[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8},"actionData","errors","root",{"_106":107,"_108":109},"routes/embed.x.$handle.$name.$",{"_9":10,"_11":12,"_13":14,"_15":16,"_17":18,"_19":20,"_21":-5},"handle","Tringelar","name","binance-arb-paper-bot","path","strategy.ts","project",{"_22":45,"_11":12,"_33":46,"_47":-5,"_48":42,"_49":50,"_51":50,"_52":53,"_54":-5,"_55":56,"_57":58,"_59":32,"_60":-5,"_61":-5,"_62":63,"_64":65,"_66":67,"_68":69,"_70":71,"_72":-5,"_73":74,"_75":74,"_76":77,"_78":77,"_79":77,"_80":74,"_81":74,"_82":83,"_84":85,"_86":87,"_88":42},"file",{"_22":23,"_11":14,"_24":25,"_26":-5,"_27":28,"_29":30,"_31":32,"_33":34,"_35":36,"_37":38,"_39":40,"_41":42,"_43":-5,"_13":44},"permalink","https://www.val.town/x/Tringelar/binance-arb-paper-bot/code/strategy.ts","extension","id","3b1b3269-2406-4553-bf15-255a2f42a0f1","fileId","8d75c760-a6fe-11f1-9e37-1607ee4eb77e","parentId","branchId","706a186d-17b0-49d7-afcd-061e66763458","branchVersion",62,"latest",true,"createdAt",["D",1788380356583],"version",21,"fileType","script","content","// ============================================================\n// Multi-Timeframe Technical Analysis Strategy Engine\n// Systematic Trading Strategy (ICT / SMC Hybrid)\n// ------------------------------------------------------------\n// Pipeline:\n// 1. Trend bias on 4H (EMA50/EMA200 + swing structure)\n// 2. POI detection on 1H/15M (Order Blocks + FVG)\n// 3. Execution trigger on 5M (reversal candle at POI retest)\n// 4. Risk management (SL behind structure, TP1/TP2 at 1:1 / 1:2)\n// 5. Adaptive learning from closed-trade outcomes\n// ============================================================\n\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\n\nconst BINANCE = \"https://data-api.binance.vision/api/v3\";\n\n// ---------- Types ----------\nexport interface Candle {\n time: number;\n open: number;\n high: number;\n low: number;\n close: number;\n volume: number;\n}\n\nexport type Side = \"LONG\" | \"SHORT\";\n\nexport interface Signal {\n symbol: string;\n side: Side;\n entry: number;\n stopLoss: number;\n takeProfit1: number;\n takeProfit2: number;\n rr: number; // reward:risk to TP2\n confidence: number; // 0-100 confluence score\n reason: string;\n timeframe: string; // POI timeframe\n timestamp: number;\n // Learning features (used to estimate historical win probability).\n poiType: \"OB\" | \"FVG\";\n trendType: Trend;\n liquidity: \"HIGH\" | \"MEDIUM\";\n}\n\nexport interface LiquidityCoin {\n symbol: string; // e.g. \"BTCUSDT\"\n base: string; // e.g. \"BTC\"\n quoteVolume: number; // 24h USDT volume\n liquidity: \"HIGH\" | \"MEDIUM\";\n}\n\n// ---------- Config ----------\nexport const LIQUIDITY_HIGH_MIN = 50_000_000; // >= $50M/day => HIGH\nexport const LIQUIDITY_MEDIUM_MIN = 5_000_000; // >= $5M/day => MEDIUM\nexport const MAX_COINS_PER_SCAN = 80;\n// Paper bet size per position — used to value open positions when computing\n// floating (unrealized) P&L. Kept in sync with strategy-bot.ts.\nexport const POSITION_SIZE_USDT = 100;\nconst STABLE_BASES = new Set([\n \"USDT\", \"USDC\", \"BUSD\", \"DAI\", \"TUSD\", \"FDUSD\", \"USDP\", \"EUR\", \"AEUR\",\n \"USTC\", \"USDD\", \"PYUSD\", \"EURI\", \"XUSD\", \"USD1\",\n]);\n\n// ---------- Binance data ----------\nexport async function fetchKlines(\n symbol: string,\n interval: string,\n limit = 200,\n): Promise {\n const res = await fetch(\n `${BINANCE}/klines?symbol=${symbol}&interval=${interval}&limit=${limit}`,\n );\n if (!res.ok) throw new Error(`klines ${symbol} ${interval} ${res.status}`);\n const raw = (await res.json()) as unknown[][];\n return raw.map((k) => ({\n time: k[0] as number,\n open: parseFloat(k[1] as string),\n high: parseFloat(k[2] as string),\n low: parseFloat(k[3] as string),\n close: parseFloat(k[4] as string),\n volume: parseFloat(k[5] as string),\n }));\n}\n\n// ---------- Liquidity universe ----------\n// Scan medium & high liquidity coins from the 24h ticker.