Quant Developer · Systematic Trader

Daniel Hughes

I research, build and run systematic trading strategies end to end — from raw options tape and volatility surfaces to out-of-sample validation and live execution — and built TradingEdge, an analytics platform used by ~2,600 paying members.

▶ Looking to join a professional quant firm as a quant trader, quant developer or software engineer.

CV GO LINKEDIN GO EMAIL GO

Projects8 ENTRIES · CLICK A TAB, OR TYPE A CODE BELOW

F1 TEDG GOAUG 2025 – PRESENTPRODUCTION

SPXW · Net dealer gamma, strike × expiryTEDG

LOADING_

The gamma heatmap members use to see where dealer hedging concentrates. Values normalised.

TradingEdge

A production market-data and analytics platform I built and run solo, used by ~2,600 paying members: dealer positioning, options flow, the volatility complex and earnings research.

Paying members: ~2,600
Commits (solo): ~1,800
Lines of code: ~120k
CLI / scheduled cmds: 95
Data providers: 10+
Test files: 109

TradingEdge is the analytics platform behind a paid trading community, and the delivery layer for most of my quant work: dealer-positioning maps from my gamma engine, IV-implied levels, the volatility term structure and a backtested setup scanner. I designed, built and operate it solo — about 1,800 commits in 14 months.

It is where I learned to ship numbers other people act on: 10+ data vendors reconciled on a dense ET schedule, guards against stale and inconsistent data, and the monitoring, security and access control a paying user base demands.

TEDG DES GO → full write-up

F2 GEX GOMAR 2026 – PRESENTPRODUCTION

SPXW · Net dealer gamma by strikeGEX

LOADING_

Midday engine snapshot, 8 Oct 2026. Bars: net dealer gamma per strike (normalised). Spot sat in a negative-gamma regime, so the flip is off-screen.

Real-Time Dealer Gamma Engine

A real-time market-microstructure system: it streams the full OPRA options tape, classifies which side dealers took on every print, and prices gamma and vanna exposure every 30 seconds.

Symbols served: 29
Snapshot cadence: 30s
Load-tested: 3.1k msg/s
Dealer-side AUC: 0.71
Tests: 642
Shadow vs prod ρ: 1.000

A from-scratch equivalent of commercial dealer-positioning products, combining three problems: trade-side classification on noisy microstructure data, volatility-surface calibration, and a low-latency streaming system that has to stay up all session.

It keeps a dealer-signed position book per strike and expiry for SPX, SPY, QQQ and 26 single names, and publishes gamma (GEX) and vanna (VEX) profiles to TradingEdge. Its recorded tape feeds a pre-registered research programme on how price behaves around gamma nodes.

GEX DES GO → full write-up

F3 IBA GOOCT 2025 – PRESENTLIVE · PAPER

NQ 1m · Example tradeIBA

LOADING_

1 Sep 2026 (backtest, real NQ). The opening range is marked, price pulls back into the entry zone and it buys with the map bias; it scales out at the first node, and both runners reach their range extensions.

IB Agent

A systematic intraday Nasdaq-futures strategy that combines opening-range structure with dealer-gamma exposure maps. Paper-trading live on IBKR and mirrored to prop accounts.

Annualised return: +50%
Max drawdown: 6.1%
Sharpe: 2.1
Profit factor: 1.90
Avg trade: +0.50%
Win rate: 49%

* Backtest, 212 trading days (Oct 2025 – Oct 2026) on real 1-second NQ prices, 72 trades. $100K account after live position limits and measured execution costs; compounding, no withdrawals. Gross of both: +81% annualised, PF 2.30.

A deterministic intraday engine for Nasdaq futures. It turns a published initial-balance method into explicit rules, adds a directional bias read from three gamma exposure maps, and tags every rule by source so my additions can be tested separately from the original.

