Market Making: Competition & Research Extension
The historical competition and later research answer different questions. The first records a simulated live team event; the second uses synthetic episodes to study how quoting policies trade fill rate, inventory and downside against simulated P&L.
Central findingMore aggressive uncertainty-scaled sizing increased mean simulated P&L but also increased drawdown and inventory; a sudden fair-value shock hurt the lower tail of every tested policy.
Evidence
The result in context
- Simulated live rounds
- 9
- Displayed leaderboard position
- 11 / 23
- Synthetic extension episodes
- 1,088
- Automated extension tests
- 25
On this page
Question
How should a market-making policy balance uncertain fair value, quote width, inventory and position sizing across changing flow conditions?
A simulated live team competition followed by an independent synthetic research extension on uncertainty, quote width, inventory and stress.
Part I · Team Project
Simulated live competition
The team completed nine simulated live rounds and finished 11th of 23 displayed leaderboard entities. Net simulated competition P&L was £46,742.40 from a £100,000 reference start.
The historical team was Husaam Ateeq, Hanad Ali, Blazej Olszta and Abdallah Ramadan.
Historical team result. Individual contribution is not broken out because the surviving project record does not assign roles in enough detail.
Competition approach
Quoting decisions under competition constraints
The supported team mechanics combined fair-value prediction, uncertainty-aware quoting and stock-specific tactics with forced-trade decisions, position sizing and leaderboard-aware adjustments.
Part II · Independent Extension
A synthetic environment for controlled policy tests
Husaam’s later independent extension runs 1,088 synthetic episodes across fixed, uncertainty-based and inventory-aware quote policies. It studies fair-value prediction, uncertainty, quote width, inventory, position sizing and stress behaviour.
Sublinear uncertainty width produced approximately £1,790 mean simulated P&L in the documented setup, but the more important result was the trade-off between fill behaviour, inventory, drawdown and stress resilience.
Two phases, two evidence sets
Competition and extension
The historical team event is kept separate from later individual research.
The leaderboard record covers nine simulated rounds from a £100,000 reference start and is not live market performance.
Interactive evidence
The independent extension compares six policies on a common simulator
Mean synthetic P&L across baseline causal-ridge runs; this is not historical competition performance.
Read: Policy differences are measured under controlled assumptions and must remain separate from the team’s rank-11 historical checkpoint.
Synthetic independent extension; not historical competition data.
Data table · 6 verified rows
| Policy | Mean Pnl | Mean Max Drawdown | Mean Absolute Inventory | Mean Passive Fill Rate Pct |
|---|---|---|---|---|
| Fixed Symmetric | 618.673394 | 650.047291 | 28.176355 | 38.240741 |
| Historical Heuristic Approximation | -99.997259 | 854.065541 | 29.330898 | 54.521605 |
| Inventory Skewed | 412.566888 | 665.522649 | 28.216684 | 37.947531 |
| Uncertainty Inventory | 1,270.589576 | 540.62978 | 27.192524 | 79.33642 |
| Uncertainty Scaled | 1,714.304704 | 460.949019 | 27.236471 | 77.777778 |
| Uncertainty Scaled Nonlinear | 1,790.219376 | 457.207579 | 27.037311 | 74.475309 |
Interactive evidence
Inventory control is visible at every simulated event
Gross and net inventory across the 324-event representative synthetic episode.
Read: The policy allows inventory to accumulate but continuously skews quotes around the current position; event-level values remain auditable.
Synthetic independent extension; one representative episode, seed 0.
Data table · 648 verified rows
| Event Id | Inventory | Inventory Type | Event Label |
|---|---|---|---|
| 0 | 10 | Gross Inventory | Event 0 |
| 0 | 10 | Net Inventory | Event 0 |
| 36 | 150 | Gross Inventory | Event 36 |
| 72 | 130 | Gross Inventory | Event 72 |
| 108 | 210 | Gross Inventory | Event 108 |
| 144 | 250 | Gross Inventory | Event 144 |
| 179 | -250 | Net Inventory | Event 179 |
| 215 | -230 | Net Inventory | Event 215 |
| 251 | -180 | Net Inventory | Event 251 |
| 287 | -52 | Net Inventory | Event 287 |
| 323 | 226 | Gross Inventory | Event 323 |
| 323 | -64 | Net Inventory | Event 323 |
Stress
Every policy remained vulnerable to abrupt repricing
A sudden fair-value shock produced negative fifth-percentile simulated P&L for every policy. Uncertainty-scaled sizing raised mean P&L while also increasing drawdown and inventory.
