All systems / Quant Prototype
Quant Prototype v2
A 7-step institutional-style process prototype, tested on real 1-minute bid/ask data.
Research / add-on
USDJPYXAUUSDGBPUSDNAS100M1 decisions · hourly holds
How it works
Quant Prototype is a learning project on how institutions design short-term systems. It is tested on real minute-by-minute bid/ask data with retail costs deducted.
- Key finding: minute-level holding is negative in every variant after real costs
- The edge exists only in some markets. With one setting across 14 markets, PF is about 1.0.
- Use it only as an add-on to Ichimoku, never on its own
Latest version v2
Updated 2026-10-06Decide every minute, hold for hours
- Signal → Forecast → Cost check → Risk sizing → Execution → Real-time risk → Post-trade
- 24-hour Donchian + D1/H4 trend + H1 squeeze · trades 07–20 UTC · flat before the weekend
- Markets chosen on 2019–22 data gave 4 markets; the 2023–26 check period has PF 1.18
Which timeframe
Best: M1 decisions · hourly holdsM1
Research result: minute-level holding is negative after real costs in every variant. Holds must be hourly or longer.
Why this many trades: Many trades because it trades 4 markets. (14.46 trades/month)
Version history
Latest version results (v2)
Hypothetical 2019-01-08 – 2026-10-02 · Compounded, 0.5% risk per trade · portfolio rules not applied (max 4 open, 2% daily stop) · Source: Python engine · Dukascopy 1-minute bid/ask 2019–2026 · retail costs deducted
Avg. return per year (compounded)
4.9%
Max drawdown
10.9%
Profit factor
1.13
Win rate
52.4%
Longest losing streak
8 trades
Trades
1,342
14.46 trades/month
Worst year
-7.8%
in 2019
Losing years
2 years
of 8 years
Profitable months
52.1%
Longest drawdown
574 days
Median hold
18.8h