We scored every entry without hearing a single note. Here's how we did.
Everyone knows that Eurovision voting is geographically tied. It's an unspoken agreement that hangs over every result — the Balkan bloc, the Scandinavian sympathy vote, the diaspora surge. I wanted to understand just how much that actually mattered. Could a substandard song accumulate enough from being simply from the right location at the right moment? Or are there more important criteria hiding underneath the politics?
The real question underneath all of that is simpler: is Eurovision actually a song contest? Or is it a system the right country can game — hit the Loreen button, optimise for the instrument rather than the art, and collect your trophy.
So we built an instrument and applied it. Without hearing a single song. Just from what's been written about them, what's been performed at national finals, and what's happening in the world right now. It might get everything wrong. But it might also show that Eurovision is considerably less random than it looks — and that the market, for all its confidence, is missing something obvious.
The results are in. Bulgaria won. We called it a dark horse at 2% implied odds. The market was looking elsewhere. The instrument wasn't.
The Grand Final is over. Bulgaria won with 516 points — the dark horse the instrument flagged at 2% implied odds. Here's the full honest reckoning: where the framework was right, where it overcalled, and what it means for the next version.
"Bulgaria scored 7.21 Combined — ranked 5th by the instrument. The market had it at 2% implied. It won. Our dark horse call was the most valuable signal in the dataset, and the most ignored one at the time."
The Instrument — Post-Contest VerdictThe instrument ranked it 5th. The market had it at 2% implied. It won both jury and televote — the first double victory since 2017. Originality scored 9, joint highest in the field. Balkan-trap with sampled Kukeri bells. The instrument saw it. The market didn't. This is the result that validates the whole project.
Top score in the instrument at 7.85 Combined. Called undervalued by the market at 8%. It placed 10th — inside the top ten, which validates the core signal, but far short of the win. The framework read the song correctly; it underweighted how the live staging would fail to translate the studio promise.
The market had Finland at 31% implied. The instrument had it third. It finished sixth. The overvaluation call was correct — it didn't win, it didn't even medal. A 25-point gap separates the instrument's merit rank from the bookmakers' implied favourite. The market paid a premium it couldn't cash.
The instrument ranked Israel 12th. It finished second with 343 points, driven by heavy televote support. The geo/diaspora scoring (9) was correct on direction — this is exactly the voting pattern the instrument described. The score underweighted how dominant that single variable would become on the night.
The instrument's "biggest overvaluation" call. Ranked Denmark 14th. Market had it 3rd. It finished 7th — better than the instrument said, but the market was still wrong to have it as a potential winner. A partial hit: the direction of overvaluation was correct, the magnitude was overstated.
Scored 7.24 Combined — ranked 4th by merit. Finished 3rd overall with 296 points. This was one of the instrument's better reads: a returning country with strong originality and staging ambition. The market underpriced it too; the instrument had it higher than the bookmakers did.
This was the instrument's prediction before the Grand Final — ranked by Combined Lens score, built entirely from written research, national final performances, and market signals. No hearing required. Compare against the actual results in the section above.
Maximum scores from both jury and public at national selection. Geo/diaspora potential of 9. The clearest undervalue in the field — the market has it at 3% implied. At that price, this is where the instrument most strongly disagrees with the bookmakers.
Highest Combined score in the entire field. Won national final with maximum points from all three voting categories. Already topping Spotify Viral 50 in Greece and Cyprus. The "mama" bridge is the emotional pivot of the contest. The bookmakers have this eighth. That gap is the story of this page.
Legitimately great. 30+ page staging document, pyrotechnics, violin virtuoso. The market favourite at 31% implied — but only third by merit. The premium the market has paid here is the defining mispricing of this contest.
"The top three by Combined Lens were Greece at 7.85, Ukraine at 7.79, and Finland at 7.57. The market priced them at 8%, 3%, and 31% respectively. The dark horse was Bulgaria at 2%. It won. The spread was the sharpest, most defensible finding in this analysis — and the result confirmed the market was wrong about what mattered."
The Instrument — Combined Lens Rankings · Pre-ContestHyper-techno banger with genuine social commentary and an authentic artist backstory. Already viral regionally. The instrument's top score across the entire field — yet the bookmakers have it level with Australia. Four places above where the market has priced it.
