.
ATLAS is your smoke detector. It will not time the top — it tells you the building is flammable.
ATLAS Ξ.1TOPOLOGICAL REGIME MAPPERLIVE// PERSISTENT HOMOLOGY H₀ · H₁L² LANDSCAPE NORM 2.14 ▲POINT CLOUD → RIPS COMPLEXH₁ LOOPPERSISTENTPERSISTENCE DIAGRAMBARCODE H₀ / H₁STRUCTURE STABLE — NO COLLAPSE SIGNATURE
LIVE ON YOUR CHARTS

Test ATLAS Ξ.1 on your charting platform. Customize parameters and integrate it into your workflow.

Λ SERIES

SPECTRA Λ.3

Random Matrix Correlation Filter

Built on random matrix theory. SPECTRA compares the live cross-asset correlation spectrum against the Marchenko–Pastur noise distribution to separate genuine structure from statistical noise, tracking when the largest eigenvalue swells — the signature of a single macro factor seizing control of everything (risk-on/risk-off compression).

  • Eigenvalue Spectrum — live decomposition of the correlation matrix
  • Noise Band Filtering — Marchenko–Pastur bounds strip false structure
  • Factor Dominance — λ₁ absorption ratio as regime-compression gauge
  • Diversification Health — effective number of independent bets
Random Matrix TheoryMarchenko–PasturSpectral Analysis
// HOW IT’S USED — SPECTRA Λ.3 PLAYBOOK
  1. Watch the λ₁ absorption ratio first. Above ~55%, one macro factor is driving everything — your “diversified” book is secretly one trade. Size total risk accordingly.
  2. Trust nothing inside the noise band. Eigenvalues under the Marchenko–Pastur ceiling are statistically indistinguishable from randomness — correlations built on them will betray you.
  3. Size the book off Diversification Health. When effective independent bets drop from 12 to 3, run a third of the risk. The market decides your diversification, not your position count.
  4. Hunt the regime turn. A falling λ₁ after extreme compression is historically where stock-picker markets begin — spread trades and relative value reopen for business.
When λ₁ swells, correlation is destiny: hedge the factor, not the positions.
SPECTRA Λ.3RANDOM MATRIX CORRELATION FILTERLIVE// Q = T/N = 4 · λ± = (1±√(1/Q))²EFFECTIVE BETS 3.2 ▼ · COMPRESSION HIGHMP NOISE BANDSIGNAL →λ₁ABSORPTION 61%ONE FACTOR RUNS THE TAPE0EIGENVALUE λCORRELATION SPECTRUM — LIVE DECOMPOSITION
LIVE ON YOUR CHARTS

Test SPECTRA Λ.3 on your charting platform. Customize parameters and integrate it into your workflow.

Η SERIES

RELIC η.2

Long-Memory Fractal Engine

Built on fractional calculus. RELIC estimates the market's live Hurst exponent and applies fractional differentiation to find the minimum memory-erasure needed to make prices stationary. It classifies whether the tape is trending, mean-reverting, or random at each timescale — and how persistent a move should be before you trust it.

  • Live Hurst Estimation — persistence vs anti-persistence per timescale
  • Fractional Differentiation — stationarity with minimal memory loss
  • Regime Grid — trend / mean-revert / random classification map
  • Memory Depth — how far back the tape still matters
Fractional CalculusHurst ExponentLong Memory
// HOW IT’S USED — RELIC η.2 PLAYBOOK
  1. Check H for your timeframe before choosing tactics. H > 0.55: momentum entries, hold winners. H < 0.45: fade extremes, take profits fast. H ≈ 0.5: the tape is coin-flips — stand down.
  2. Match every strategy to the Regime Grid cell it lives in. Trend systems die in anti-persistent cells; mean-reversion bleeds in persistent ones. The grid tells you which book to run today.
  3. Set lookbacks with Memory Depth. If the market's memory is 340 bars, an indicator tuned to 1,000 bars is measuring ghosts. Calibrate windows to live memory, not habit.
  4. Respect the flip. When H crosses 0.5 on your timescale, retire the old playbook immediately — regime transitions punish slow adapters most.
Persistence is permission: only press winners while the tape remembers.
RELIC η.2LONG-MEMORY FRACTAL ENGINELIVE// FRACTIONAL DIFF d* = 0.42MEMORY DEPTH 340 BARS · REGIME: TRENDSAME NOISE — THREE MEMORIESH = 0.72PERSISTENT · TRENDH = 0.50RANDOM WALKH = 0.28ANTI-PERSISTENT · FADETREND ZONEMEAN-REVERT ZONEH=0.5 RANDOMH(τ)=0.63LIVE HURST BY TIMESCALE1m → 1D
LIVE ON YOUR CHARTS

