Run census

Momentum runs: 32 (~32 per session)
Run length distribution: {2: 2, 3: 4, 4: 2, 5: 1, 7: 1, 8: 1, 9: 1, 10: 2, 11: 3, 12: 1, 13: 1, 14: 1, 16: 1, 17: 6, 18: 1, 20: 4}
Runs lasting <= 2 candles: 2 (6.2%)
Direction at start: {'DOWN': 16, 'UP': 16}
Boundary structure at start: {'none': 32}

Self-consistency

Self-consistency (labeled direction vs the run's own realized drift):
  mismatched runs: 15/32 (46.9%) | by run length: {'>2': 15}

Predictiveness

Predictiveness of run-start direction (signed forward return):
  +5 candles: hit rate 46.9% over 32 runs, average signed return -0.00025
  +10 candles: hit rate 35.5% over 31 runs, average signed return -0.00022
  by starting state (+10 candles):
    STARTING_DOWN      n=  10 hit= 40.0% avg=+0.00024
    STARTING_UP        n=  10 hit= 40.0% avg=+0.00003
    DECELERATING_DOWN  n=   4 hit= 50.0% avg=+0.00120
    DECELERATING_UP    n=   4 hit= 25.0% avg=-0.00138
    ACCELERATING_UP    n=   2 hit=  0.0% avg=-0.00140
    ACCELERATING_DOWN  n=   1 hit=  0.0% avg=-0.00599
  by boundary trigger (+10 candles):
    non-structural     n=  31 hit= 35.5% avg=-0.00022

Boundary churn

Boundary churn:
  direction reversals between consecutive runs: 16
  flip-flops (reversal after a run of <= 2 candles): 2
  flip-flops by the reversing run's boundary trigger: {'non-structural': 2}

Change rate

Absolute change_rate at run start (vs MOMENTUM_MIN_CHANGE_RATE=0.0001):
  DOWN           n=  16 median=0.000409 below threshold: 12.5%
  UP             n=  16 median=0.000534 below threshold: 6.2%

Anchor angle

No anchor-angle observations to measure.

Velocity quality

Kinematic velocity quality (recency-weighted velocity / noise, #642), signed +10-candle forward return:
  |q| 0.0-0.1    n=  106 hit= 43.4% avg=-0.00014
  |q| 0.1-0.2    n=   86 hit= 45.3% avg=-0.00034
  |q| 0.2-0.3    n=   96 hit= 52.1% avg=-0.00008
  |q| 0.3-0.4    n=   52 hit= 48.1% avg=-0.00032
  |q| >= 0.4     n=   39 hit= 64.1% avg=+0.00006

Occurrence census

Trend occurrences: 186 (~186.0 per session)
Occurrence length distribution (candles): {1: 170, 2: 11, 13: 1, 26: 1, 27: 1, 50: 1, 60: 1}
Occurrences lasting <= 5 candles: 181 (97.3%)
Direction at start: {'DOWN': 117, 'UP': 69}
Starting state: {'BEARISH_EMERGING': 117, 'BULLISH_EMERGING': 69}

Self-consistency

Self-consistency (labeled direction vs the occurrence's own realized drift):
  mismatched occurrences: 13/186 (7.0%)

Predictiveness

Predictiveness of occurrence-start direction (signed forward return):
  +10 candles: hit rate 41.4% over 186 occurrences, average signed return -0.00035
  +30 candles: hit rate 38.5% over 169 occurrences, average signed return -0.00036
  by starting state (+30 candles):
    BEARISH_EMERGING     n= 113 hit= 48.7% avg=+0.00008
    BULLISH_EMERGING     n=  56 hit= 17.9% avg=-0.00125

Occurrence churn

Occurrence churn:
  direction reversals between consecutive occurrences: 15
  flip-flops (reversal after an occurrence of <= 5 candles): 11

