Add signal quality version comparison to analytics dashboard
- Created /api/analytics/version-comparison endpoint - Shows performance metrics for v1, v2, v3 scoring logic - Compares: trade count, win rate, P&L, quality scores, MFE/MAE - Special focus on extreme positions (< 15% or > 85% range) - Tracks weak ADX count (< 18) for each version - Visual indicators for current version (v3) - Data collection progress notice for v3 (need 20+ trades) - Legend explaining MFE, MAE, extreme positions, weak ADX Enables data-driven optimization by comparing algorithm performance with clean, version-tagged datasets.
This commit is contained in:
mindesbunister
124
docs/analysis/SIGNAL_QUALITY_VERSION_ANALYSIS.sql
Normal file
124
docs/analysis/SIGNAL_QUALITY_VERSION_ANALYSIS.sql
Normal file
@@ -0,0 +1,124 @@
|
||||
-- Signal Quality Version Analysis
|
||||
-- Compare performance between different scoring logic versions
|
||||
|
||||
-- 1. Count trades by version
|
||||
SELECT
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
COUNT(*) as trade_count,
|
||||
ROUND(100.0 * COUNT(*) / SUM(COUNT(*)) OVER(), 1) as percentage
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL
|
||||
GROUP BY "signalQualityVersion"
|
||||
ORDER BY version;
|
||||
|
||||
-- 2. Performance by version
|
||||
SELECT
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
COUNT(*) as trades,
|
||||
ROUND(AVG("realizedPnL")::numeric, 2) as avg_pnl,
|
||||
ROUND(SUM("realizedPnL")::numeric, 2) as total_pnl,
|
||||
ROUND(100.0 * SUM(CASE WHEN "realizedPnL" > 0 THEN 1 ELSE 0 END) / COUNT(*)::numeric, 1) as win_rate,
|
||||
ROUND(AVG("signalQualityScore")::numeric, 1) as avg_quality_score,
|
||||
ROUND(AVG("maxFavorableExcursion")::numeric, 2) as avg_mfe,
|
||||
ROUND(AVG("maxAdverseExcursion")::numeric, 2) as avg_mae
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL AND "exitReason" NOT LIKE '%CLEANUP%'
|
||||
GROUP BY "signalQualityVersion"
|
||||
ORDER BY version;
|
||||
|
||||
-- 3. Version breakdown by exit reason
|
||||
SELECT
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
"exitReason",
|
||||
COUNT(*) as count,
|
||||
ROUND(AVG("realizedPnL")::numeric, 2) as avg_pnl
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL AND "exitReason" NOT LIKE '%CLEANUP%'
|
||||
GROUP BY "signalQualityVersion", "exitReason"
|
||||
ORDER BY version, count DESC;
|
||||
|
||||
-- 4. Quality score distribution by version
|
||||
SELECT
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
CASE
|
||||
WHEN "signalQualityScore" >= 80 THEN '80-100 (High)'
|
||||
WHEN "signalQualityScore" >= 70 THEN '70-79 (Good)'
|
||||
WHEN "signalQualityScore" >= 60 THEN '60-69 (Pass)'
|
||||
ELSE '< 60 (Block)'
|
||||
END as score_range,
|
||||
COUNT(*) as trades,
|
||||
ROUND(AVG("realizedPnL")::numeric, 2) as avg_pnl,
|
||||
ROUND(100.0 * SUM(CASE WHEN "realizedPnL" > 0 THEN 1 ELSE 0 END) / COUNT(*)::numeric, 1) as win_rate
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL AND "signalQualityScore" IS NOT NULL
|
||||
GROUP BY "signalQualityVersion", score_range
|
||||
ORDER BY version, score_range DESC;
|
||||
|
||||
-- 5. Extreme position entries by version (< 15% or > 85%)
|
||||
SELECT
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
direction,
|
||||
COUNT(*) as trades,
|
||||
ROUND(AVG("pricePositionAtEntry")::numeric, 1) as avg_price_pos,
|
||||
ROUND(AVG("adxAtEntry")::numeric, 1) as avg_adx,
|
||||
ROUND(AVG("realizedPnL")::numeric, 2) as avg_pnl,
|
||||
ROUND(100.0 * SUM(CASE WHEN "realizedPnL" > 0 THEN 1 ELSE 0 END) / COUNT(*)::numeric, 1) as win_rate
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL
|
||||
AND "pricePositionAtEntry" IS NOT NULL
|
||||
AND ("pricePositionAtEntry" < 15 OR "pricePositionAtEntry" > 85)
|
||||
GROUP BY "signalQualityVersion", direction
|
||||
ORDER BY version, direction;
|
||||
|
||||
-- 6. Recent v3 trades (new logic)
|
||||
SELECT
|
||||
"createdAt",
|
||||
symbol,
|
||||
direction,
|
||||
"entryPrice",
|
||||
"exitPrice",
|
||||
"exitReason",
|
||||
ROUND("realizedPnL"::numeric, 2) as pnl,
|
||||
"signalQualityScore" as score,
|
||||
ROUND("adxAtEntry"::numeric, 1) as adx,
|
||||
ROUND("pricePositionAtEntry"::numeric, 1) as price_pos
|
||||
FROM "Trade"
|
||||
WHERE "signalQualityVersion" = 'v3'
|
||||
AND "exitReason" IS NOT NULL
|
||||
ORDER BY "createdAt" DESC
|
||||
LIMIT 20;
|
||||
|
||||
-- 7. Compare v3 vs pre-v3 on extreme positions
|
||||
WITH version_groups AS (
|
||||
SELECT
|
||||
CASE WHEN "signalQualityVersion" = 'v3' THEN 'v3 (NEW)' ELSE 'pre-v3 (OLD)' END as version_group,
|
||||
*
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL
|
||||
AND "pricePositionAtEntry" IS NOT NULL
|
||||
AND ("pricePositionAtEntry" < 15 OR "pricePositionAtEntry" > 85)
|
||||
AND "adxAtEntry" IS NOT NULL
|
||||
)
|
||||
SELECT
|
||||
version_group,
|
||||
COUNT(*) as trades,
|
||||
ROUND(AVG("adxAtEntry")::numeric, 1) as avg_adx,
|
||||
COUNT(*) FILTER (WHERE "adxAtEntry" < 18) as weak_adx_count,
|
||||
ROUND(AVG("realizedPnL")::numeric, 2) as avg_pnl,
|
||||
ROUND(SUM("realizedPnL")::numeric, 2) as total_pnl,
|
||||
ROUND(100.0 * SUM(CASE WHEN "realizedPnL" > 0 THEN 1 ELSE 0 END) / COUNT(*)::numeric, 1) as win_rate
|
||||
FROM version_groups
|
||||
GROUP BY version_group
|
||||
ORDER BY version_group DESC;
|
||||
|
||||
-- 8. Daily performance by version (last 7 days)
|
||||
SELECT
|
||||
DATE("createdAt") as trade_date,
|
||||
COALESCE("signalQualityVersion", 'v1/null') as version,
|
||||
COUNT(*) as trades,
|
||||
ROUND(SUM("realizedPnL")::numeric, 2) as daily_pnl
|
||||
FROM "Trade"
|
||||
WHERE "exitReason" IS NOT NULL
|
||||
AND "createdAt" >= NOW() - INTERVAL '7 days'
|
||||
GROUP BY trade_date, "signalQualityVersion"
|
||||
ORDER BY trade_date DESC, version;
|
||||
Reference in New Issue
Block a user