why scalping backtests look great and then die live

[Shadow Trader]

Backtest says a profit factor of 1.8. Live says break-even at best. The gap usually isn't a mystery. It's a set of boring data problems that hit scalpers harder than anyone, because the edge per trade is a few pips.

- Tick data quality. Many platforms build default history from one-minute bars with modelled ticks in between. The path inside a candle is guesswork, and that's exactly where scalp stops and targets get decided.
- Fixed spread. A lot of tests assume one spread all day. Real spreads widen around rollover, news and thin Asian hours, and a 5-pip target feels every pip of it.
- No slippage or rejections. In the test, fills happen at the exact price. Live, stops fill worse and limit orders sometimes touch without filling.
- Bid-only data. Candles are usually built from bid prices, so long entries need an ask-side adjustment that's easy to forget.
- Over-fitting. Tune eight settings on one year and the result describes that year.

Scale of the problem: if the test spread is 0.6 and the live average is 1.1, that half pip is 10% of a 5-pip target on every single trade. A thin edge can vanish from that alone.

A scalper can get closer with real tick data from a reputable source, a variable-spread feed, a deliberate slippage penalty, and a test period the settings never saw. Even then it's an estimate.

Forward testing on demo or tiny live size helps more than another backtest, though demo fills are often gentler than live ones, depending on the broker.

What's the biggest gap you found between backtest and live results, and which of these turned out to be the cause?