\nexport async function getLiquidCoins(): Promise {\n const res = await fetch(`${BINANCE}/ticker/24hr`);\n if (!res.ok) throw new Error(`24hr ${res.status}`);\n const all = (await res.json()) as {\n symbol: string;\n quoteVolume: string;\n }[];\n\n const coins: LiquidityCoin[] = [];\n for (const t of all) {\n if (!t.symbol.endsWith(\"USDT\")) continue;\n const base = t.symbol.slice(0, -4);\n if (STABLE_BASES.has(base)) continue;\n const qv = parseFloat(t.quoteVolume);\n if (!Number.isFinite(qv) || qv <= 0) continue;\n if (qv >= LIQUIDITY_HIGH_MIN) {\n coins.push({ symbol: t.symbol, base, quoteVolume: qv, liquidity: \"HIGH\" });\n } else if (qv >= LIQUIDITY_MEDIUM_MIN) {\n coins.push({ symbol: t.symbol, base, quoteVolume: qv, liquidity: \"MEDIUM\" });\n }\n }\n\n // Sort by volume desc, take the top N (mix of medium & high).\n coins.sort((a, b) => b.quoteVolume - a.quoteVolume);\n return coins.slice(0, MAX_COINS_PER_SCAN);\n}\n\n// ---------- Indicators ----------\nexport function ema(values: number[], period: number): number[] {\n const k = 2 / (period + 1);\n const out: number[] = [];\n let prev = values[0];\n for (let i = 0; i < values.length; i++) {\n if (i === 0) {\n out.push(values[0]);\n prev = values[0];\n } else {\n prev = values[i] * k + prev * (1 - k);\n out.push(prev);\n }\n }\n return out;\n}\n\nexport function atr(candles: Candle[], period = 14): number {\n if (candles.length < period + 1) return 0;\n let sum = 0;\n for (let i = 1; i <= period; i++) {\n const c = candles[i];\n const p = candles[i - 1];\n const tr = Math.max(\n c.high - c.low,\n Math.abs(c.high - p.close),\n Math.abs(c.low - p.close),\n );\n sum += tr;\n }\n return sum / period;\n}\n\n// ---------- Swing structure ----------\ninterface Swing {\n index: number;\n price: number;\n type: \"high\" | \"low\";\n}\n\n// Fractal swing detection: a candle is a swing high if its high is the max of\n// the surrounding `left` candles on each side.\nexport function findSwings(candles: Candle[], left = 2): Swing[] {\n const swings: Swing[] = [];\n for (let i = left; i < candles.length - left; i++) {\n let isHigh = true;\n let isLow = true;\n for (let j = i - left; j <= i + left; j++) {\n if (j === i) continue;\n if (candles[j].high >= candles[i].high) isHigh = false;\n if (candles[j].low <= candles[i].low) isLow = false;\n }\n if (isHigh) swings.push({ index: i, price: candles[i].high, type: \"high\" });\n if (isLow) swings.push({ index: i, price: candles[i].low, type: \"low\" });\n }\n return swings;\n}\n\nexport type Trend = \"bullish\" | \"bearish\" | \"range\";\n\n// Trend bias from EMA50/EMA200 on the last candle + swing structure.\nexport function detectTrend(candles: Candle[]): Trend {\n if (candles.length < 200) return \"range\";\n const closes = candles.map((c) => c.close);\n const e50 = ema(closes, 50);\n const e200 = ema(closes, 200);\n const last = closes.length - 1;\n const price = closes[last];\n\n const swings = findSwings(candles, 3).filter((s) => s.index > last - 60);\n const highs = swings.filter((s) => s.type === \"high\");\n const lows = swings.filter((s) => s.type === \"low\");\n\n // Higher highs & higher lows => bullish; lower highs & lower lows => bearish.