Maximum jury and public scores at Vidbir. Bandura-led Ukrainian folk tradition. Geo/diaspora potential of 9. Eight positions above where the market has placed it. At 3% implied, this is the single largest rank gap between merit and market in the dataset.
Finland is not bad. Finland is just not this good. Critics note a weak hook and chorus-title disconnect. The market has it first; the instrument has it third. Two positions — closer than it looks — but the 31% implied probability implies near-certainty the scores don't support.
The biggest single overvaluation in this dataset. The instrument ranks Denmark 14th — a solid jury-bait ballad, low on originality, low on earworm. The market has it third. Eleven positions separate the bookmakers from the scores. That gap needs a better explanation than the music provides.
Balkan-Trap-Pop with sampled Kukeri bells — originality scored 9, the joint highest in the field. Returning country after a three-year absence. The ESC Insight predictive model had been moving this upward week on week. The market was at 2% implied. The instrument said dark horse. It won with 516 points — dominating both jury and televote.
Jury lens of 7.52 is genuinely strong — elegant, polished, classic jury-bait. But the public lens drops to 6.69 because the earworm simply isn't there. If France wins, the televote will have been split and a jury-led result will have carried it. Seven positions above its merit rank.
Every entry scored and ranked. Position Delta shows the gap between where the bookmakers rank an entry and where the instrument ranks it by merit. Positive = market has underestimated it. Negative = market has paid a premium it hasn't earned.
Both the betting market (by implied win %) and the instrument (by Combined score) produce a ranking of all 35 entries. The delta is bookmaker rank minus merit rank. Positive = Undervalued — ranked further down the field than the scores say it deserves. Negative = Overvalued — placed higher than the merit justifies. Within ±3 positions = Solidly Priced. Tied entries share an average rank.
Click any entry to expand full scores. Position Delta = bookmaker rank minus merit rank. Positive = Undervalued · Negative = Overvalued · Within ±3 = Solidly Priced.
The weights were fixed before a single entry was scored. Ten criteria, scored 1–10. Three differently weighted lenses. Every number is visible and every weight is arguable — that's the whole point.
What professional juries reward. Vocal ceiling and originality carry the most weight. Geo/diaspora potential registers at just 3% — versus 15% in the public lens. That structural gap is where the two systems most fundamentally disagree.
What casual viewers and diaspora communities actually vote for. Earworm and emotional resonance dominate. Geo/diaspora potential at 15% here versus 3% in the jury lens — the most structurally honest acknowledgement in this whole framework that Eurovision is not purely a song contest.
The instrument's best model of what actually wins — a hybrid that rewards entries capable of performing across both voting systems. This is the primary ranking lens. An entry that dominates on jury but collapses on public, or vice versa, will not top this ranking.
The same note that runs on every Sandbox project. The instrument was built before any scores were applied. These are not predictions. They are a structured opinion, made visible and arguable.
Structured estimates built from publicly available information — national final results, critical reception, fan community reaction, streaming data, and betting market signals, cross-referenced against each other. Every score is arguable. The methodology is there to be challenged.
Predictions. Empirical data. Live performance evaluations — staging won't be fully visible until Vienna. The geo/diaspora scores are based on historical voting patterns. Israel's score of 9 reflects documented government advertising campaigns from prior years, not a value judgement on the entry.
Greece was upgraded most significantly from an earlier pass: earworm to 9, staging ambition to 9, emotional resonance to 8 — all research-justified after reviewing national final performances and fan response. Poland was downgraded: originality and earworm both reduced after multiple reviewers called it technically strong but generically forgettable.
Bulgaria won. Our dark horse — flagged at 2% market implied odds, scored 7.21 Combined, ranked 5th by merit — took the trophy with 516 points. Greece placed 10th (predicted 1st). Finland placed 6th (market said 1st, we said overvalued). 7 of our predicted top 10 finished in the actual top 10. The framework didn't get the winner right by Combined score, but it got the market's biggest mispricing right.
Built in conversation with Claude by The Unanswered. Part of the Sandbox alongside No Guilty Pleasures and No Soft Opinions. Not an Anthropic product. Not sponsored by any bookmaker.