Latency Is a Tax on Reactive Trading
The faster the move, the more your click becomes a donation.
Data
Backtests, feeds, bias, and the lies your dataset tells.
Backtests Underestimate Slippage Twice
Backtests pay the spread once and assume clean fills. Live trading pays it twice, plus a tax you didn’t model.
Slippage Is a Regime, Not an Error
Slippage clusters. If you treat it as rare, your risk model is fiction.
Filters Shrink Sample Size and Inflate Confidence
Every extra filter can make a backtest look smarter while making the result less trustworthy by starving the sample size.
Backtests Are Biographies, Not Prophecies
A backtest describes a past environment; it does not guarantee the future microstructure will keep paying you the same way.
Parameter Sensitivity Reveals Fragility
If tiny parameter tweaks flip results, you didn’t find structure, you found noise with a good story.
Your Edge Might Be a Data Artifact
If your edge only exists on one feed, one session cutoff, or one candle construction, it’s not an edge, it’s an artifact.
Alpha Often Lives in Time-of-Day, Not Pattern
Many ‘patterns’ disappear when you control for time-of-day, because the real driver is liquidity cycles, not chart magic.
The Cleanest Backtest Is Often the Most Lying
A backtest that looks unusually smooth usually means you removed the messy friction that will show up live and break expectancy.
Data Is Not Neutral
Your ‘edge’ can be an artifact of data choices: feeds, session cutoffs, revisions, survivorship, and cleaning rules.
Paper Trading Is Polite Fiction
Paper trading removes the two things that break strategies: execution friction and human fear.
Survivorship Bias Is a Hall of Mirrors
You mostly see winners because losers disappear, so your ‘base rates’ are fantasy.
Regimes Kill Models
Most strategies are secretly ‘one-regime strategies’ pretending to be universal.