\n let hh = 0, hl = 0, lh = 0, ll = 0;\n for (let i = 1; i < highs.length; i++) {\n if (highs[i].price > highs[i - 1].price) hh++; else lh++;\n }\n for (let i = 1; i < lows.length; i++) {\n if (lows[i].price > lows[i - 1].price) hl++; else ll++;\n }\n\n const emaBull = price > e50[last] && e50[last] > e200[last];\n const emaBear = price < e50[last] && e50[last] < e200[last];\n\n // Strong sustained trend: EMA alignment confirmed by structure count.\n if (emaBull && hh + hl >= lh + ll - 2) return \"bullish\";\n if (emaBear && lh + ll >= hh + hl - 2) return \"bearish\";\n\n // Weaker but still directional: rely on EMA alignment alone.\n if (emaBull) return \"bullish\";\n if (emaBear) return \"bearish\";\n\n return \"range\";\n}\n\n// ---------- POI detection ----------\nexport interface POI {\n type: \"OB\" | \"FVG\";\n side: Side;\n top: number;\n bottom: number;\n index: number;\n}\n\n// Order Block: the last opposite candle before a strong directional move.\nexport function findOrderBlocks(candles: Candle[], lookback = 60): POI[] {\n const pois: POI[] = [];\n const start = Math.max(1, candles.length - lookback);\n for (let i = start; i < candles.length - 1; i++) {\n const move = candles[i + 1].close - candles[i].close;\n const body = Math.abs(candles[i].close - candles[i].open);\n // Strong up move preceded by a bearish candle => demand OB (LONG)\n if (move > 0 && candles[i].close < candles[i].open && body > 0) {\n pois.push({\n type: \"OB\",\n side: \"LONG\",\n top: Math.max(candles[i].open, candles[i].close),\n bottom: Math.min(candles[i].open, candles[i].close),\n index: i,\n });\n }\n // Strong down move preceded by a bullish candle => supply OB (SHORT)\n if (move < 0 && candles[i].close > candles[i].open && body > 0) {\n pois.push({\n type: \"OB\",\n side: \"SHORT\",\n top: Math.max(candles[i].open, candles[i].close),\n bottom: Math.min(candles[i].open, candles[i].close),\n index: i,\n });\n }\n }\n return pois;\n}\n\n// Fair Value Gap: gap between candle[i-1] and candle[i+1].\nexport function findFVG(candles: Candle[], lookback = 60): POI[] {\n const pois: POI[] = [];\n const start = Math.max(1, candles.length - lookback);\n for (let i = start; i < candles.length - 1; i++) {\n const prev = candles[i - 1];\n const next = candles[i + 1];\n // Bullish FVG: next.low > prev.high\n if (next.low > prev.high) {\n pois.push({\n type: \"FVG\",\n side: \"LONG\",\n top: next.low,\n bottom: prev.high,\n index: i,\n });\n }\n // Bearish FVG: next.high < prev.low\n if (next.high < prev.low) {\n pois.push({\n type: \"FVG\",\n side: \"SHORT\",\n top: prev.low,\n bottom: next.high,\n index: i,\n });\n }\n }\n return pois;\n}\n\n// ---------- Reversal candle detection (execution trigger) ----------\n// Returns true if the last candle is a valid reversal trigger for `side`.\nexport function isReversalTrigger(candles: Candle[], side: Side): boolean {\n if (candles.length < 3) return false;\n const c = candles[candles.length - 1];\n const p = candles[candles.length - 2];\n const body = Math.abs(c.close - c.open);\n const range = c.high - c.low;\n if (range === 0) return false;\n const lowerWick = Math.min(c.open, c.close) - c.low;\n const upperWick = c.high - Math.max(c.open, c.close);\n\n if (side === \"LONG\") {\n // Hammer: small body near top, long lower wick (>= 2x body)\n const hammer = body > 0 && lowerWick >= 2 * body && upperWick <= body;\n // Bullish engulfing: green candle engulfs previous red body\n const engulf = c.close > c.open && p.close < p.open &&\n c.close >= p.open && c.open <= p.close;\n return hammer || engulf;\n } else {\n // Hanging man: small body, long lower wick at top of range (bearish)\n const hangingMan = body > 0 && lowerWick >= 2 * body && c.close >= p.close;\n // Bearish engulfing\n const engulf = c.close < c.open && p.close > p.open &&\n c.close <= p.open && c.open >= p.close;\n return hangingMan || engulf;\n }\n}\n\n// ---------- Full signal generation for one coin ----------\nexport async function generateSignal(\n symbol: string,\n liquidity: \"HIGH\" | \"MEDIUM\" = \"MEDIUM\",\n): Promise {\n // Fetch the three timeframes we need.