The same engine code makes every decision in backtest and live, once per 5-second tick, so there is no separate research implementation to drift. Live, it trades an IBKR paper account with a resting stop and mirrors into four prop accounts through an MT5 bridge, with tighter loss limits than the firm’s own.

IBA DES GO → full write-up

Backtest return, % of account

LOADING_

$100K account, compounding. Amber: after live position limits and measured execution costs. Grey: gross of both. Dashed: QQQ buy-and-hold over the same dates. The strategy lines are flat from December to mid-February because only six sessions of gamma-map history exist in that stretch: missing data, not a quiet patch.

F4 GEXR GOBACKTEST 2022 – 2026LIVE · PROP

SPX 15m · Example tradeGEXR

LOADING_

5 Sep 2025 (backtest). Price tags the put wall late morning, the strategy buys it, and the target is hit on the next bar.

GEX Reverter

Systematic mean reversion at dealer-gamma walls on the S&P 500, trading live on a prop-firm account.

Trades: 499
Win rate: 72.7%
Profit factor: 2.60
Sharpe: 6.69
Max drawdown: 6.7%
Walk-forward PF: 2.72

* 4-year backtest (2022–2026), the configuration closest to what is deployed, with prop-firm costs. Not live results.

Thesis: when dealers are long gamma, their hedging dampens moves and tends to pin the index between the largest gamma strikes. Every 15 minutes the strategy rebuilds those levels from SPX and SPXW greeks and open interest, then fades the walls, with gamma-regime, volatility-regime and trend filters deciding whether and where to engage.

Execution is split across two machines: a Python engine computes levels and sends commands over a TLS socket to an MQL5 expert advisor, which places limit orders with stop and target attached and reports fills back. The broker is the source of truth for positions.

GEXR DES GO → full write-up

Backtest equity

LOADING_

4-year backtest with prop-firm costs.

F5 WHAL GOBACKTEST SEP 2023 – OCT 2025LIVE

QUBT 1D · Example tradeWHAL

LOADING_

Nov 2024: about $755K of bullish call premium against a ~$413M market cap put QUBT at the top of the ranking, twice in a week. Both entries were held to their time exit (+433% and +267%).

Whale Flow

A cross-sectional options-flow signal: buy the stock with the most bullish call premium relative to its market cap. Trading live on Alpaca.

Trades: 426
Return: +224%
SPY same period: +57%
Win rate: 50.7%
Max drawdown: 44.2%
Best trade: QUBT +433%

* Backtest Sep 2023 – Oct 2025. Signal after the close, filled the next trading day with slippage.

Hypothesis: when call buying is very large relative to a company’s size, informed money may be positioning. After each close the system filters the day’s options flow to bullish, longer-dated call buying, sums premium per ticker, normalises by market cap and ranks the cross-section.

It trades the shares, not the options, so the signal is tested apart from the options’ leverage and decay. Execution is fully automated: the evening job selects, and the next morning it closes matured positions and opens the new one.

WHAL DES GO → full write-up

Backtest vs SPY

LOADING_

Cumulative return, strategy vs SPY (dividend-adjusted).

F6 BTA GO2026RESEARCH · PAPER

10:00 bias · LLM vs rulesBTA

LOADING_

Same engine and execution over 147 days; only the 10:00 bias read changes. Claude finished below “always short”.

LLM Trader & Backtest Agent

A controlled experiment: can a blind LLM read gamma maps as well as rules? Same days and same execution — only the decision-maker changes.

Rules bias P&L: $8.1k
LLM bias P&L: $2.9k
LLM, disagree days: −$271
Test days: 147
Blind agent-days: 184
Blind-run PF: 2.45

* LLM vs rules: 147 days, same engine and execution, 4 MNQ. Blind runs: 184 agent-days.

Experiment: a Claude agent reads text renderings of intraday gamma maps at fixed checkpoints, blind to outcomes, and writes trade tickets; a deterministic tool sets price, size and stops. To isolate the LLM’s contribution, I swapped only the 10:00 bias read inside the IB Agent engine and kept everything else fixed.

