16 lines
509 B
Markdown
16 lines
509 B
Markdown
|
# Experiments
|
||
|
|
|
||
|
|
Define every pre-registered experiment here (E1..E8 from VALIDATION_PROTOCOL.md).
|
||
|
|
|
||
|
|
Initial (bootstrap) experiments — all are `NOT TESTED` pending data+harness runs:
|
||
|
|
|
||
|
|
- E1 ARIMA > naive
|
||
|
|
- E2 SAX > naive
|
||
|
|
- E3 SAX adds info beyond ARIMA
|
||
|
|
- E4 ARIMA adds info beyond SAX
|
||
|
|
- E5 agreement improves conditional performance
|
||
|
|
- E6 disagreement identifies poor/unstable regimes
|
||
|
|
- E7 survives transaction costs
|
||
|
|
- E8 stable across walk-forward periods
|
||
|
|
|
||
|
|
Do not add model complexity until E1..E8 are answered.
|