01 / Matchup selection
Useful comparisons before random ones.
SirDeal maintains valid product pairs within a category. The selection engine gives most opportunities to close or uncertain products, reserves some comparisons for newer products, and keeps a small exploration share for unexpected signals.
- 70%
- Uncertainty: close scores and under-tested pairs
- 20%
- Discovery: newer products against established ones
- 10%
- Exploration: valid pairs outside the obvious path
02 / Ranking model
Preference strength, not raw win rate.
A perfect record from five battles is not automatically stronger than a well-tested product with hundreds of mixed outcomes. SirDeal uses a Bradley–Terry preference model, which estimates how likely each product is to be chosen relative to the others it faced.
The public score is a readable category signal. It is accompanied by battle count and confidence so an early estimate cannot masquerade as a mature ranking.
03 / Price weighting
A vote keeps its price snapshot.
The products and prices shown are stored with the decision. Recent votes carry more weight; older votes decay. When a current price changes substantially, older choices made under a different trade-off are discounted further.
- 0–30 days: full recency weight
- 31–90 days: reduced weight
- 91–180 days: substantially reduced weight
- Older than 180 days: residual historical weight
04 / Confidence
“Early” is a result, not a verdict.
Products begin neutral with no visible preference score. As useful battles accumulate, confidence progresses from early to growing and then high. The label describes evidence depth, not product quality.
05 / Limits
What the ranking does not prove.
SirDeal measures the choices of participating visitors within the products and prices presented. It is not a controlled representative survey, lab test, durability review, or objective specification verdict. Availability, audience mix, and product selection can affect results.
Commercial separation: retailer commission is never an input to matchup selection, preference scores, or ranking order.