Skip to content
View talalmalikllc-debug's full-sized avatar

Block or report talalmalikllc-debug

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse

Hi, I'm Muhammad Talal Malik πŸ‘‹

Quantitative Researcher & Data Analyst Β· Founder of Khurjeen.pk
πŸ“ Pakistan Β· 🌍 Open to roles in finance

LinkedIn Email Store

I do academic-driven quantitative research. I start from published findings in empirical finance, replicate them on instruments that people actually trade, and then build and test my own hypotheses where the original result is weak, too costly to trade, or not fully explained. Every idea is a falsifiable hypothesis with a stated null. Each one is tested with out-of-sample discipline and reported honestly, including the ideas that fail.

I also run my own private-label e-commerce brand, so I'm used to making decisions with real money on the line.


πŸ§ͺ How I research

flowchart TB
    subgraph S1 [" Idea "]
    direction LR
    A[πŸ“„ Academic paper] --> B[πŸ” Replicate on a tradable market] --> C[πŸ”Ž Find where the effect lives or dies]
    end
    subgraph S2 [" Test "]
    direction LR
    D[πŸ’‘ My own hypothesis Hβ‚€ / H₁] --> E[πŸ”’ Pre-register rules] --> F[πŸ“Š Train β†’ Validate β†’ Test]
    end
    subgraph S3 [" Decide "]
    direction LR
    G[πŸ’Έ Costs & robustness checks] --> H[🧊 Freeze β†’ forward test]
    end
    S1 --> S2 --> S3
Loading
  1. Start from the literature. Take a published result, such as an anomaly, return predictability or a market-microstructure effect, and read it critically.
  2. Replicate before innovating. Rebuild the paper's main test on liquid, tradable futures and compare my numbers with the authors'.
  3. Audit the data first. Bar-timestamp conventions, session calendars, contract rolls, 1-min vs 5-min consistency and leakage tests come before any result is looked at.
  4. Form my own hypothesis. Ask where and when the effect is strongest, and turn that into a testable Hβ‚€ / H₁ with a concrete trading rule.
  5. Pre-register and split time. Fix the candidate rules in advance, choose on training data only, then judge on untouched validation and test periods.
  6. Make it survive reality. Model transaction costs under several scenarios and stress-test with the bootstrap, best-trade removal, delayed entries and parameter neighbours.
  7. Report everything. Disclose all rules tried, all limitations and all failures. Freeze the surviving rule and evaluate it only on new data.

πŸ“š Research log

Idea Source My hypothesis Outcome
Overnight return β†’ first-half-hour reversal Iwanaga & Sakemoto (2026), SSRN 5807282 The paper's SPY reversal also exists in Nasdaq-100 futures βœ… Replicated, but weaker: Ξ²Γ—100 = βˆ’4.15 (t = βˆ’2.42) vs βˆ’9.19 for SPY. Most of the effect sits in the first ~5 minutes, which is too small to trade after costs.
Momentum instead of reversal Extension of the above Trading with the overnight move works better ❌ Rejected in every holding window
Post-open rebound after extreme drops My own hypothesis, built on the replication The reversal is concentrated in extreme, volatility-scaled drops once a bounce has started βœ… Out of sample: +50.9 pts/trade after costs, t β‰ˆ 2.5 (18 trades). Frozen for forward testing. β†’ repo
Overnight drift (close β†’ open) Overnight-return anomaly literature Holding NQ overnight earns a premium after costs ⚠️ Positive over 2016–2026 (t β‰ˆ 2.5), but concentrated before 2021. 26 follow-up filters and enhancements tested: 0 significant, all reported.
Volatility regimes: scale vs direction Volatility-clustering and compression literature A quiet run-in forecasts the size and direction of the next move βœ… Size: yes. A quiet hour cuts the odds of a volatility spike by 13–55 pp, in 24/24 market Γ— session cells; 67/102 tests survive BH-FDR. ❌ Direction: no. 0/132 survive. β†’ repo
Weekday structure in NASDAQ-100 breakouts My own hypothesis Opening-range breakouts follow through more on some weekdays βœ… Friday β‰ˆ 3Γ— the average (t = 4.58), and it survives a selection-aware permutation test (P = 0.0017, 100k shuffles). Positive in every year 2022–2026. β†’ repo
FX fixing reversal Krohn, Mueller & Whelan (2024), Journal of Finance USD reverses around the ECB and London fixes in EURUSD ❌ Falsified: gross βˆ’0.43 pips/day (t = βˆ’0.31) vs the paper's β‰ˆ +6; a placebo anchor does as well as the real fix. β†’ repo
Intraday momentum Gao, Han, Li & Zhou (2018), Journal of Financial Economics The first half-hour predicts the last half-hour ❌ Does not replicate: β₁ is negative on all 7 instruments; 0 of 19 development-significant specs survive the 2025–26 holdout. β†’ repo

πŸ”¬ Featured research

NQ Post-Open Rebound After Extreme Overnight Drops Do Nasdaq-100 futures rebound between 09:35 and 10:00 New York time after an extreme overnight drop?