\n const [c4h, c1h, c5m] = await Promise.all([\n fetchKlines(symbol, \"4h\", 250),\n fetchKlines(symbol, \"1h\", 100),\n fetchKlines(symbol, \"5m\", 80),\n ]);\n\n const trend = detectTrend(c4h);\n if (trend === \"range\") return null; // no bias => no trade\n\n const last4h = c4h[c4h.length - 1];\n const price = last4h.close;\n const a = atr(c4h, 14);\n if (a <= 0) return null;\n\n // POIs from 1H (medium timeframe).\n const ob = findOrderBlocks(c1h, 60);\n const fvg = findFVG(c1h, 60);\n const pois = [...ob, ...fvg].filter((p) => p.index > c1h.length - 40);\n\n // Pick the side aligned with the trend.\n const side: Side = trend === \"bullish\" ? \"LONG\" : \"SHORT\";\n\n // Find a POI in the direction of the trend whose zone the price is near.\n const zone = pois\n .filter((p) => p.side === side)\n .map((p) => ({ ...p, dist: Math.abs(price - (p.top + p.bottom) / 2) }))\n .sort((x, y) => x.dist - y.dist)[0];\n\n if (!zone) return null;\n\n // Execution: price must be inside / at the POI zone AND a reversal trigger\n // must appear on 5M (retest confirmation).\n const inZone = price >= zone.bottom * 0.995 && price <= zone.top * 1.005;\n const trigger = isReversalTrigger(c5m, side);\n if (!inZone || !trigger) return null;\n\n // ---- Risk management ----\n // SL behind the nearest swing structure beyond the zone.\n const swings = findSwings(c4h, 2).filter((s) => s.index > c4h.length - 60);\n let stopLoss: number;\n if (side === \"LONG\") {\n const support = swings.filter((s) => s.type === \"low\" && s.price < zone.bottom)\n .sort((x, y) => y.price - x.price)[0];\n stopLoss = support ? support.price : zone.bottom - a * 0.5;\n stopLoss -= a * 0.1; // buffer vs stop hunts\n } else {\n const resistance = swings.filter((s) => s.type === \"high\" && s.price > zone.top)\n .sort((x, y) => x.price - y.price)[0];\n stopLoss = resistance ? resistance.price : zone.top + a * 0.5;\n stopLoss += a * 0.1;\n }\n\n const risk = Math.abs(price - stopLoss);\n if (risk <= 0) return null;\n\n // TP1 = 1:1, TP2 = 1:2 (measured move / R:R targets).\n let tp1: number, tp2: number;\n if (side === \"LONG\") {\n tp1 = price + risk;\n tp2 = price + risk * 2;\n } else {\n tp1 = price - risk;\n tp2 = price - risk * 2;\n }\n\n // R:R gate: reject trades below 1:2 to TP2.\n const rr = (Math.abs(tp2 - price)) / risk;\n if (rr < 2) return null;\n\n // Confluence score.\n let confidence = 40;\n if (zone.type === \"OB\") confidence += 15;\n if (ob.some((o) => o.side === side && o.index > c1h.length - 30)) confidence += 10;\n if (trend !== \"range\") confidence += 15;\n if (isReversalTrigger(c5m, side)) confidence += 10;\n\n return {\n symbol,\n side,\n entry: price,\n stopLoss,\n takeProfit1: tp1,\n takeProfit2: tp2,\n rr,\n confidence,\n reason: `${trend} trend | ${zone.type} ${side} @ ${zone.top.toFixed(6)}-${zone.bottom.toFixed(6)} | 5m trigger`,\n timeframe: \"1H\",\n timestamp: Date.now(),\n poiType: zone.type,\n trendType: trend,\n liquidity,\n };\n}\n\n// ---------- DB ----------\nexport async function initStrategyDb() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS