How to test it properly:
- Define the spike exactly: 10 pips or more from the 08:00 open, within three M1 bars.
- Log every qualifying spike, not only those that look fadeable.
- Use real fills, or at least add realistic slippage. Stops right after the open often fill worse than planned.
- Mark news days, including minor European releases at 08:00 or 08:15 London.
Two filters are logical candidates. Spikes driven by news should continue more often, so removing news days may help. Spikes with wide spreads will have poor fills. But each filter shrinks the sample, and with 30 or 40 trades left you can't tell a real improvement from luck.
Has anyone run a similar fade test on the open spike? Did you find a filter that separates the ones that continue?