  • 10 years of intraday NQ futures data (2016 β†’ 2026), with timestamp, contract-roll and leakage checks
  • Pre-registered rule selection on training data, then separate validation and test periods
  • Out of sample: 18 trades, +50.9 points/trade after costs, t β‰ˆ 2.5, tested under 4 cost scenarios
  • Robustness: bootstrap confidence interval, best-trade removal, delayed entry, and all candidate rules disclosed
  • Started as a replication of an academic paper (Iwanaga & Sakemoto, 2026) on overnight-return reversal

Python pandas statsmodels SciPy matplotlib Jupyter

Intraday Volatility Regimes: What Is Forecastable and What Is Not A pre-registered, 330-test research programme on six markets (NASDAQ-100, S&P 500, Dow, DAX, gold, EUR/USD).

  • Volatility deseasonalised by time of day, so "quiet" means quiet for that clock time, not just the Asian session
  • Persistence tested against a first-order Markov chain, with day-block bootstrap confidence intervals
  • Every test logged with an explicit pre-specified/exploratory flag; Benjamini–Hochberg FDR control by family
  • Result: volatility scale is strongly forecastable 1–3 hours ahead, while direction is not (0 of 132 tests survive)

Day-of-Week Structure in NASDAQ-100 Opening-Range Breakouts 1,082 breakouts from 2022 to 2026, with forward returns net of the spread paid at each fill.

  • Five-test grading harness: quintiles, rank correlation, per-year sign, in-sample/out-of-sample split, placebo
  • Selection-aware permutation null: the best-of-5-days search is repeated inside each of 100,000 shuffles
  • Friday follow-through β‰ˆ 3Γ— the average (P = 0.0017); not an option-expiry artefact, and limitations stated

Replicating Published Intraday Anomalies Under Realistic Costs Three published effects re-tested on a different asset class, with every trade charged the spread quoted at that moment.

  • Journal of Finance 2024 (FX fixing), Journal of Financial Economics 2018 (intraday momentum), and Della Corte et al. 2022 (overnight–intraday reversal)
  • Newey–West t-statistics, stratified placebos, mirror controls and a sealed 2025–26 holdout
  • Documents the data-pipeline bugs found along the way (clock offsets, column-shifted contract specs, a bar-boundary error that faked a result)

Python NumPy pandas matplotlib Jupyter Β· each repo has an executed walkthrough notebook


πŸ› οΈ Skills

Area What I use
Quant research Backtesting, out-of-sample testing, transaction-cost modelling, bootstrap and Monte Carlo, regression, hypothesis testing
Market data Intraday futures bars, session calendars, contract rolls, look-ahead and leakage checks
Data analysis Python, pandas, NumPy, SciPy, statsmodels, Jupyter, data cleaning and validation
Visualisation matplotlib, equity curves, drawdowns, distribution and bucket analysis
E-commerce Shopify store management, private-label product sourcing, nationwide fulfilment

πŸ›οΈ Entrepreneurship: Khurjeen.pk

I own and run Khurjeen, a private-label bags brand in Pakistan, which I co-run with my partner. We sell backpacks, laptop bags, luggage, travel bags and accessories through our own Shopify store, with cash on delivery across Pakistan.


🎯 What I'm looking for

Opportunities in finance, including investment research, equities, trading, risk and financial data analysis. I'm happy to walk through any of my research in detail. The best way to reach me is LinkedIn or email.

πŸ“« Contact

Popular repositories Loading

  1. nq-post-open-rebound nq-post-open-rebound Public

    Pre-registered quant study: do Nasdaq-100 futures (NQ) rebound 09:35–10:00 NY after extreme overnight drops? Train/validate/test, cost scenarios, robustness checks.

    Jupyter Notebook

  2. talalmalikllc-debug talalmalikllc-debug Public

    Profile README

  3. volatility-forecasting-study volatility-forecasting-study Public

    Pre-registered 330-test study on 6 markets: intraday volatility scale is forecastable (67/102 survive BH-FDR), direction is not (0/132). Python, day-block bootstrap, Markov nulls.

    Jupyter Notebook

  4. nasdaq-orb-forward-returns nasdaq-orb-forward-returns Public

    Day-of-week structure in NASDAQ-100 opening-range breakouts: Friday follow-through ~3x average, survives a selection-aware permutation test (P=0.0017, 100k shuffles).

    Jupyter Notebook

  5. market-anomaly-replications market-anomaly-replications Public

    Replicating published intraday anomalies (JF 2024 FX fixing, JFE 2018 intraday momentum, overnight-intraday reversal) under realistic per-event costs and holdout tests.

    Jupyter Notebook