Test RELIC η.2 on your charting platform. Customize parameters and integrate it into your workflow.

Σ SERIES · NEW

PYTHIA σ.2

Conditional Quantile Oracle

Built on conditional distribution theory. PYTHIA forecasts the session’s high, low, and close as conditional quantiles: a HAR volatility engine (Corsi 2009) scales the day, overnight-regime × NR7-compression conditioning selects the right history, and Koenker–Bassett quantile levels print a probability fan at the open — 50/70/90% boundaries for today’s highs and lows. A five-signal ensemble of published intraday effects votes on direction, weighted only by demonstrated walk-forward edge — and every claim carries a Wilson confidence interval. If the edge isn’t statistically real, PYTHIA says so.

  • Quantile Fan — 50/70/90% conditional boundaries for session highs and lows, printed at the open
  • Edge-Weighted Ensemble — five published directional signals; no walk-forward edge, no vote
  • Range-Completion Projection — live projected final high/low from intraday volatility periodicity
  • Honest Inference — Wilson 95% intervals; flagged significant only when the CI clears 50%
HAR (Corsi 2009)Koenker–BassettNR7 CompressionWilson Intervals
// HOW IT’S USED — PYTHIA σ.2 PLAYBOOK
  1. Run it on intraday futures charts. 1m–15m with ETH data on; pick your product preset (ES/NQ/YM/RTY, GC/SI, or CL). At the RTH open, PYTHIA prints the day’s conditional quantile fan.
  2. Trade the fan as a map of “days like today.” The 90% levels are statistical exhaustion zones — natural profit targets and fade locations. The 50% levels are routine territory, not edges.
  3. Wait for the ensemble to lock. Thirty minutes in, five published signals vote with edge-earned weights. Respect a NEUTRAL reading — it means no expert currently has demonstrated edge.
  4. Check the verdict before trusting the bias. Only a Wilson CI fully above 50% reads SIGNIFICANT. Anything else is statistically a coin flip — and PYTHIA is honest enough to say so.
The fan says how far days like today travel. The ensemble says which way. The CI says whether to believe it.
PYTHIA σ.2CONDITIONAL QUANTILE ORACLELIVE// HAR σ (CORSI 2009) · KOENKER–BASSETT QUANTILESWALK-FORWARD · WILSON 95% CICONDITIONAL QUANTILE FAN — TODAY'S SESSIONRTH OPENOPEN 5,214.2590% +38.5070% +24.7550% +14.2550% −13.0070% −22.5090% −35.25MEDIAN CLOSE · 1ST BAR UPPROJ. FINAL HIGH70% TAGGEDCONDITIONING: ON-UP × NR7 → n=47 · FALLBACK JOINT→ON→ALLSIGNAL ENSEMBLES1 ON DRIFTw .18S2 1ST CANDLEw .12S3 CLOSE LOC·w 0S4 PRIOR DIRw .05S5 FIRST 30Mw .21▲ BULLISH +0.46RECORD 57.8% [51.2–63.9] n=142VERDICT: SIGNIFICANTWILSON 95% CI FULLY ABOVE 50%
LIVE ON YOUR CHARTS

Built for futures — ES, NQ, YM, RTY, GC, SI, CL — on 1m–15m charts. Session presets and alerts included.

// DIVISION 02

HFT ALGOS

Three autonomous execution systems. Every bot is fully systematic, risk-managed, and auditable — no discretion, no emotion, no exceptions.

