Images from ParamScape3D
Report example from TS-ParamScapeRank
ParamScapeRank Analysis Report
Objective: All:NetProfit
Purpose:
This report grades each input from 1 to 10 based on how useful it appears to be for the selected objective. Higher grades are better candidates for further investigation. Decimal-capable inputs may be tested at finer fractional resolution; integer-only inputs should be refined with whole-number steps, narrowed ranges, shifted ranges, or held static when already at step 1.
1. PS3D_VolMom:AvgLen
Grade:
10 out of 10. This is the report-card score for how much attention this input deserves in the next optimization pass.
Recommendation:
High priority
Analysis:
Keep varying this input, but refine it with smaller whole-number steps only. This parameter is marked Integer Based Only, so fractional values should not be used. The plateau score suggests the exact value may be forgiving, so a range may matter more than a single perfect setting.
Result Notes:
Strong integer-only candidate: this input matters, but refinement should use smaller whole-number steps only. Broad plateau: several nearby values appear to work similarly, so the exact setting may be less fragile. Integer-only override: this input is being treated as whole-number based, so suggested center, low, high, and step values stay on integer values.
Evidence:
Composite 1.00; RF 1.00; Corr 1.00; Plateau 0.60; Knee 0.41; Type Integer Based Only; Tested min step 2.
2. PS3D_VolMom:MomLen
Grade:
5 out of 10. This is the report-card score for how much attention this input deserves in the next optimization pass.
Recommendation:
Moderate priority
Analysis:
Keep it in the sweep, but use a practical range. If search space is tight, refine this after the high-priority inputs.
Result Notes:
Finest integer resolution: this input is marked integer-only and the step is already 1, so the next improvement must come from narrowing, shifting, or freezing the range rather than using smaller decimal steps.
Evidence:
Composite 0.47; RF 0.40; Corr 0.57; Plateau 0.50; Knee 0.24; Type Integer Based Only; Tested min step 1.
3. PS3D_VolMom:ATRPcnt
Grade:
3 out of 10. This is the report-card score for how much attention this input deserves in the next optimization pass.
Recommendation:
Static candidate
Analysis:
Consider holding this input static near a sensible value so you can spend more combinations on stronger inputs. It does show a possible bend, so review it before locking it down permanently.
Result Notes:
Possible bend point: performance may change quickly around part of this range, so a focused pass can help locate that turn.
Evidence:
Composite 0.22; RF 0.20; Corr 0.25; Plateau 0.10; Knee 0.90; Type Decimal resolution allowed; Tested min step 0.025.
4. PS3D_VolMom:ATRLen
Grade:
1 out of 10. This is the report-card score for how much attention this input deserves in the next optimization pass.
Recommendation:
Static candidate
Analysis:
Consider holding this input static near a sensible value so you can spend more combinations on stronger inputs.
Result Notes:
Safe-to-freeze candidate: this input did not show much evidence that changing it improves the selected objective. Finest integer resolution: this input is marked integer-only and the step is already 1, so the next improvement must come from narrowing, shifting, or freezing the range rather than using smaller decimal steps.
Evidence:
Composite 0.00; RF 0.00; Corr 0.00; Plateau 0.50; Knee 0.30; Type Integer Based Only; Tested min step 1.
Suggested Ranges Report
Suggested Range Explanation
Parameter: PS3D_VolMom:AvgLen
Objective: All:NetProfit
Suggested Range:
Use 22.000 as the center value, then test from 16.000 to 30.000, using a step of 2.000.
Plain-English Reason:
The strongest runs did not clearly press against either edge of the tested range. The suggested range stays inside the tested range because the best evidence appears to be within the values already explored. This input is marked Integer Based Only, so the app will only suggest whole-number center, low, high, and step values.
How to Use This:
If this parameter received a strong grade in the Analysis report, give this range more resolution when the parameter type allows it. For integer-only inputs already at step 1, use the report to narrow the range, shift the range, or decide whether to hold the parameter static.
Trend Scope:
This trend is local to the currently loaded optimization file. A broad full-search run may point one way, while a later reduced-search run may point the other way after it brackets a peak or enters a narrower neighborhood. Treat that as a signal to confirm the range, not as an automatic contradiction.
Result-Specific Notes:
– Local upward pressure: higher tested values generally looked better inside the currently loaded range.
– Integer-only override: this input is being treated as whole-number based, so suggested center, low, high, and step values stay on integer values.
Technical Details:
Tested Range: 10.000 to 30.000.
Edge Pressure Low: 0.01. This is the share of elite runs near the tested low boundary.
Edge Pressure High: 0.31. This is the share of elite runs near the tested high boundary.
Sensitivity: 0.41. Higher means the objective changed more as this input changed.
Trend: increasing. Monotonic increasing means the objective generally improved as this input got larger inside the currently loaded search range.
Projected Values: 8. This is how many values Low/High/Step should generate before any platform-specific rounding.
Audit Trail:
Elite 10% -> next range [16, 30], tested [10, 30], within tested range, minDelta=2; sens=0.41, mono=increasing, edgeLow=0.01, edgeHigh=0.31, fractionalStep=False, integerOnly=True, finestIntegerResolution=False, lowBoundaryOk=False, highBoundaryOk=True, impScale=1.10, sensScale=1.11, gridAligned=False, lowBoundaryGuard=False
