01First Place
ClockSense Trading
BITS Pilani
Shubham Nahata · Aayush Mittal
HFTC 2025
First international edition — retrospective
Five themed stages on the SHIFT platform, with teams from universities in the United States and abroad.
- Teams
- 22
- Students
- 70
- Universities
- 9
- Award ceremony
- May 7, 2025
Competition overview
The competition’s first international edition.
The 2025 Stevens High Frequency Trading Competition (HFTC-25) was the event’s first international edition, drawing 22 teams and 70 students from 9 universities in the United States and India. Students designed algorithms to trade against one another in a high-frequency market on the SHIFT simulation platform — the only competition of its kind in the world.
The competition ran across five themed stages: Up and Down, Swing Days, the VIX Stage, Zero Intelligence, and Reinforcement Learning. The award ceremony was held on May 7, 2025.
Who competed
70students
44 Undergraduate 63%
26 Graduate 37%
Top programs: computer science 17 · quantitative finance 16 · financial engineering 11
Universities all 9
- Stevens 50
- Georgia Tech 4
- Rutgers 4
- UT Austin 4
- Baruch 3
- BITS Pilani 2
- Boston College 1
- DePaul 1
- Fordham 1
Competition stages
- 01Up and Down
- 02Swing Days
- 03VIX Stage
- 04Zero Intelligence
- 05Reinforcement Learning
The award ceremony
May 7, 2025
Teams presented their strategies and received their awards at the close of the five-stage competition.
“Consistency. Simplicity. Execution.”
Final rankings
Winning teams
02Second Place
Spread Snipers
Stevens Institute of Technology
Sharif Haason · Saugat Shreshtha · Akbar Pathan · Scott Henriquez
03Third Place
Money Never Sleeps
Stevens Institute of Technology
Bence Marton · Ke Ren
Case studies
Highlighted strategies
How the top teams approached the 2025 markets.
ClockSense Trading
“Consistency. Simplicity. Execution.”
ClockSense treated the competition like an internship: disciplined pre-round research and backtesting, live monitoring during trading, and post-round analysis of every result. They won three consecutive rounds — Day 1 “Up or Down” (momentum-weighted ticker selection: 40% momentum, 30% volatility, 30% liquidity), Day 2 “Swing Days” (mean-reversion with a McClellan oscillator and stop-losses), and Day 3 “VIX Stage” (momentum/reversal classification with dynamic position sizing and stops) — each with the top or near-top Sharpe ratio. They took 2nd on Day 4 “Zero Intelligence” with a market-making approach, and — in their own candid post-mortem — slipped on the Day 5 “Reinforcement Learning” round, attributing it to over-aggression.
Strategy snapshot
- Approach
- Round-specific strategies with disciplined pre- and post-round analysis
- Market condition
- Five themed stages, from “Up or Down” to Reinforcement Learning
- Key idea
- Momentum-weighted ticker selection; mean reversion with a McClellan oscillator; momentum/reversal classification
- Result
- Won rounds 1–3 with top or near-top Sharpe; 2nd in round 4
Spread Snipers
Adaptive Market Making
Spread Snipers evolved a single market-making core to fit each week’s market. A naive best-bid/ask quoter failed in trending markets, so they added moving-average trend filters and quoted above or below mid-price accordingly, logging every order to CSV for analysis. Later rounds layered in VWAP, MACD, and Bollinger Bands (favoring VWAP in low-volatility windows), a reservation price with dynamic order sizing from inventory and order-book imbalance, and stale-order cancellation every 30 seconds. In the final reinforcement-learning round they optimized latency and reaction speed to post the second-highest PnL. Their takeaways: order-book analysis is richer than technical indicators alone, staying market-neutral can beat directional bets, and results hinge on tuning speed and inventory control.
Strategy snapshot
- Approach
- Adaptive market making
- Market condition
- Weekly-changing market conditions
- Key idea
- Trend filters, a VWAP reservation price, and sizing driven by inventory and order-book imbalance
- Result
- Second-highest PnL in the reinforcement-learning round
Money Never Sleeps
Avellaneda–Stoikov Market Making
The team implemented the Avellaneda–Stoikov optimal market-making model, setting bid and ask quotes around a reservation price driven by current inventory, asset volatility, order-flow intensity, and a risk-aversion parameter to balance spread capture against inventory risk. Their central challenge was inventory control — wide spreads and limited reaction time made unwinding positions costly — which they addressed by adding tighter inventory-penalty terms to the model’s utility function.
Strategy snapshot
- Approach
- Avellaneda–Stoikov optimal market making
- Key idea
- Quotes around a reservation price set by inventory, volatility, order-flow intensity, and risk aversion
- Key challenge
- Inventory control, addressed with tighter inventory-penalty terms
- Result
- 3rd place overall