strategy_wallet (\n id INTEGER PRIMARY KEY CHECK (id = 1),\n balance REAL NOT NULL DEFAULT 1000.0,\n initial_balance REAL NOT NULL DEFAULT 1000.0,\n open_positions INTEGER NOT NULL DEFAULT 0\n )`);\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS strategy_trades (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n symbol TEXT NOT NULL,\n side TEXT NOT NULL,\n entry REAL NOT NULL,\n stop_loss REAL NOT NULL,\n take_profit1 REAL NOT NULL,\n take_profit2 REAL NOT NULL,\n rr REAL NOT NULL,\n confidence INTEGER NOT NULL,\n reason TEXT NOT NULL,\n timeframe TEXT NOT NULL,\n status TEXT NOT NULL DEFAULT 'open',\n pnl REAL,\n opened_at INTEGER NOT NULL,\n closed_at INTEGER,\n poi_type TEXT,\n trend_type TEXT,\n liquidity TEXT\n )`);\n await sqlite.execute(`INSERT OR IGNORE INTO strategy_wallet (id) VALUES (1)`);\n await initLearningDb();\n\n // Migration: add learning-feature columns if this table was created by an\n // older version of the bot (ALTER TABLE fails if the column already exists,\n // so each is wrapped and ignored).\n const migs = [\n \"ALTER TABLE strategy_trades ADD COLUMN poi_type TEXT\",\n \"ALTER TABLE strategy_trades ADD COLUMN trend_type TEXT\",\n \"ALTER TABLE strategy_trades ADD COLUMN liquidity TEXT\",\n ];\n for (const m of migs) {\n try {\n await sqlite.execute(m);\n } catch {\n // column already exists — fine\n }\n }\n}\n\nexport async function getStrategyWallet() {\n const r = await sqlite.execute(\"SELECT * FROM strategy_wallet WHERE id = 1\");\n return r.rows[0] as unknown as {\n balance: number;\n initial_balance: number;\n open_positions: number;\n };\n}\n\n// Total unrealized (floating) P&L across all open positions, valued at the\n// latest close price. Returns the dollar amount, the average % (weighted by\n// position notional), and the number of open positions.\nexport async function getFloatingPnL(): Promise<{\n usd: number;\n pct: number;\n positions: number;\n}> {\n const res = await sqlite.execute(\n \"SELECT symbol, side, entry FROM strategy_trades WHERE status = 'open'\",\n );\n const open = res.rows as unknown as {\n symbol: string;\n side: string;\n entry: number;\n }[];\n let usd = 0;\n for (const t of open) {\n let price: number;\n try {\n const k = await fetchKlines(t.symbol, \"15m\", 1);\n price = k[0].close;\n } catch {\n continue; // skip if price fetch fails for this symbol\n }\n const pct = t.side === \"LONG\"\n ? (price - t.entry) / t.entry\n : (t.entry - price) / t.entry;\n usd += pct * POSITION_SIZE_USDT;\n }\n const notional = open.length * POSITION_SIZE_USDT;\n return {\n usd,\n pct: notional > 0 ? (usd / notional) * 100 : 0,\n positions: open.length,\n };\n}\n\nexport async function openTrade(sig: Signal) {\n await sqlite.execute(\n `INSERT INTO strategy_trades\n (symbol, side, entry, stop_loss, take_profit1, take_profit2, rr, confidence,\n reason, timeframe, opened_at, poi_type, trend_type, liquidity)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,\n [sig.symbol, sig.side, sig.entry, sig.stopLoss, sig.takeProfit1,\n sig.takeProfit2, sig.rr, sig.confidence, sig.reason, sig.timeframe, sig.timestamp,\n sig.poiType, sig.trendType, sig.liquidity],\n );\n await sqlite.execute(\n \"UPDATE strategy_wallet SET open_positions = open_positions + 1 WHERE id = 1\",\n );\n}\n\n// Close any open positions that have hit SL / TP1 / TP2 based on the latest price.\n// Also feeds every outcome into the adaptive learning system.