FLAGSHIP

VANTA α.1

MRKT Theory Alpha. The flagship engine — fuses all five theory models into one composite conviction score and executes it systematically.

Ensemble AlphaSignal FusionSystematic Execution
ACTIVE

BREACH β.2

Range Breakout Bot. Hunts volatility compression inside defined ranges and strikes on confirmed structural breaks — never on the fakeout.

Compression DetectionBreakout LogicFalse-Break Filter
ACTIVE

RECOIL ρ.1

Retracement Bot. Buys fear inside strength — waits for impulsive moves, measures the pullback, and enters at confluence with the trend intact.

Pullback LogicConfluence ZonesTrend Filter
Α SERIES · FLAGSHIP

VANTA α.1

MRKT Theory Alpha Engine

The flagship. VANTA ingests live output from all five MRKT Theory models — path signatures, cascade criticality, topological structure, spectral compression, and memory state — and fuses them into a single composite conviction score. When the theories agree, VANTA acts; when they conflict, it stands down. Position size scales with structural clarity, never with hope.

  • Composite Signal Fusion — five models, one conviction score
  • Regime-Aware Sizing — exposure scales with structural clarity
  • Asymmetric Risk Engine — cuts losers fast, presses confirmed winners
  • Full Audit Trail — every entry mapped to the model state that caused it
Ensemble AlphaConviction ScoringRisk Overlay
// HOW IT’S USED — VANTA α.1 PLAYBOOK
  1. Allocate, then supervise — don't override. VANTA runs autonomously; your job is choosing the risk budget it manages, not second-guessing entries.
  2. Understand the conviction logic. Composite score above 0.7 with model agreement = deployment; models in conflict = automatic stand-down. Flat is a position, and VANTA takes it often.
  3. Review the audit trail weekly. Every fill maps to the exact model state that caused it — you can always answer “why did it trade?” That's the discipline manual traders never get.
  4. Judge it on the risk-adjusted curve, not single trades. The asymmetric engine cuts losers fast and presses confirmed winners — expect many small scratches and occasional large runs.
Five theories, one vote. When they agree, act; when they argue, cash.
VANTA α.1MRKT THEORY ALPHA ENGINEEXECUTING// ENSEMBLE FUSION · 5 MODELS · 1 SCORERISK ENGINE: ASYMMETRIC · AUDIT: FULLΨSIGILχSWARMΞATLASΛSPECTRAηRELICFUSION CORECOMPOSITE CONVICTION0.81LONG5/5 MODELS AGREE → DEPLOYEQUITY
AUTONOMOUS EXECUTION

Fully systematic deployment of the MRKT Theory stack. Backtested, forward-tested, risk-capped.

Β SERIES

BREACH β.2

Range Breakout Bot

BREACH hunts compression. It identifies volatility coiling inside statistically defined ranges — energy building with nowhere to go — and strikes only on confirmed structural breaks. A false-break filter demands follow-through in order flow before committing, so BREACH skips the liquidity-grab fakeouts that bleed manual breakout traders dry.

  • Compression Detection — flags ranges where volatility is coiling
  • False-Break Filter — requires flow confirmation before entry
  • Dynamic Targets — ATR-scaled objectives adapt to the regime
  • Session Awareness — behavior tuned to session liquidity profiles
Volatility CompressionStructural BreaksATR Sizing
// HOW IT’S USED — BREACH β.2 PLAYBOOK
  1. Deploy on compression-prone instruments — index futures, FX majors, liquid crypto. BREACH needs ranges that coil; it has no edge on already-trending tape.
  2. Let it wait. BREACH acts only when compression ranks extreme (90th percentile+) and a structural break confirms with real order flow. Days of silence are the strategy working.
  3. Read skipped fakeouts as saved losses. The false-break filter demands follow-through before committing — every liquidity-grab it ignores is the drawdown you didn't take.
  4. Manage risk at the account level. Targets auto-scale with ATR; in violent sessions widen the risk cap rather than overriding trade logic mid-flight.
Energy stored in the coil pays the breakout. Patience is the position.
BREACH β.2RANGE BREAKOUT BOTARMED// COMPRESSION → CONFIRMED BREAK ONLYFALSE-BREAK FILTER: ACTIVE · ATR TARGETSRANGE HIGHRANGE LOW✕ FAKEVOLATILITY COILINGSTRUCTURAL BREAK ✓FLOW CONFIRMED — EXECUTECOMPRESSION92ND PCTL
AUTONOMOUS EXECUTION

Systematic breakout capture with false-break protection. Backtested, forward-tested, risk-capped.

