1. 焦点更新条件变更

2. 图像更新，使用30秒的trade数据与3分钟的trade数据长度生成犯错概率

strategy/src/predictor.rs

@@ -67,6 +67,7 @@ pub struct Predictor {
 
     pub prev_insert_time: Decimal,
     pub prev_save_time: Decimal,
+    pub init_time: Decimal,
 
     pub params: Params,
 
@@ -79,8 +80,8 @@ pub struct Predictor {
 impl Predictor {
     // 时间窗口大小（微秒）
     // const MAX_TIME_RANGE_MICROS: i64 = 3 * 60_000_000;
-    const TIME_DIFF_RANGE_MICROS: i64 = 10 * 60_000_000;
-    const TRADE_LONG_RANGE_MICROS: i64 = 3 * 60_000_000;
+    const TIME_DIFF_RANGE_MICROS: i64 = 15 * 60_000_000;
+    const TRADE_LONG_RANGE_MICROS: i64 = 60_000_000;
     // const SPREAD_RANGE_MICROS: i64 = 15 * 60_000_000;
     const TRADE_SHORT_RANGE_MICROS: i64 = 30_000_000;
     // const ONE_MILLION: Decimal = dec!(1_000_000);
@@ -159,7 +160,7 @@ impl Predictor {
             bid_delta: Default::default(),
 
             spread_sma_1000_time_vec: FixedTimeRangeDeque::new(10_000),
-            fair_price_time_vec: FixedTimeRangeDeque::new(10_000),
+            fair_price_time_vec: FixedTimeRangeDeque::new(10_000_000),
             mid_price_time_vec: FixedTimeRangeDeque::new(100_000),
             fair_price: Default::default(),
             fair_rate_focus: Default::default(),
@@ -182,6 +183,7 @@ impl Predictor {
 
             prev_insert_time: Default::default(),
             prev_save_time: Decimal::from(Utc::now().timestamp_millis()),
+            init_time: Decimal::from(Utc::now().timestamp_millis()),
 
             params,
 
@@ -342,7 +344,7 @@ impl Predictor {
             }
 
             // 更新程序关注的变化幅度焦点
-            if self.fair_rate_focus.is_zero() {
+            if self.fair_rate_focus.is_zero() && self.inventory.is_zero() {
                 // 只有有强度的rate才有资格被称为针
                 if rate.abs() > self.params.open_activate {
                     // 向上涨，并且fair下穿mid，视为观测阶段开始
@@ -569,11 +571,11 @@ impl Predictor {
 
         let inventory = self.inventory;
 
-        let last_spread_sma = self.spread_sma_1000_time_vec.deque.iter().last().unwrap();
-        let first_spread_sma = self.spread_sma_1000_time_vec.deque[0];
+        let last_fair_price = self.fair_price_time_vec.deque.iter().last().unwrap();
+        let first_fair_price = self.fair_price_time_vec.deque[0];
 
-        let sigma_square = last_spread_sma - first_spread_sma;
-        let gamma = now - self.last_update_time;
+        let sigma_square = (last_fair_price - first_fair_price) / first_fair_price;
+        let gamma = self.error_rate;
 
         let kappa = self.fair_rate_focus;