summary: the lookback period controls how far back WIP reaches for historical data. you can choose 3 months, 6 months, 1 year, or 2 years — shorter windows react faster to recent conditions, longer windows give you a bigger sample and smoother averages.
the lookback period controls how far back WIP goes when calculating the historical data behind every setup on the dashboard.
edgeful offers 4 options: 3 months, 6 months, 1 year, and 2 years. the one you choose changes the numbers you see — and the story the data is telling you. think of it as a spectrum: the shorter the window, the more it weights what the market has been doing lately; the longer the window, the more data it pulls in and the smoother the averages get.
3 months — most recent, most reactive
the 3-month lookback focuses on what the market has been doing lately. markets go through regimes — periods where volatility contracts, expands, or behaves differently than the longer-term average.
if the last 3 months have been unusually choppy or trending hard in one direction, that shift may not show up clearly in a longer window. the 3-month lookback surfaces it. use it when you want to weight recent behavior more heavily — especially during periods of elevated volatility or after a significant change in market conditions.
6 months — the balanced default
the 6-month lookback pulls in more historical occurrences for each setup. more occurrences means a larger sample size — and larger sample sizes produce more reliable averages.
if a setup has filled 68% of the time over 6 months, that number is built on more data points than the same setup at 3 months — it's less likely to be a fluke or a short-term blip. 6 months is the best default for most traders: the patterns are more statistically grounded, and the averages smooth out short-term noise without reaching so far back that stale market regimes dominate the number.
1 year and 2 years — the largest samples
the 1-year and 2-year lookbacks reach back the furthest and give you the biggest sample of all. they're the most useful for two things:
setups that don't trigger often. if a setup only fires a handful of times per quarter, a 3- or 6-month window can leave you with a thin sample. stretching to 1 or 2 years builds the occurrence count up to something you can actually trust.
seeing the long-run baseline. a 1- or 2-year number tells you how a setup behaves across multiple market regimes, not just the current one. it's the steadiest read on the "true" long-term tendency.
the trade-off is reactivity: the further back you reach, the more a long-gone market regime can weigh on the average and mask what's happening right now. that's the exact thing the 3-month window is there to catch — so use the long lookbacks for the baseline, and check a shorter one to see if recent behavior has drifted from it.
a note on sample size
some setups don't trigger that often. if a setup only occurred 8 times over 3 months, a 75% fill rate is based on 6 occurrences — that's a thin sample.
the same setup over 6 months might have 20+ occurrences, and over 1 or 2 years more still — the same 75% now means a lot more. this is exactly when the longer lookbacks earn their keep.
always pay attention to the occurrence count alongside the percentage. a stat built on a small sample can look compelling and still be unreliable. reaching for a longer lookback is the fastest way to fix a thin sample — more data, more confidence. or let WIP do the check for you: set the frequency filter to a minimum sample and any card built on fewer past sessions drops off the dashboard (see data filters on what's in play).
which one to use
a practical approach: run more than one and compare.
if the numbers are consistent across a short and a long window, the pattern is holding up across time — that's a stronger edge. if a short window has diverged significantly from a longer one, the market may have shifted recently and it's worth knowing why.
a simple way to think about it:
3 months — what's working right now
6 months — the balanced default for most traders
1 year / 2 years — the long-run baseline, and the fix for thin samples on setups that rarely trigger
when in doubt, default to 6 months. the sample size gives you a stable foundation, and you can always drop to 3 months to check if recent behavior has changed, or stretch to 1–2 years to firm up a thin sample.