Data layer: a credit-aware collector stores the maps as raw JSON plus derived Parquet minute tables, with each expiry netting kept as a separate view so they can’t be mixed up. The agent has paper-traded every weekday since October 2026.

BTA DES GO → full write-up

Cumulative P&L by bias read

LOADING_

Rule-based vs Claude vs a naive baseline, same days.

F7 VOL GO2017 – 2026 DATARESEARCH

Confirmatory resultsVOL

LOADING_

Each pre-registered test with its estimate and 95% block-bootstrap interval. Hover a row for the note.

Sector Vol Surface Research

Pre-registered statistical tests of whether sector-ETF option surfaces predict forward realised volatility.

Sessions: 72,168
ETFs: 29
History from: 2017
Bootstrap draws: 10,000
Hypotheses passed: 1 of 3
H-INV effect: +0.8–2.1 vol pts

A volatility-forecasting study: rebuild sector-ETF implied-volatility surfaces from 2017 onward and test, under pre-registration, whether their shape predicts realised volatility. Hypotheses, horizons and thresholds were fixed before looking at results, and multiple-testing correction was applied across them.

Inverted term structures do precede volatility expansion (+0.8 to +2.1 vol points at 5–10 day horizons; 7 of 9 pre-registered cells pass, the 21-day tail cells don’t). The risk-reversal hypothesis failed, and a cross-sector spill-over result was withdrawn after overlap placebos showed it was an errors-in-variables artefact.

VOL DES GO → full write-up

F8 RLOG GO9 KILLED

Research Log — what didn’t work

Most ideas don’t survive contact with out-of-sample data. These are the ones I killed, and why.

Idea	Sample	Result	Verdict	What I learned
GARMA A discretionary liquidity-sweep playbook, coded rule for rule	155 days, real NQ	245 trades · PF 1.08 · negative without the top 5 wins	NO EDGE	Looked fine on a proxy price series (PF ~1.7); the edge disappeared on the real futures tape. Validate on the instrument you trade.
GARMA-Wicks Wick-and-reclaim at session levels with a volume burst	677 days, NQ 1m	1,263 trades · PF 0.90 · −77.7R total	NEGATIVE	Above zero in only 6% of bootstrap resamples. Never traded.
Mechanical gamma setups Eight rule-based setups read off intraday gamma maps	Minute-bar backtest	93 trades · PF 0.76	NEGATIVE	Map “shapes” alone don’t pay; context (time, opening range, regime) carries the edge.
LLM map reader Let Claude set the morning bias instead of rules	147 days	$2.9k vs $8.1k for rules	REJECTED	The deterministic rule wins. The LLM stays a proposer, not a decider.
IB reversal setups Sweep and flip reversals inside the IB strategy	Hold-out	Lost on the hold-out	REJECTED	Switched off in production; the core continuation setup carries the strategy.
RS Pullback Relative-strength pullbacks to weekly EMAs on the S&P 500	Two hold-outs	13,826 trades · +0.6% / trade vs random · ~0% account CAGR	NOT TRADABLE	A real but tiny edge that can’t beat holding the index after capital constraints.
Sector vol H1 Risk-reversal percentile predicts forward vol	72k sessions	Failed the pre-registered test	FAILED	Reported as a null result rather than re-specified until it passed — no p-hacking.
Sector vol H2 SMH’s surface informs XLK’s forward vol	72k sessions	Passed, then withdrawn	WITHDRAWN	Overlap placebos showed an errors-in-variables artefact from shared holdings.
Third-party risk on/off Rebuild a published daily SPY regime signal	2019 – 2026	Rebuild: 12.9% CAGR vs 32.7% claimed	REJECTED	The original was chosen from ~100k candidates in-sample; it trails SPY since 2023.

RLOG DES GO → full write-up