Interactive evidence
Policy rankings shift across nine synthetic stress regimes
Mean synthetic P&L across nine stress scenarios and six fixed-size policies.
Read: Mean outcomes change materially by regime and policy. These means do not show the negative fifth-percentile shock outcomes documented in the source analysis.
Synthetic independent extension; mean scenario outcomes, not realised trading P&L or lower-tail risk.
Data table · 54 verified rows
| Scenario | Policy | Mean Pnl | Mean Max Drawdown |
|---|---|---|---|
| Baseline | Fixed Symmetric | 707.94394 | 565.44254 |
| Baseline | Historical Heuristic Approximation | -205.87256 | 920.286952 |
| Greater Volatility | Fixed Symmetric | 778.635679 | 1,251.687236 |
| Higher Prediction Error | Fixed Symmetric | 688.317714 | 646.57999 |
| Inventory Accumulation | Fixed Symmetric | 943.556282 | 1,478.700028 |
| Narrower Competitor Spreads | Fixed Symmetric | 166.561428 | 680.227665 |
| Narrower Competitor Spreads | Uncertainty Scaled Nonlinear | 252.637573 | 683.661487 |
| One Sided Flow | Uncertainty Scaled Nonlinear | 1,551.08299 | 1,123.160915 |
| Persistent Directional Moves | Uncertainty Scaled Nonlinear | 1,583.915332 | 634.512162 |
| Sudden Fair Value Shock | Uncertainty Scaled Nonlinear | 974.708273 | 3,302.109695 |
| Wider Competitor Spreads | Uncertainty Scaled | 1,854.337998 | 408.278949 |
| Wider Competitor Spreads | Uncertainty Scaled Nonlinear | 2,024.83958 | 391.504549 |
Interactive evidence
Wider quotes trade off against passive fill probability
Policy-bucket means from the independent synthetic extension.
Read: The relationship is policy-dependent rather than a universal curve, so points remain grouped and labelled by policy.
Synthetic independent extension.
Data table · 22 verified rows
| Policy | Bucket | Observations | Mean Quote Width | Passive Fill Probability Pct |
|---|---|---|---|---|
| Fixed Symmetric | 1 | 6,480 | 2.41 | 38.240741 |
| Historical Heuristic Approximation | 1 | 1,309 | 0.338144 | 79.526356 |
| Historical Heuristic Approximation | 2 | 1,287 | 0.790047 | 85.081585 |
| Historical Heuristic Approximation | 5 | 1,296 | 10.575208 | 0 |
| Uncertainty Inventory | 1 | 1,301 | 0.649931 | 78.01691 |
| Uncertainty Inventory | 3 | 1,292 | 1.333367 | 79.334365 |
| Uncertainty Scaled | 1 | 1,327 | 0.691341 | 78.824416 |
| Uncertainty Scaled | 3 | 1,298 | 1.461364 | 79.583975 |
| Uncertainty Scaled | 5 | 1,252 | 2.829617 | 71.166134 |
| Uncertainty Scaled Nonlinear | 3 | 1,294 | 1.65456 | 76.043277 |
| Uncertainty Scaled Nonlinear | 4 | 1,376 | 2.030574 | 72.093023 |
| Uncertainty Scaled Nonlinear | 5 | 1,200 | 2.739133 | 74.833333 |
Limitations
What this evidence does not establish
- Extension results come from a synthetic research environment and are not historical competition or live-trading performance.
- The historical leaderboard shows 23 displayed entities; no larger registered-team claim is made without independent evidence.
- The surviving competition record does not include quote, fill or round-ledger data, so the historical rounds cannot be deterministically replayed.
Source and reproducibility
Trace the evidence
Source code, evaluation outputs and supporting material are available in the repository.
View repository- Historical competition leaderboardassets/leaderboard.jpegCommit / evidence ID: ed4e4bb24b7d9af0595316557a1614baf3653594
- Historical competition summaryreports/historical_competition_summary.mdCommit / evidence ID: ed4e4bb24b7d9af0595316557a1614baf3653594
- Final independent extension commitresearch_extension/results/policy_summary.csvCommit / evidence ID: ed4e4bb24b7d9af0595316557a1614baf3653594