\nexport async function manageOpenPositions(): Promise {\n const res = await sqlite.execute(\n \"SELECT * FROM strategy_trades WHERE status = 'open'\",\n );\n const open = res.rows as unknown as {\n id: number;\n symbol: string;\n side: string;\n entry: number;\n stop_loss: number;\n take_profit1: number;\n take_profit2: number;\n opened_at: number;\n poi_type: string;\n trend_type: string;\n liquidity: string;\n }[];\n\n let closed = 0;\n for (const t of open) {\n let price: number;\n try {\n const k = await fetchKlines(t.symbol, \"15m\", 1);\n price = k[0].close;\n } catch {\n continue;\n }\n let status: string | null = null;\n let pnlPct = 0;\n if (t.side === \"LONG\") {\n if (price >= t.take_profit2) { status = \"tp2\"; pnlPct = 2; }\n else if (price >= t.take_profit1) { status = \"tp1\"; pnlPct = 1; }\n else if (price <= t.stop_loss) { status = \"sl\"; pnlPct = -1; }\n } else {\n if (price <= t.take_profit2) { status = \"tp2\"; pnlPct = 2; }\n else if (price <= t.take_profit1) { status = \"tp1\"; pnlPct = 1; }\n else if (price >= t.stop_loss) { status = \"sl\"; pnlPct = -1; }\n }\n if (status) {\n await sqlite.execute(\n `UPDATE strategy_trades SET status = ?, pnl = ?, closed_at = ? WHERE id = ?`,\n [status, pnlPct, Date.now(), t.id],\n );\n await sqlite.execute(\n `UPDATE strategy_wallet SET open_positions = open_positions - 1,\n balance = balance + ? WHERE id = 1`,\n [100 * (pnlPct / 100)],\n );\n // Teach the learning system: a win is reaching at least TP1.\n const won = status === \"tp1\" || status === \"tp2\";\n await teachResult({\n poiType: t.poi_type ?? \"OB\",\n trendType: t.trend_type ?? \"bullish\",\n liquidity: t.liquidity ?? \"MEDIUM\",\n side: t.side,\n }, won);\n // Bump the closed-trade counter used to arm the learned gate.\n await sqlite.execute(\n `INSERT INTO strategy_meta (key, value) VALUES ('closed_trades', 1)\n ON CONFLICT(key) DO UPDATE SET value = value + 1`,\n );\n closed++;\n }\n }\n return closed;\n}\n\n// ============================================================\n// Adaptive Learning System\n// ------------------------------------------------------------\n// Learns from every closed trade. Each trade is tagged with its\n// factors (poi type, trend, liquidity, side). After we accumulate\n// enough samples, we compute a historical win-rate for each factor\n// and derive an acceptance gate: a new signal is opened only if its\n// weighted win probability clears the dynamic threshold.\n// ============================================================\n\n// Minimum closed trades before the learned gate kicks in.\nexport const LEARN_MIN_SAMPLES = 15;\n// How much a factor's positive history must beat break-even to be\n// considered an edge. With a ~1:1 to 1:2 payoff we want win rate > ~40%.\nconst WIN_RATE_FLOOR = 0.40;\n\nexport async function initLearningDb() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS strategy_factors (\n factor TEXT PRIMARY KEY,\n wins INTEGER NOT NULL DEFAULT 0,\n losses INTEGER NOT NULL DEFAULT 0\n )`);\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS strategy_meta (\n key TEXT PRIMARY KEY,\n value REAL\n )`);\n}\n\n// Record a win/loss against each factor tag of a trade.\nexport async function teachResult(\n features: { poiType: string; trendType: string; liquidity: string; side: string },\n won: boolean,\n) {\n const factors = [\n `poi:${features.poiType}`,\n `trend:${features.trendType}`,\n `liq:${features.liquidity}`,\n `side:${features.side}`,\n ];\n const res = await sqlite.execute(\n `SELECT factor FROM strategy_factors WHERE factor IN (${factors.map(() => \"?