Spotting jumps in tick data before they spoil a statistic

[Amanhasnoname]

Price changes in forex are mostly small, but occasionally a jump occurs: a sudden move of many standard deviations in a single update or within a few seconds. Jumps arise from news, large orders, and liquidity gaps.

In statistical work, jumps distort volatility estimates, correlations, and risk measures. A single event can dominate a day's variance.

Separating jumps from the continuous component helps. Simple methods flag any return larger than a multiple of local volatility, for example five times the rolling standard deviation. More elaborate approaches compare realized variance with bipower variation.

For a scalper, a jump has two implications. Stops and targets within the jump are skipped over, and signals based on volatility spike. Knowing how often jumps occur in the traded window quantifies gap risk.

Counting jumps by hour reveals when they cluster. The pattern often aligns with scheduled data, but a share comes from unscheduled events.

Strategies can then treat the periods after a jump differently, from avoiding entries to switching style.

How many jump events per week show up in the pair being traded, and when do they cluster?