Ρ SERIES

RECOIL ρ.1

Retracement Bot

RECOIL buys fear inside strength. It waits for an impulsive move to qualify — strong, persistent, structurally clean — then measures the pullback against mapped confluence zones and enters only while trend integrity holds. If the retracement cuts too deep and the structure degrades, RECOIL walks away. There is always another trade.

  • Impulse Qualification — only trades pullbacks of moves worth trusting
  • Confluence Mapping — retracement, VWAP, and structure zones stacked
  • Trend Integrity Check — continuous validation that the move is alive
  • Adaptive Stops — invalidation placed where the thesis actually dies
Pullback LogicConfluence ZonesTrend Filter
// HOW IT’S USED — RECOIL ρ.1 PLAYBOOK
  1. It trades second chances, not bottoms. RECOIL activates only after an impulsive move qualifies — strong, persistent, structurally clean. No impulse, no interest.
  2. Entries fire only inside stacked confluence. Retracement zone + VWAP + prior structure must align while trend integrity holds. One missing pillar = no trade.
  3. Respect the walk-away. If the pullback cuts too deep and structure degrades, RECOIL abandons the setup permanently — there is no chase logic, by design.
  4. Expect selectivity, not frequency. Fewer, higher-quality fills with invalidation placed where the thesis actually dies — that asymmetry is the whole edge.
Buy fear inside strength — but only where three reasons agree on the price.
RECOIL ρ.1RETRACEMENT BOTSTALKING// BUY FEAR INSIDE STRENGTHCONFLUENCE: FIB + VWAP + STRUCTURE0.5000.618CONFLUENCE ZONEVWAPIMPULSE ✓ QUALIFIEDPULLBACK MEASUREDENTRYINVALIDATIONTREND INTEGRITY: HOLDING
AUTONOMOUS EXECUTION

Systematic pullback entries at mapped confluence. Backtested, forward-tested, risk-capped.

// DIVISION 03

TERMINALS

Professional workstations built for a single purpose: total situational awareness. One screen for probability markets, one command center for everything else.

BETA

DELPHI Π.1

Prediction Market Terminal. Polymarket, Kalshi, and event markets aggregated into one screen — odds, flow, and cross-venue mispricings in real time.

Cross-Venue OddsMispricing ScannerEvent Flow
FLAGSHIP

MQT-1

The MRKT Quant Terminal. A Bloomberg-class command center — multi-asset data, institutional charting, and the full MRKT Theory stack overlaid on every instrument.

Multi-Asset DataModel OverlaysCommand Line
Π SERIES · PREDICTION MARKETS

DELPHI Π.1

Prediction Market Terminal

Named for the oracle. DELPHI pulls every liquid prediction market — Polymarket, Kalshi, and event venues — into a single professional screen. It normalizes odds into implied probabilities, tracks flow and liquidity shifts, and runs a continuous mispricing scan: when two venues price the same event differently, or when market odds diverge from model-estimated probabilities, DELPHI flags it before the crowd closes the gap.