\").join(\",\")})`,\n factors,\n );\n const existing = new Set((res.rows as { factor: string }[]).map((r) => r.factor));\n const upserts = factors.filter((f) => !existing.has(f)).map((f) => [f]);\n if (upserts.length) {\n await sqlite.batch(upserts.map(([f]) => ({\n sql: `INSERT OR IGNORE INTO strategy_factors (factor, wins, losses) VALUES (?, 0, 0)`,\n args: [f],\n })));\n }\n const col = won ? \"wins\" : \"losses\";\n await sqlite.batch(factors.map((f) => ({\n sql: `UPDATE strategy_factors SET ${col} = ${col} + 1 WHERE factor = ?`,\n args: [f],\n })));\n}\n\n// Win rate for a single factor (0..1), or null if too few samples.\nasync function factorWinRate(factor: string): Promise {\n const r = await sqlite.execute(\n \"SELECT wins, losses FROM strategy_factors WHERE factor = ?\",\n [factor],\n );\n const row = r.rows[0] as { wins: number; losses: number } | undefined;\n if (!row) return null;\n const total = row.wins + row.losses;\n if (total < LEARN_MIN_SAMPLES) return null;\n return row.wins / total;\n}\n\n// Historical edge adjustment: returns a delta in confidence points built from\n// how each factor of the signal has performed historically.\nexport async function learnScoreAdjustment(sig: {\n poiType: string;\n trendType: string;\n liquidity: string;\n side: string;\n}): Promise {\n const factors = [\n `poi:${sig.poiType}`,\n `trend:${sig.trendType}`,\n `liq:${sig.liquidity}`,\n `side:${sig.side}`,\n ];\n let known = 0;\n let edge = 0;\n for (const f of factors) {\n const wr = await factorWinRate(f);\n if (wr === null) continue; // not enough data yet — no signal\n known++;\n // Edge is how far this factor's historical win rate beats the floor.\n edge += wr - WIN_RATE_FLOOR;\n }\n if (known === 0) return 0; // still in the learning/warm-up phase\n // Map cumulative edge into a bounded confidence-points offset.\n return Math.max(-15, Math.min(15, edge * 15));\n}\n\n// Adaptive acceptance decision. After enough history, reject a signal whose\n// learned win probability (adjusted confidence) fails the dynamic floor.\nexport async function shouldAccept(sig: {\n confidence: number;\n poiType: string;\n trendType: string;\n liquidity: string;\n side: string;\n}): Promise<{ accept: boolean; adjusted: number; reason: string }> {\n const adjustment = await learnScoreAdjustment(sig);\n const adjusted = sig.confidence + adjustment;\n\n // How many closed trades have we learned from so far?\n const meta = await sqlite.execute(\n \"SELECT value FROM strategy_meta WHERE key = 'closed_trades'\",\n );\n const closed = (meta.rows[0] as { value: number } | undefined)?.value ?? 0;\n\n if (closed < LEARN_MIN_SAMPLES) {\n // Warm-up: accept everything that already passed technical gates; the\n // learning gate engages once we have enough outcomes.\n return { accept: true, adjusted, reason: \"warm-up\" };\n }\n\n // Dynamic threshold: learned win probability must beat break-even.\n // With payoff ~1:1 to 1:2, we require adjusted confidence >= 55.\n const threshold = 55;\n if (adjusted >= threshold) {\n return { accept: true, adjusted, reason: `learned-edge (+${adjustment.toFixed(1)})` };\n }\n return { accept: false, adjusted, reason: `weak-history (${adjusted.toFixed(0)}<${threshold})` };\n}\n// ---------- Learning report ----------\n// Build a human-readable summary of the paper wallet + what the bot has\n// learned so far: cumulative win rate and per-factor win rates.\nexport async function buildLearningReport(): Promise {\n const meta = await sqlite.execute(\n \"SELECT value FROM strategy_meta WHERE key = 'closed_trades'\",\n );\n const closed = Math.round((meta.rows[0] as { value: number } | undefined)?.value ?? 0);\n\n // Wallet summary.