Forex Optimization: Fix the Feed, Session and Cost Model

An FX backtest needs more than a pair name. Record the TradingView provider, symbol, quote precision, account currency, date range and session rule. A chart feed is not a guarantee of the bid, ask, fill or financing conditions offered by the intended broker. Keep those assumptions visible when comparing parameter sets. Define session windows by their intended local timezone rather than assuming a London–New York overlap always has the same UTC hours. Daylight-saving transitions can change those offsets. Pine’s chart display timezone does not itself change the script’s time calculations; use and verify the strategy’s actual timezone rule. Session filtering must already exist in the strategy if you want an exposed input to control it. Pineify does not add that rule automatically. For a hypothetical cost check, suppose a 100,000-unit EURUSD trade incurs a full bid–ask spread of 1 pip, defined here as 0.0001 USD per EUR. That spread corresponds to $10 before commission, financing or additional slippage. Do not count it twice if the model already uses separate bid and ask execution prices. Evaluate cost sensitivity before interpreting a small gross edge as an improvement. An optimization summary without individual entry/exit timestamps and position sizes is insufficient to reconstruct historical swap charges.