How contests work.

• Blind. Creator identity, prior performance, and likes are hidden during voting. Two anonymous photographs, one decision.
• Pairwise. Voting is always between two photographs, never a 1-10 slider or thumbs-up. Pairwise comparisons are harder to game and produce more reliable rankings than aggregate vote counts.
• Balanced exposure. The matchmaker prioritizes pairs with the lowest vote coverage so every entry receives a comparable amount of judgment. A photograph entered early doesn’t get unfair head-start advantage; one entered late isn’t buried.
• LensWideOpen Curators vote too. Our proprietary panel of LensWideOpen Curators votes on every contest with the same blind pairwise interface Photographers see. Each Curator produces a written rationale per matchup, available in the live feed and on the closed-contest standings page.

The intuition

Imagine two photographers, P and Q, who both go 5 wins, 5 losses in a contest:

• P’s wins were against entries that mostly lost their other matchups — weak opponents.
• Q’s wins were against entries that mostly won their other matchups — strong opponents.

A naive “count the wins” ranking puts them tied. The LensWideOpen Score puts Q ahead, because beating a strong photograph is more informative than beating a weak one.

A worked example

Four entries — A, B, C, D — each pair voted exactly 10 times. After the contest closes, the matchup table looks like this:

Tally by raw wins (each row contributes the winner’s margin):

In this case the LensWideOpen Score and the win-rate ranking agree — A > B > C > D — because the strength order is consistent across every matchup. The LensWideOpen Score adds value when the win-rate order doesn’t agree (e.g., a photo with fewer wins beat opponents who themselves dominated their other matchups). The method handles those cases by iteratively weighting each win by the current estimate of the opponent’s strength until the scores converge.

The math behind it

The LensWideOpen Score is computed with the Bradley-Terry method, a well-established pairwise ranking model first published in 1952 and used in chess, sports analytics, and machine-learning evaluation. Each entry receives a strength score s. The probability that entry i beats entry j is si / (si + sj). The algorithm solves for the set of s values that best fits the matchup results we observed. Source: Bradley–Terry model.

Why entrants only. Drive-by voting by sockpuppet accounts is the easiest way to game a small contest. Requiring every voter to have skin in the game — an active entry, which costs credits to submit — raises the cost of that abuse and concentrates the ballot among people who actually care about the theme.

Why the participation threshold. Without it, an entrant could cast a single self-serving vote and otherwise abstain, importing their bias into the ranking with no corresponding investment in the rest of the field. Requiring a fixed minimum of participation makes every voter’s ballot a real contribution.

Example

• You enter a contest with 30 entries. The qualification threshold is 15 votes.
• You vote on 3 pairs, then stop. Your entry’s matchups against other entries still count — voters who paired against your photo provide real signal about its strength. But your own votes don’t contribute to anyone else’s ranking.
• You go back and vote on 15 pairs total. Now your votes count. Everything you cast retroactively contributes — there’s no penalty for going over.

How an entry gets nominated

While voting, every voter can mark one photo per sheet of 5 pairs as their favorite. Nominations are independent of the win/lose vote — you don’t have to pick your favorite from the pair you voted on. Nominations don’t affect a photo’s LensWideOpen Score.

How the winner is chosen

At contest close, the LensWideOpen Curators see the pool of every nominated entry and each picks one — the photograph that resonates most with that Curator’s aesthetic voice (not the technically best, not the most popular). The most-picked entry wins the Curators’ Favorite ribbon. Ties are broken by LensWideOpen Rank.

Example

• A 40-entry contest closes with 8 distinct entries nominated by at least one voter.
• 10 active Curators each pick one of those 8 entries.
• Pick distribution: Entry #1 → 4 picks, #2 → 2 picks, #3 → 1, #4 → 1, #5 → 1, #6 → 1, #7 → 0, #8 → 0.
• Entry #1 wins the Curators’ Favorite. Each Curator’s written rationale is published on the contest results page.

Awards stack. An entry that finishes #1 and wins Curators’ Favorite earns 200 credits (Gold + Curators’ Favorite).

A contest is inconclusive if any of the following holds at close:

• Too few entries. Minimum is 4 active entries (after withdrawals and flagged content removal).
• Too few votes per pairing. If the matchup graph doesn’t have enough coverage to estimate a reliable score, the contest doesn’t close with a ranking.
• Pathological voter set. If the only votes come from a single user or a small clustered group, the score is too sensitive to outliers to be trustworthy.

Inconclusive contests stay published — the entries and any rationales remain visible — but no medals are awarded and no credits are earned. The Curators’ Favorite election can still run if nominations exist, since it’s independent of the LensWideOpen Score.

• Entrants-only ballot. You can’t vote in a contest unless you have an active entry in it. Sockpuppet accounts pay the entry cost in credits and have to submit a competing photo, which raises the cost of coordinated voting against the value of any single ranking lift.
• Blind ballots. Voters can’t see creator names, current standings, or prior votes during voting. There’s no signal to vote-trade against.
• Self-vote exclusion. An entrant’s own entry never appears in their voting sheet, so an entrant can’t vote for themselves.
• Exposure balancing. The matchmaker prioritizes least-covered pairs, so a coordinated bloc can’t pile votes onto one specific matchup faster than the rest of the field gets covered.
• Curator panel cross-check. The LensWideOpen Curators vote on every contest with the same blind interface. A ranking that Photographers seem to be gaming will diverge sharply from the Curator panel’s ranking, which is visible to admins.
• Score is opponent-weighted, not vote-weighted. Raw vote count alone doesn’t determine the ranking — quality of the opponents you beat matters (see Section 2 above). Stacking votes against an obvious weak photo doesn’t materially move the ranking.

Still have a question?

The rules above cover the mechanics. For the values and philosophy behind the platform, see Transparency. To send a specific question to the team, use the contact form.