Research: measured findings
Studies run against our own corpus of confirmed sales. Each one states its method, its sample and its limits — and when the evidence cannot carry a conclusion, it says that instead of publishing a number. Findings recompute as the corpus grows.
What do collectors actually pay for a numbered parallel?
Method. For every player with a confirmed sale of an unnumbered card, we take the median delivered price as that player's base, then the median delivered price of their sales in each serial band, and divide. That gives one ratio per player per band, so a heavily traded player counts once rather than dominating. The published figure is the median of those ratios. The 'we assume' column is the multiplier our own fallback valuation uses — this study exists partly to audit it.
Not enough evidence yet
This study needs confirmed sales of numbered and unnumbered cards. We currently hold 0 where 25 is the minimum we will publish on. Rather than print a number a handful of sales cannot support, we are saying so — the study runs again automatically as the corpus grows.
What does a graded copy actually command over raw?
Method. The same per-player normalisation, keyed on the full grade string rather than the number, because a different grader is a different market. This measures the premium a slab commands — not the probability of achieving that grade, which is a separate question the grading calculator answers.
Not enough evidence yet
This study needs confirmed sales of graded and raw copies. We currently hold 0 where 25 is the minimum we will publish on. Rather than print a number a handful of sales cannot support, we are saying so — the study runs again automatically as the corpus grows.
Why these and not price predictions: a study we can measure honestly is worth more than a forecast we cannot. The scoring formulas these sit beside are documented on the methodology page, the evidence rules in how to read sold comps, and the assumption ladder being audited above on the parallels hub.