1. Start with the official boundary
The publisher-provided Steam listing describes the core risk: raid loot matters only if the player gets it out. That establishes useful qualitative context for separating a find from an extraction. It does not publish the numeric target rates, route probabilities, market values, or timings required by a calculator.
2. Define one observable success event
Write down what success means before collecting data. “Target appeared at the location,” “target was picked up,” and “target was successfully extracted” are not interchangeable. A reusable definition should specify:
- the exact item, category, enemy, room, container, or other event being counted;
- the route and locations checked in every eligible raid;
- mode, party rule, date, game state, and any temporary event conditions;
- what makes a raid eligible or excluded;
- whether duplicate targets in one raid count once or separately.
3. Grade the probability input
- Publisher or official table: preserve its exact scope, date, and wording.
- Extracted game data: identify the project, version, file, extraction method, and record-to-target mapping.
- Observed sample: retain successes, eligible trials, exclusions, dates, and uncertainty.
- What-if assumption: label it plainly and use it to explore a scenario, not to describe the live game.
A number copied without its denominator and selection rule cannot support a rate claim. A small personal sample can still be useful for planning, but it should remain a personal estimate and should not be presented with more precision than the observations justify.
4. Separate finding from extracting
Let f be the chance that an eligible raid finds the target and e be the chance of extracting after that target has been found. The combined scenario chance is p = f × e. An overall survival rate is not automatically the same as e: finding the item may change route, time, risk, or player behavior. Estimate the conditional event that the formula actually uses.
5. Time a complete repeatable raid cycle
Start before queueing and stop when the next comparable attempt can begin. Include matching, loading, traversal, checks, fights or avoidance, extraction or death, healing, stash management, selling or storing, and re-queueing. Record several full cycles rather than timing only the best run. Use a median for a typical route comparison and keep the raw values when variability matters.
6. Keep tracked-item contribution narrower than raid profit
For a target value v, its gross expected contribution per raid is p × v. Subtracting a direct per-cycle cost produces a target-only net contribution, not total raid EV. A full economic model would also need every other loot outcome, carried and secured items, sales costs, equipment-loss treatment, insurance outcomes, recovery, tasks, price timing, and correlations. Do not label a one-target calculation as a universal profit route.
7. Compare scenarios with identical boundaries
Change one assumption at a time. Use the same target definition, eligibility rule, extraction condition, time boundary, price timestamp, and direct-cost rule. Save one calculator URL per route and keep the evidence note beside it. A route wins only the stated scenario; a different target, build, risk preference, or patch can reverse the comparison.
8. Keep a real observation log
date,mode,route,target,eligible_raids,target_finds,extracted_targets,full_cycle_minutes,value_snapshot,direct_cost_rule,source_url,notes YYYY-MM-DD,mode label,route label,target definition,0,0,0,0,0,describe the cost,URL or personal sample,record actual observations
This is an empty schema, not claimed player data. Do not seed it with invented rows. If the route definition changes, start a new series instead of silently merging incompatible observations.
Use the scenario calculator
Enter your supported assumptions and measured cycle in the EFT Raid Target & Extraction Scenario Calculator, then save the URL with the observation note. The tool exposes cumulative thresholds and expected values without adding a hidden map or item preset.