Momentum-trend alignment

Momentum-trend alignment (momentum run starts, signed +5-candle forward return):
  with-trend     n=  18 hit= 50.0% avg=-0.00045
  counter-trend  n=  13 hit= 46.2% avg=+0.00006
  no-trend-yet   n=   1 hit=  0.0% avg=-0.00059

Trades: 15 | wins 4 (26.7%) | net -405.18
Positions (round trips): 15 | wins 4 (26.7%)

Session and cohort net PnL

  spy_2026-06-10     trades=  15 wins=  4 net=   -405.18  [training]
  training   ( 1 sessions): trades=  15 wins=  4 net=   -405.18
  all        ( 1 sessions): trades=  15 wins=  4 net=   -405.18

Net PnL by session regime

  RANGE          ( 1 sessions): trades=  15 wins=  4 net=   -405.18

By exit path

  STOP_MARKET    n=  11 wins=  0.0% pnl=   -871.15
  TAKE_PROFIT    n=   4 wins=100.0% pnl=   +465.98

By direction

  DOWN           n=   8 wins= 25.0% pnl=   -244.65
  UP             n=   7 wins= 28.6% pnl=   -160.52

Hold duration in candles (all closes): {0: 2, 1: 4, 2: 3, 5: 1, 6: 1, 9: 2, 20: 1, 29: 1}
Hold duration in candles (stop-outs):  {0: 2, 1: 3, 2: 2, 5: 1, 6: 1, 9: 2}
Longest hold: 29 wall-clock minutes | cross-session holds: 0 (must stay 0 with the session flush enabled)

Round-trip friction / 1R: median=0.05 p90=0.13 (#612 target: <= 0.20)
Stop distance as fraction of entry price: p10=0.00026 median=0.00074 p90=0.00129
Entry risk/reward ratio: min=1.55 median=2.00 max=6.94
Maximum favorable excursion before stop-out (in R): median=0.00 p75=1.33 | >=0.5R 45.5%, >=1R 45.5% of stopped trades

MAE/MFE in R (closed-candle excursions vs initial risk)

  winners n=   4 | MAE median=1.08R p75=1.11R p90=1.11R | MFE median=2.00R p75=2.03R
  losers  n=  11 | MAE median=1.20R p75=1.22R p90=1.22R | MFE median=0.00R p75=1.61R

Win rate by entry-confidence decile

  0.4-0.5: n=   6 wins= 50.0%
  0.5-0.6: n=   7 wins=  0.0%
  0.6-0.7: n=   2 wins= 50.0%

Entry expectancy by 30-minute session bucket (minutes since open)

     0-30   n=  2 wins=   0% net=  -182.86 meanR=-1.05
    60-90   n=  1 wins=   0% net=   -45.51 meanR=-1.08
    90-120  n=  1 wins=   0% net=   -44.23 meanR=-1.09
   120-150  n=  2 wins=  50% net=   -70.09 meanR=+0.18
   150-180  n=  1 wins= 100% net=  +126.36 meanR=+1.64
   180-210  n=  2 wins=   0% net=  -122.06 meanR=-1.09
   210-240  n=  2 wins=  50% net=   -38.11 meanR=+0.47
   240-270  n=  1 wins=   0% net=   -69.02 meanR=-1.05
   270-300  n=  1 wins=   0% net=   -52.76 meanR=-1.07
   300-330  n=  1 wins= 100% net=  +170.19 meanR=+1.63
   330-360  n=  1 wins=   0% net=   -77.08 meanR=-1.05

Churn checks (both should stay at 0 after the 2026-07-08 fixes)

  REDUCE fills on in-profit positions: 0 (of 0 REDUCE fills)
  Same-direction re-entries within 3 candles of a stop-out: 0 (of 11 stop-outs)

Bracket vs realized outcome

Filled candidates: 15
Win-direction agreement rate: 93.3%
Realized-R mean absolute error: 0.295

Regime

Regime: RANGE
Signed efficiency: -0.0573
Mean absolute return: 0.0005
Flip density: 0.5195