lfg2025/botfights
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: add answer confidence differential as scoring factor

Browse Source 
When both bots are correct, the one with higher checkAnswer confidence
(exact match 1.0 vs fuzzy match 0.8) gets up to +1.0 bonus points.
This rewards precise answers over approximate ones, adding another
competitive dimension beyond pure speed.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Dorian
2026-03-13 04:45:53 +00:00
co-authored by Claude Opus 4.6
parent cb8a45cfb3
commit c6c792dd9e
2 changed files with 26 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
server/src/engine/scoring.test.ts
+19
View File
@@ -218,6 +218,25 @@ describe('scoreRound', () => {
expect(typeof result.narration).toBe('string')
expect(result.narration.length).toBeGreaterThan(10)
})
it('answer confidence gives bonus — exact match (1.0) beats fuzzy match (0.8)', () => {
// Use answers where one answer is a substring of another
// "satoshi" against ["satoshi nakamoto"] scores 0.8 (containment)
// "satoshi nakamoto" against ["satoshi nakamoto"] scores 1.0 (exact)
const challenge = makeChallenge({ answers: ['satoshi nakamoto'] })
// Both answer at equal speed, but A has exact match, B has fuzzy
const result = scoreRound(
challenge, botA, botB,
makeResponse('satoshi nakamoto', 500), // exact → 1.0
makeResponse('nakamoto', 500), // contained → 0.8
null, 0, 0,
)
// A should score higher due to confidence differential
expect(result.botAScore).toBeGreaterThan(result.botBScore)
expect(result.winnerId).toBe('a1')
})
})
describe('calculateElo', () => {
server/src/engine/scoring.ts
+7 -4
View File
@@ -131,12 +131,15 @@ export function scoreRound(
const confA = Math.min(correctA, 1)
const confB = Math.min(correctB, 1)
// Confidence differential: exact match (1.0) vs fuzzy (0.8) gives up to +1.0 bonus
const confDiff = confA - confB
if (aFaster) {
scoreA = FACTUAL_FASTER_BASE + (1 - speedRatio) * SPEED_ADVANTAGE_MULTIPLIER + confA
scoreB = FACTUAL_SLOWER_BASE + speedRatio * SPEED_RATIO_MULTIPLIER + confB * CONFIDENCE_BONUS
scoreA = FACTUAL_FASTER_BASE + (1 - speedRatio) * SPEED_ADVANTAGE_MULTIPLIER + confA + Math.max(0, confDiff)
scoreB = FACTUAL_SLOWER_BASE + speedRatio * SPEED_RATIO_MULTIPLIER + confB * CONFIDENCE_BONUS + Math.max(0, -confDiff)
} else {
scoreA = FACTUAL_SLOWER_BASE + speedRatio * SPEED_RATIO_MULTIPLIER + confA * CONFIDENCE_BONUS
scoreB = FACTUAL_FASTER_BASE + (1 - speedRatio) * SPEED_ADVANTAGE_MULTIPLIER + confB
scoreA = FACTUAL_SLOWER_BASE + speedRatio * SPEED_RATIO_MULTIPLIER + confA * CONFIDENCE_BONUS + Math.max(0, confDiff)
scoreB = FACTUAL_FASTER_BASE + (1 - speedRatio) * SPEED_ADVANTAGE_MULTIPLIER + confB + Math.max(0, -confDiff)
}
} else if (correctA > 0 && correctB === 0) {
scoreA = ONE_CORRECT_WINNER_BASE + correctA * CONFIDENCE_BONUS