On this page
Four families of automated strategies How interval grids actually behave Averaging ladders and safety orders Statistical pullback approaches Breakout entries from defined levels Starting configurations worth keeping How these systems fail — a written recordFour families of automated strategies
Every retail trading bot in the wild descends from a small set of archetypes: trend-continuation, statistical pullback, relative-value spreads, and liquidity provision. This archive keeps field notes on the two archetypes that dominate deployed systems — and within those, the four patterns you will meet most often: interval grids, averaging ladders, oscillator pullbacks, and range breakouts.
| Pattern | Underlying assumption | Observed win rate | Dominant failure mode |
|---|---|---|---|
| Interval grid | Price keeps crossing a range | sixty to seventy-five percent | Inventory stacks up in a trend |
| Averaging ladder | Mean entry will be recovered | Depends on ladder depth | Asset never comes back |
| Oscillator pullback | Stretched values snap back | fifty-five to sixty-five percent | A trend makes stretching permanent |
| Range breakout | Momentum carries past a level | thirty-five to forty-five percent | Failed breaks and fake-outs |
How interval grids actually behave
A grid engine posts paired limit orders on both sides of a reference price. When one side fills, the engine immediately places a counter-order one interval away. Earnings come from the interval itself, not from direction.
The three decisions that matter: interval width (fixed percent or volatility-scaled), level count (this sets the worst-case inventory), and a hard inventory cap.
A starting configuration
grid: spacing_mode: atr # fixed, percent, or atr spacing_value: 0.5 # percent, or ATR multiplier levels_above: 6 levels_below: 6 order_size: 0.5 percent # of portfolio per level inventory_cap: 8 # max one-sided levels stop_loss: 12 percent # emergency exit
Averaging ladders and safety orders
Ladder bots open a base position and schedule additional entries at fixed intervals below the first fill. Later steps are usually sized up — the classic safety-order pattern — which drags the average entry down quickly. The take-profit is measured from the averaged price, not the first entry.
Statistical pullback approaches
Pullback engines watch a statistical distance — RSI, z-score against a moving mean, or band width — and enter against the stretch, exiting near the mean. They earn their keep in range-bound, persistently volatile markets.
They lose it when a regime change turns a stretched price into the new resting level. Regime filters (trend strength, volatility bands) belong in the same file as the entry logic.
Breakout entries from defined levels
Breakout engines act when price crosses a marked level — prior extreme, trendline, volatility band — and require confirmation, usually a close beyond the level or a volume condition. Exits trail or take fixed targets; they do not wait for reversal signals.
Starting configurations worth keeping
Every pattern has a canonical shape. Keeping templates makes backtests reproducible and parameters auditable.
shared_base:
risk:
max_portfolio_risk: 2 percent
max_drawdown_stop: 15 percent
max_concurrent_positions: 3
execution:
order_type: limit
post_only: true
time_in_force: gtc
monitoring:
heartbeat_interval: 60s
report_interval: 1h
How these systems fail — a written record
Most automated strategies die from one of four causes: a regime shift the entry logic cannot see, parameter drift after over-tuning, infrastructure failure at the exact wrong moment, or inventory that can no longer be unwound. Each topic page in this archive ends with a failure section for exactly that reason.
Browse the full topic index for notes on specific indicators, engines, and operational practices.