Strategy-Specific Stop Modeling
How to determine stop-loss distance using drawdown and persistence distributions instead of ATR.
1 — Inputs Required for Strategy-Specific Stops
Every quantified strategy must define the following before stops can be modeled:
-
Entry Trigger Conditions The Boolean rules that determine when a trade begins.
-
Expected Hold Horizon (bars or minutes) The minimum time window during which your strategy has a positive statistical edge.
-
Distribution of Maximum Drawdown over that hold horizon Measured in USD or pips based on the strategy’s historical sample.
-
Indicator Persistence Distribution How long the statistical-edge condition(s) remain satisfied after entry.
This is not necessarily the same as the entry condition.
- Entry Conditions define when to open a trade.
- Statistical-Edge Conditions define why the trade maintains expectancy after entry.
In simple strategies, the same indicator may serve both roles (e.g., SMA slope), but conceptually they are distinct:
- Entry tells you when to act.
- Persistence tells you how long the edge lasts.
Your stop-loss must be based on persistence behavior, not merely on the entry trigger.
-
Regime Filters Volatility compression/expansion, funding environment, trend alignment, etc., which influence drawdown shape and persistence duration.
These inputs determine the stop distance using the strategy’s own behavior, not ATR.
2 — Example Strategy: SMA Slope Persistence
Entry Conditions
entry_signal = [
"14-period 5-min SMA slope ≥ 0.10",
"Condition B",
"Condition C"
]
Expected Hold Horizon
expected_hold_horizon = 20 minutes
(= 4 bars on a 5-minute chart)
Slope Persistence Distribution
Measured from the entry bar's slope value:
| Bars After Entry | P(Slope ≥ 0.08) |
|---|---|
| 4 bars | 67% |
| 5 bars | 54% |
| 6 bars | 52% |
Interpretation:
- The strategy’s edge reliably persists for ~4 bars.
- Probability decays afterward.
- The stop must be wide enough to survive the strategy’s normal volatility within these 4 bars.
3 — Drawdown Distribution (in USD % or pips)
Drawdowns measured relative to the close of the entry bar:
1 bar forward
-1σ : 10 pips
median : 40 pips
+1σ : 100 pips
2 bars forward
-1σ : 17 pips
median : 52 pips
+1σ : 175 pips
3 bars forward
-1σ : 35 pips
median : 67 pips
+1σ : 225 pips
This represents the strategy’s unique volatility profile, which can differ dramatically from generic ATR volatility.
4 — Stop Distance = Drawdown Envelope for Expected Hold
Because expected hold = 4 bars, we must choose a stop that survives the strategy’s normal unfavorable excursions during this window.
This is never ATR-based.
Correct logic:
max_expected_dd_usd =
max(dd_1bar_usd, dd_2bar_usd, dd_3bar_usd, ...)
Using +1σ as a conservative envelope:
max_expected_dd_usd = 225 pips × pip_value_usd
This becomes the strategy-specific stop distance for this entry.
ATR may be used as a sanity check but never as the primary driver.
5 — Why Strategy-Specific Stops Are Superior
These stops are built from:
- real drawdown behavior
- true indicator persistence
- edge duration
- volatility regime
- structural signal characteristics
—not from generic volatility measures.
They:
- shrink during volatility compression
- widen when the strategy historically requires space
- match actual distributional behavior
- avoid violating the statistical edge
- prevent premature exits
- resist noise but respect structural boundaries
This is how quant systems survive their own expected behavior.
6 — Position Sizing with Strategy-Specific Stops
Once stop_distance_usd is known:
position_size_sol = MLPT_usd / stop_distance_usd
This ensures:
- MLPT stays constant
- position sizing adapts to strategy volatility
- account health remains stable
- size is correct regardless of SOL price
Position sizing is always the last step.
7 — When to Use Strategy-Specific Stops vs. ATR
Use strategy-specific stops when:
- drawdown behavior is measurable
- persistence behavior is known
- expected hold horizon is defined
- volatility is strategy-dependent
Use ATR stops when:
- strategy is discretionary
- no persistence model exists
- conditions are not fixed
- hold horizon is undefined
ATR is the fallback. Distributional stops are the primary method for any quantified system.
8 — Final Rule
**When a strategy has measurable drawdown/persistence behavior,
use that distribution — not ATR — to set the stop.**
ATR is optional. The distribution is mandatory.
If you'd like:
- front matter for this doc
- a matching diagram
- cross-links inserted to your other articles
…I can add those.