Archived research experiment. This page describes an earlier version of the model. For the current methodology, see How the estimates work.
Research results · September 18, 2026
What party size changes in the draw
Whole-party allocation is now tested with a simulation engine. A party receives tags only when every member fits. The experiments below show why the size of the penalty depends on the competing applications.
Completed: 1,944,020 simulated draws. That includes 1,000,020 benchmark draws, 800,000 small-population sensitivity draws, and 144,000 synthetic elk draws using the site's 2026 report counts.
Your 10-tag example
Assume 100 competing solo applications, plus your application, all seeking the same hunt. Each application receives one random position. Your group of four must be reached with at least four tags left: one of the first seven positions out of 101.
| Your party | Exact chance | Simulation |
|---|---|---|
| 1 | 9.90% | 9.93% |
| 2 | 8.91% | 8.93% |
| 3 | 7.92% | 7.98% |
| 4 | 6.93% | 6.97% |
These are exact probabilities under the stated example. The engine also passed tests for zero tags, excess tags, a party too large to fit, competing groups, and progression through all three choices.
The other applicants matter
Here the number of competing people stays at 100 and the hunt has 10 tags. Regrouping those people changes the number of applications and the tags left near the end. Your party is an additional application.
| Competing population | Applications | Solo | Party of 2 | Party of 3 | Party of 4 |
|---|---|---|---|---|---|
| Solo competitors | 100 | 9.90% | 8.93% | 7.93% | 6.95% |
| Mostly solo | 88 | 10.39% | 9.24% | 8.20% | 7.19% |
| Mixed parties | 64 | 11.36% | 9.57% | 8.41% | 7.36% |
| More groups | 42 | 15.60% | 10.68% | 8.91% | 7.69% |
A larger target party had lower odds than a solo applicant within every tested population. But the percentage decrease is not a universal multiplier. Solo applicants can also benefit from tags left behind by larger parties.
Your elk choices
1st: 16D First Archery → 2nd: 16A First Archery → 3rd: 16C First Archery.
We created 48 synthetic applicant populations across 322 elk hunt codes, preserving the published number of applicant people at every choice position. Each population went through 3,000 randomized draws. At the same draw position, we compared your application with one, two, three, or four hunters.
The table shows how much the chance of receiving any of your three hunts decreased when changing from solo to a party of four, across the tested assumptions.
| Pool | Party-of-4 decrease versus solo |
|---|---|
| Resident | 0.80–1.90 percentage points |
| Nonresident | 1.22–5.48 percentage points |
| Outfitter | 4.35–13.98 percentage points |
For this particular application, the resident penalty was relatively small because the modeled resident allocations were much larger. This result does not apply to every hunt.
See a complete choice-by-choice example
One assumed population: resident pool, mixed competing party sizes, and stronger association between the weapon on earlier and later choices. These numbers describe that synthetic scenario; they are not measured odds for actual resident parties.
| Party | 1st: 16D | 2nd: 16A | 3rd: 16C | Any of the three |
|---|---|---|---|---|
| 1 | 24.10% | 12.68% | 14.55% | 51.33% |
| 2 | 23.67% | 12.85% | 14.28% | 50.80% |
| 3 | 23.30% | 13.08% | 13.93% | 50.32% |
| 4 | 23.15% | 13.08% | 13.82% | 50.05% |
Each choice percentage here is the chance of winning at that position on this complete application. They are mutually exclusive simulation outcomes, so they sum to “Any.” This differs from the site's historical rate among people who listed an individual hunt at a given position.
What the historical check revealed
Matching applicant counts does not automatically reproduce who won. For example, one solo-competitor scenario allocated about 17.5 resident awards to 16D second-choice applicants, compared with 8 in the report. That difference shows that assumed links between choices matter.
The allocation mechanism passed its benchmarks. The applicant-population model still needs improvement. Running more repetitions would reduce random simulation noise; it would not identify the missing party sizes or linked choices.
Compare simulated and published awards
Resident pool, all competitors solo, independently assigned later choices, first synthetic cohort. Simulated values are mean awarded people over 3,000 draws, compared with the same year's published counts. This is a reconstruction check, not a test against a future year.
| Hunt | Choice | Published awards | Simulated awards |
|---|---|---|---|
| ELK-2-291 | 1 | 61 | 53.4 |
| ELK-2-291 | 2 | 8 | 17.5 |
| ELK-2-291 | 3 | 7 | 5.1 |
| ELK-2-272 | 1 | 121 | 117.8 |
| ELK-2-272 | 2 | 68 | 68.8 |
| ELK-2-272 | 3 | 21 | 23.4 |
| ELK-2-284 | 1 | 50 | 45.5 |
| ELK-2-284 | 2 | 35 | 40.7 |
| ELK-2-284 | 3 | 36 | 34.8 |
The live application outlook now uses a separate eight-scenario dataset for six species. Historical per-choice rates and simplified party scenarios remain separately labeled. The live range is experimental; its full method is explained on the main Methodology page.