  • Cross-Venue Aggregation — every major prediction market on one tape
  • Mispricing Scanner — flags probability spreads between venues in real time
  • Calibration Engine — implied odds vs model-estimated probabilities
  • Event Radar — catalysts, resolution dates, and liquidity migration
PolymarketKalshiProbability CalibrationArb Detection
// HOW IT’S USED — DELPHI Π.1 PLAYBOOK
  1. Build your event watchlist first. DELPHI normalizes every venue's odds into one implied-probability tape — Fed decisions, elections, crypto milestones, macro prints, side by side.
  2. Trade the scanner's spread flags. When Polymarket and Kalshi price the same event 4¢ apart, buy the cheap venue, sell the rich one, and let resolution — or convergence — pay you.
  3. Use the Calibration Engine for directional edge. Market implied 62%, model estimate 57% = the crowd is overpaying. Fade it or stand aside; never bet with a negative-edge number.
  4. Watch the Event Radar into resolution. Liquidity migrates late and spreads widen early — position before the crowd arrives, exit before it stampedes out.
Odds are prices. Prices disagree. DELPHI finds the disagreement first.
DELPHI Π.1PREDICTION MARKET TERMINALBETA// VENUES: POLYMARKET · KALSHI · EVENTSCAN CYCLE 240ms · 1,204 MARKETSEVENTPM¢KL¢ΔFED CUT — MAR FOMC62.458.2+4.2ARBBTC > $150K BY Q131.830.9+0.9US CPI < 2.5% FEB PRINT44.145.0-0.9ETH ETF NET INFLOW WK71.669.4+2.2OPEC CUT EXTENSION55.354.8+0.562.462.9FED CUT — MARIMPLIED 0.62MODEL 0.57EDGE +5.0%RESOLVES 18DDL>scan --spread 3% --venue all3 SPREADS > 2% · LIQUIDITY MIGRATING → PM · NEXT RESOLUTION 04:12:33
BETA ACCESS

One screen for every probability market. Early access rolling out to the waitlist.

MQ SERIES · FLAGSHIP

MQT-1

MRKT Quant Terminal

The command center. MQT-1 unifies futures, equities, crypto, and macro data into one keyboard-first workspace — Bloomberg-class density, MRKT QUANT mathematics. Every chart can run the full theory stack live: SIGIL signatures, SWARM criticality, ATLAS topology, SPECTRA compression, and RELIC memory, overlaid on any instrument, any timeframe, with alerting wired to all of it.

  • Unified Market Data — futures, equities, crypto, and macro in one workspace
  • Theory Stack Overlays — all five models live on any chart
  • Command-Line Control — keyboard-first workflow, terminal style
  • Custom Workspaces — layouts, watchlists, and a full alerting engine
Multi-AssetInstitutional ChartingModel IntegrationAlerting Engine
// HOW IT’S USED — MQT-1 PLAYBOOK
  1. Make it the primary screen. Futures, equities, crypto, and macro in one keyboard-first workspace — stop alt-tabbing between six half-tools.
  2. Overlay the full theory stack on any chart. SIGIL, SWARM, ATLAS, SPECTRA, RELIC live on one instrument — model confluence (4/5+ agreeing) is the highest-value signal the platform produces.
  3. Drive it from the command line. overlay sigil --depth 4 · alert swarm n>0.9 · layout macro — muscle memory beats mouse hunting.
  4. Alert on model states, not price levels. Get pinged when structure changes — criticality rising, topology collapsing, memory flipping — not after the move already happened.
One screen, five theories, zero blind spots. The market in a single pane of glass.
MQT-1MRKT QUANT TERMINAL — COMMAND CENTERFLAGSHIP// FEED: LIVE · LAT 0.4ms · 6 ASSET CLASSESWORKSPACE MACRO-01 · ALERTS 14 ARMEDWATCHLISTES+0.42%NQ+0.87%CL-1.13%BTC+2.41%DXY-0.22%ZN+0.08%NQ · 5M · THEORY STACK ONΨ.87χ.94THEORY STACKΨSIGIL0.87χSWARMn.94ΞATLAS2.14ΛSPECTRA61%ηRELICH.63CONFLUENCE: 5/5 ▲MQ>overlay sigil --depth 4 --alert rhyme>0.8✓ OVERLAY LIVE
FLAGSHIP WORKSTATION

The full MRKT QUANT stack in one command center. Request access below.

// DIVISION 04

RESEARCH & DEVELOPMENT

The mathematics underpinning every MRKT QUANT model — foundational and frontier research, mapped tool by tool.

