Methodology

What the numbers are

Every percentage is a prior-draw probability: given the published quota, applicants, and point distribution for one draw year, the chance an applicant at a given point level would have drawn a first-choice tag. We never forecast next year. Trends (point creep) are only shown with at least three draw years of history.

Data

• Sources are official agency draw reports, fee schedules, and regulations only. Each row stores the source URL, retrieval date, and page. Secondary sites are never a source of a number.
• Loads fail — they are not partially applied — when tags issued exceed quota, first-choice applicants are fewer than first-choice tags, per-level pool sums disagree with reported totals, or any row lacks a source.
• A blank cell means the agency did not publish that figure. It is never a zero.

Strategies by draw system

• Preference points (Colorado, Utah general deer, Iowa, Kansas, Nebraska, South Dakota): 100% above the level that drew out, 0% below. At the cut level, the share is tags issued ÷ applicants at that level when the state publishes per-level pools; otherwise we show the cut as unknown rather than guess. Agency per-level tables are not monotone on their own — group applications draw at their lowest member's points, applicants withdraw or are ineligible, and Colorado's hybrid pool hands a few licenses to applicants below the cut — so a table can show 3 of 4 drawing at 4 points and 14 of 17 at 3. We take the monotone envelope: a level is credited with the best share of any level at or below it, every level strictly above the published drawn-out level is a sure thing, and the realised share at the cut itself is kept.
• Hybrid preference + random (Wyoming nonresident, California, Oregon): the preference share of the quota is allocated by points with the same monotone envelope as above; the random share is drawn uniformly from everyone left, so the random-pass rate is random tags ÷ unsuccessful applicants. Closed form, no simulation. Published bands cover every whole point level inside them (Montana's ".2–1.8" row is the 1-point row; Wyoming's residual "<13" row is everyone from 0 to 12).
• Bonus points (Arizona, Nevada, Utah, Montana, North Dakota): each applicant gets (points + 1), (points + 1)² or points² + 1 entries per the state's rule; an optional top-points pass awards a share first. Computed by Monte Carlo over published pools (weighted sampling without replacement, 4,000 iterations, 95% Wilson interval shown) or, for very large pools, the equivalent expected-draw recursion. Applicants an agency prints with a blank point level ("none" in Montana's 2006–2023 tables) are zero-point applicants and count in full. Without per-level pools we show the unweighted ratio and label it as such.
• Random (Idaho, New Mexico): tags ÷ applicants within the residency bucket — where the agency prints first-choice draws we use that exact ratio, and the same for 2nd/3rd-choice draws.
• Nevada: one random-number order across all five choices (a 2nd-choice application with a lower number beats a 1st-choice one), so every application listing the hunt competes in one weighted pool with (points² + 1) entries.
• Montana: permits use (points² + 1) entries with residents and nonresidents drawn together under a 10% cap; the nonresident combination licenses are a preference draw whose 25% random share goes only to zero-point applicants.
• Common pools with a cap (Arizona, Colorado, Montana, North Dakota, Oregon): residents and nonresidents are drawn from one pool and a residency is capped at its share. In bonus systems the cap is applied inside the simulated draw the way the agencies run it: once a residency has its allowance it is skipped and the tag goes to the next applicant, in both the bonus pass and the random pass. Arizona additionally limits nonresidents to half of their cap in the bonus pass (R12-4-114(G)), which is why a 28-point nonresident can miss the pass while residents draw down to 25. Nonresident "quota" on those pages is the cap, "undersubscribed" means every first-choice applicant drew, and a residency row with no quota and no applicants is a hunt closed to that residency, shown without odds.
• Washington publishes no per-point-level tables since 2020, so its odds are the unweighted permits ÷ applicants ratio and are labelled as such — a 10-point applicant's true chance is far higher than shown, a 0-point applicant's far lower.
• Unpublished point levels (nobody applied with exactly that many points): in an undersubscribed hunt they are 100%. Otherwise, in bonus systems the bonus-pass share follows the published pass cut and the random-pass share is scaled by entry weight from the nearest level that still had applicants in the random pass, under the same residency cap; in preference systems a level between two published levels takes the envelope, and above the drawn-out level it is a sure thing.
• Lower choices: one number per choice rank, from the choice-ordered random pass or the agency's issued-to-other-choices count. A choice nobody listed has no odds (shown as a dash).
• Known limitations: Colorado hunt codes with a 20% hybrid pool are modelled as pure preference draws, so the small random chance at 5–19 points shows only where a hybrid winner appears in CPW's table (7 nonresident and 301 resident code-years). South Dakota's strict 2+/1+/0+ tiers with cubed entries inside each tier are shown as the realised per-level draw rate, not a weighted-lottery probability, so small pools look jagged. Nevada, Washington and Pennsylvania draw every choice on points, but the lower-choice number shown is an average over all lower-choice applicants rather than one per point level.

Guards

• Hunts with fewer than five tags show tags/applicants counts instead of a percentage.
• All money figures show the fee year used; when a draw year has no fee schedule we use the latest earlier year.
• Back-testing: see below. States whose latest back-test error exceeds 15 points show a warning banner.

Back-test: how well did last year's odds anticipate the next draw?

The forecast a hunter actually makes with these pages is "last year's odds at my point level." For every hunt code and point level published in two consecutive draws we score that forecast against what the next draw's odds turned out to be. MAE is the mean absolute error in probability (weighted by applicants at each level where the state publishes pools); within 10 pts is the share of code/level pairs off by no more than 10 percentage points; cut shift is how far the drawn-out level moved on average (positive = creep).

Expect the weighted number to be much larger than the plain one. Odds are near 0% or 100% at most point levels and only change at the cut; applicants concentrate at exactly that cut (that is why it is the cut), so the error they experience is the boundary error. Read the weighted MAE as "if you are sitting at last year's cut, this is roughly how wrong last year's number will be for you."

No back-test yet — needs two consecutive draw years with computed odds.

Coverage