All scenario results
These are the chances of receiving any of the three example elk hunts under each synthetic population assumption. Comparing rows shows sensitivity to the missing information; the range across rows is not a confidence interval.
Open all 24 scenarios
Resident pool
| Competing parties | Choice linkage | Solo | Party of 2 | Party of 3 | Party of 4 |
|---|---|---|---|---|---|
| solo | independent | 45.12% | 44.78% | 44.52% | 44.22% |
| solo | same weapon | 50.57% | 50.23% | 50.05% | 49.77% |
| mostly solo | independent | 46.15% | 45.67% | 45.42% | 45.12% |
| mostly solo | same weapon | 49.60% | 49.30% | 48.97% | 48.65% |
| mixed | independent | 47.45% | 46.78% | 46.38% | 46.18% |
| mixed | same weapon | 51.33% | 50.80% | 50.32% | 50.05% |
| group heavy | independent | 47.42% | 46.23% | 45.87% | 45.52% |
| group heavy | same weapon | 50.48% | 49.48% | 49.15% | 48.78% |
Nonresident pool
| Competing parties | Choice linkage | Solo | Party of 2 | Party of 3 | Party of 4 |
|---|---|---|---|---|---|
| solo | independent | 4.87% | 4.50% | 3.98% | 3.52% |
| solo | same weapon | 5.18% | 4.77% | 4.32% | 3.92% |
| mostly solo | independent | 5.57% | 5.03% | 4.60% | 4.10% |
| mostly solo | same weapon | 5.37% | 4.88% | 4.35% | 4.15% |
| mixed | independent | 5.78% | 4.73% | 4.08% | 3.70% |
| mixed | same weapon | 5.90% | 5.10% | 4.52% | 4.08% |
| group heavy | independent | 7.13% | 5.13% | 4.38% | 3.93% |
| group heavy | same weapon | 9.83% | 5.57% | 4.87% | 4.35% |
Outfitter pool
| Competing parties | Choice linkage | Solo | Party of 2 | Party of 3 | Party of 4 |
|---|---|---|---|---|---|
| solo | independent | 26.58% | 25.17% | 23.68% | 22.23% |
| solo | same weapon | 27.58% | 26.03% | 24.43% | 23.08% |
| mostly solo | independent | 28.90% | 26.82% | 25.37% | 23.57% |
| mostly solo | same weapon | 28.62% | 26.57% | 25.00% | 23.53% |
| mixed | independent | 31.03% | 27.88% | 26.12% | 24.57% |
| mixed | same weapon | 30.97% | 28.03% | 26.03% | 24.48% |
| group heavy | independent | 33.25% | 27.58% | 24.80% | 23.23% |
| group heavy | same weapon | 36.80% | 27.58% | 25.07% | 22.82% |
How to interpret this archived experiment
The experiments below predate the audit fixes. The current live populations prohibit repeated hunts, enforce hunt-specific party limits, and publish any small marginal residuals. Current ranges also include a reconstruction sensitivity allowance.
- Competing party mixes were assumed, not observed. Nominal application-size weights for sizes 1/2/3/4 were: solo 100/0/0/0; mostly solo 85/10/4/1; mixed 55/25/15/5; and group-heavy 25/35/25/15 percent. Realized counts are saved in the results.
- Later choices were assigned either independently from remaining counts or with an eightfold preference for the first choice's weapon. Both preserve aggregate applicant counts. Neither reconstructs actual individual applications.
- Resident, nonresident, and outfitter pools used fixed report allocations and were simulated separately. Cross-pool transfers, mixed-residency parties, individual eligibility, and special quota adjustments were not reconstructed. Duplicate synthetic hunt choices were allowed.
- The target application is additional to each fixed competitor population. Changing its size does not change the competing population. This is a controlled comparison, not a reconstruction of a known person's original draw.
- Each scenario has 6,000 random draw orders over two synthetic populations. Saved 95% Monte Carlo intervals measure random simulation uncertainty conditional on those populations, not uncertainty about real-world odds.
- Before adopting a forecast, the next step is to fit and test applicant-population assumptions against multiple historical years, with some years reserved for evaluation. These experiments alone do not establish predictive accuracy.
Download detailed results (JSON) · Download 10-tag experiments (JSON)
Source: the site's 2026 elk report snapshot. Input SHA-256: 87138316e2d6bd34794d6d2799ae1d36484838f601c2b905ec0eef766b8fde74. Seed sequence begins at 260918. The C++ engine, Python population generator, benchmarks, and report builder are saved with the site source.