The simulation engine strategies must clear
Before an idea is allowed near the order path it has to survive a simulator that fills on the next bar, pays real Indian charges by product class, and refuses to look at data it could not have seen.
- strategies to live
- 8
Built here.
- 01
A configuration-driven engine: a declarative contract describes instruments, entries, legs, exits and lifecycle rules, so a new strategy is a config rather than a fork.
- 02
A bar-by-bar lifecycle engine holding portfolio, strategy and group state — linked exits, hedge injection, re-entry guards, end-of-day flatten, drawdown kill.
- 03
A fill model where the default is the next bar and the clock runs signal → route → exchange acknowledgement → broker acknowledgement → fill.
- 04
Cost factories by instrument class: securities transaction tax, exchange and regulatory charges, brokerage, stamp duty and GST are not one rate across options, cash and futures.
- 05
Greeks, strike selection and margin heuristics for multi-leg option structures, plus walk-forward selection that freezes training-window decisions before they touch test sessions.
- 06
Equity and futures paths through the same engine, so the desk is not maintaining three simulators.
Failure modes.
- Lookahead, which is treated as a defect rather than a modelling choice.
- Sparse or missing option chains at a requested strike, which fail loudly instead of substituting a made-up price.
- Overfitting through a walk-forward split where the training window's decisions are frozen before test sessions are scored.
Most ideas die here, and that is the engine working. Eight strategies cleared it into live capital — intraday index options and overnight stock options.