SIGIL_Ψ.1

A Primer on the Signature Method in Machine Learning

I. Chevyrev, A. Kormilitzin

Path signatures as universal nonlinear features for sequential data — the bridge from rough path theory to practical time-series learning.

SIGIL_Ψ.1

Rough Paths, Signatures and the Modelling of Functions on Streams

T. Lyons

Foundational treatment by the creator of rough path theory: the signature as a faithful, hierarchy-ordered description of a data stream.

SWARM_χ.2

Hawkes Processes in Finance

E. Bacry, I. Mastromatteo, J.-F. Muzy

Definitive survey of self-exciting point processes in market microstructure — estimation, kernels, and applications to order flow.

SWARM_χ.2

Critical Reflexivity in Financial Markets: A Hawkes Process Analysis

S. Hardiman, N. Bercot, J.-P. Bouchaud

Shows markets operate near criticality with branching ratios approaching one — most activity is self-generated rather than news-driven.

ATLAS_Ξ.1

Topological Data Analysis of Financial Time Series: Landscapes of Crashes

M. Gidea, Y. Katz

Persistence landscapes detect early-warning signals ahead of the 2000 and 2008 crashes — the core empirical result behind ATLAS.

ATLAS_Ξ.1

Topology and Data

G. Carlsson

The landmark survey of topological data analysis: persistent homology, barcodes, and why shape is a robust lens on noisy data.

SPECTRA_Λ.3

Noise Dressing of Financial Correlation Matrices

L. Laloux, P. Cizeau, J.-P. Bouchaud, M. Potters

The classic result: most of the empirical correlation spectrum is indistinguishable from random matrix noise — with profound portfolio implications.

SPECTRA_Λ.3

Cleaning Large Correlation Matrices: Tools from Random Matrix Theory

J. Bun, J.-P. Bouchaud, M. Potters

Comprehensive review of RMT-based estimators for denoising correlation matrices — the modern toolkit SPECTRA is built on.

RELIC_η.2

Long-Term Memory in Stock Market Prices

A. W. Lo

Rigorous test of long-range dependence in equity returns, introducing the modified rescaled-range statistic that sharpened Hurst analysis.

RELIC_η.2

A Brief History of Long Memory

T. Graves, R. Gramacy, N. Watkins, C. Franzke

From Hurst's Nile studies to fractionally integrated processes — the intellectual lineage of long-memory modeling RELIC draws on.

Σ

See the Math in Action

These papers power our models. Explore MRKT Theory to see mathematics become market structure.

Watch It Execute

See the theory deployed live — VANTA, BREACH, and RECOIL turn research into autonomous execution.

// THE DESK

THE TEAM

One desk. One standard. Institutional experience, deployed for the trader on the other side of the screen.

@mrktmoh

FOUNDER · HEAD OF RESEARCH · MRKT THEORY ARCHITECT

Spent 1.5 years at Citadel as an L5 Quant, building and stress-testing systematic models inside one of the most competitive quantitative trading environments on earth — where an edge either survives the math or it doesn't ship.

Left to launch his own hedge fund and now develops MRKT Theory full time — the proprietary framework behind every model, bot, and terminal on this site. The same rigor that governed institutional capital now runs through every product carrying the MQ mark.

Along the way he's taught 500+ traders — turning retail chart-watchers into systematic thinkers, with documented student success across futures, equities, and prediction markets.

L5
QUANT @ CITADEL
1.5 YRS
INSTITUTIONAL DESK
500+
TRADERS TAUGHT
10+
MODELS & SYSTEMS
// THE MISSION

Make MRKT QUANT the one-stop shop for serious traders. The theory to understand markets, the indicators to read them, the bots to execute them, and the terminals to command it all — one ecosystem, one standard, no fluff. Everything a trader needs to stop gambling and start engineering.

RISK DISCLAIMER: Trading involves substantial risk of loss and is not suitable for every investor. MRKT QUANT tools are analytical instruments, not financial advice. Past performance — live, backtested, or simulated — does not guarantee future results. Never trade capital you cannot afford to lose.
MRKT QUANT
© 2026 MRKT QUANT. All rights reserved.