About Duelist
Duelist is a community-powered ranking engine. Instead of relying on opaque editorial decisions or weighted metrics, we let people vote head-to-head and let the math do the rest. Pick any topic, vote, and watch the rankings emerge.
What is ELO?
ELO is a rating system invented by physicist Arpad Elo to rank chess players. It works by comparing players head-to-head and updating both ratings after every match — the winner gains points, the loser loses the same. The key insight is that how many points change depends on how surprising the outcome was.
If the top-rated item beats a lower-rated one, that's expected — ratings barely move. But if the lower-rated item wins, that's an upset — ratings shift dramatically, pulling the underdog up and the favorite down. Over thousands of matchups, this self-corrects into a highly accurate ranking.
How the math works
Starting Rating
1500
Every item begins equal. Rankings emerge entirely from votes — no editorial bias.
K-Factor
32
Controls how much each vote shifts ratings. 32 is moderately volatile — rankings move meaningfully but don't swing wildly from a single vote.
Expected Score Formula
The probability that item A beats item B. A 400-point gap means the higher-rated item wins ~91% of the time. The larger the gap, the less surprising the outcome.
After Each Vote
Upsets move ratings more. Expected outcomes move ratings less. The system is zero-sum: every point gained by one item is lost by the other.
Why ELO beats traditional rankings
No hidden weights
Traditional rankings combine dozens of metrics using weights that change year to year. The formula is opaque. Duelist has one input: your vote.
Self-correcting over time
Early votes have high variance. But as comparisons accumulate, rankings converge toward genuine community consensus. The more people vote, the more accurate it becomes.
Topic-aware
A single overall ranking rarely captures reality. Duelist lets you vote — and see rankings — scoped to specific topics, so niche expertise surfaces where it matters.
Resistant to gaming
The only way to move an item's rating is to win individual matchups against real opponents. Ballot stuffing is rate-limited and IP-tracked.
How matchups are chosen
70% of matchups use smart pairing: the item with the fewest comparisons is selected, then matched against a similarly-rated opponent (within ~300 ELO points). This accelerates convergence by prioritising underexposed items and fair fights.
The other 30% are random pairings. This adds exploration — surfacing unexpected matchups that might reveal surprising community preferences, and preventing the algorithm from getting stuck in local patterns.