\n const w = await getStrategyWallet();\n const pnl = w.balance - w.initial_balance;\n const pnlPct = (pnl / w.initial_balance) * 100;\n\n // Unrealized (floating) P&L across all currently open positions.\n const floating = await getFloatingPnL();\n\n // Overall stats from closed trades.\n const stats = await sqlite.execute(\n `SELECT\n COUNT(*) AS total,\n SUM(CASE WHEN status IN ('tp1','tp2') THEN 1 ELSE 0 END) AS wins\n FROM strategy_trades WHERE status != 'open'`,\n );\n const s = stats.rows[0] as { total: number | null; wins: number | null };\n const total = s.total ?? 0;\n const wins = s.wins ?? 0;\n const winRate = total > 0 ? (wins / total) * 100 : 0;\n\n const lines = [\n \"🧠 Strategy Learning Report\",\n \"\",\n \"💰 Paper Wallet\",\n `• Balance: ${w.balance.toFixed(2)}`,\n `• PnL: ${pnl >= 0 ? \"+\" : \"\"}${pnl.toFixed(2)} (${pnlPct.toFixed(2)}%)`,\n `• Floating P&L: ${floating.usd >= 0 ? \"+\" : \"\"}${floating.usd.toFixed(2)} (${floating.pct.toFixed(2)}%) across ${floating.positions} open`,\n `• Open positions: ${w.open_positions}`,\n `• Closed trades: ${closed}`,\n `• Win rate: ${winRate.toFixed(0)}% (${wins}W/${total - wins}L)`,\n \"\",\n ];\n\n const factorsRes = await sqlite.execute(\n \"SELECT factor, wins, losses FROM strategy_factors ORDER BY factor\",\n );\n const factors = factorsRes.rows as unknown as {\n factor: string;\n wins: number;\n losses: number;\n }[];\n\n if (factors.length === 0) {\n lines.push(\"No closed trades to learn from yet.\");\n } else {\n lines.push(\"Per-factor win rates:\");\n for (const f of factors) {\n const total = f.wins + f.losses;\n const wr = total > 0 ? ((f.wins / total) * 100).toFixed(0) : \"0\";\n const bar = total > 0 ? \"█\".repeat(Math.min(10, Math.round((f.wins / total) * 10))) : \"\";\n const state = total >= LEARN_MIN_SAMPLES ? \"\" : \" (warm-up)\";\n lines.push(\n `• ${f.factor} — ${wr}% (${f.wins}W/${f.losses}L) ${bar}${state}`,\n );\n }\n }\n if (closed >= LEARN_MIN_SAMPLES) {\n lines.push(\"\");\n lines.push(\"✅ Learned gate active — filtering weak signals.\");\n } else {\n lines.push(\"\");\n lines.push(`⏳ Learned gate arms after ${LEARN_MIN_SAMPLES - closed} more closed trades.`);\n }\n return lines.join(\"\\n\");\n}\n","deleted",false,"label",[14],"4656a5b0-561d-4ff3-812e-b3949dee3507",["D",1788372422505],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Binance triangular arbitrage paper bot","imageUrl","userId","a79beaf8-5d23-4a58-a785-4a70bab11255","pinnedFileId","01a0634d-9e20-77ee-876e-733794a7e2e2","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","01a0634a-0ca9-7584-9a97-35a1aeb44011","tursoName","4656a5b0561d4ff3812eb3949dee3507","tursoHostname","4656a5b0561d4ff3812eb3949dee3507-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789501263826],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_9":10,"_105":100},"org",{"_9":10,"_99":100,"_101":100,"_55":56,"_102":103,"_104":42},"branch",{"_35":89,"_11":90,"_22":28,"_91":45,"_92":-5,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":12,"_98":10},"isOwned",78,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","","avatar","tier","free","canRestrictVals","profileImageUrl","ctx",{"_82":-5,"_139":140,"_141":142,"_143":144,"_145":42,"_146":42,"_147":42,"_148":32,"_149":-5,"_150":-7,"_151":-7,"_152":63,"_153":42,"_154":42,"_155":-5,"_156":42,"_157":42,"_158":42,"_159":-5},"ENV",{"_110":111,"_112":113,"_114":90,"_115":116,"_117":118,"_119":120,"_121":122,"_123":124,"_125":126,"_127":128,"_129":130,"_131":132,"_133":134,"_135":136,"_137":138},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town","orgs",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]