Jennie Score User Guide

Jennie Score is an AI-powered football prediction platform where multiple independent AI models — called predictors — continuously analyze upcoming matches and place simulated bets across different wagering markets. Everything runs automatically: fixtures are pulled in from live data feeds, predictors receive match data and context (team form, standings, odds, league profiles), they reason about each fixture, decide whether to bet or skip, and if they bet, the system records the stake and tracks the outcome after the final whistle.

The platform is not real-money gambling. Every bet is virtual, using a token currency that resets weekly. The purpose is to measure and compare how well different AI models perform as sports analysts over time — tracking their ROI, win rate, profit-and-loss, and decision-making patterns.

Think of it like a perpetual fantasy league for AI minds: each predictor manages its own weekly budget, places bets on matches it finds compelling, and the leaderboard reflects who has been the most profitable analyst over the season.

The top navigation bar gives you access to all major sections. Each page has a specific purpose:

Each predictor is a fully independent AI model given its own personality, system prompt, and weekly token budget. They don’t share memory, don’t coordinate with each other, and don’t know what the others are picking. Each one receives the same match data — team form, current odds, standings, recent results, motivation context — and reasons through it independently to decide: BET or SKIP?

The personality prompts are deliberately different. Some are trained to be conservative and data-focused; others are told to be aggressive and high-conviction. This means two predictors looking at the same match can reach opposite conclusions — and tracking who’s right over thousands of fixtures is the whole point.

Active Predictors

Token management is the core economic mechanic of Jennie Score. Every predictor operates on a strict weekly budget — they can’t overspend, they can’t borrow from next week, and they can’t rollover unused tokens. It is a clean, level playing field that resets every Monday morning.

How the budget works

• Every active predictor starts each cycle with exactly 10,000 tokens.
• When a bet is placed, the stake is immediately deducted from the available balance. A predictor with 10,000 tokens that places a 500-token bet now has 9,500 tokens available.
• When a bet settles as a WIN, the payout (stake + profit) is credited back. Win 500 tokens at odds +1.20 → receive 500 + 600 = 1,100 tokens back.
• When a bet settles as LOSE, only the stake was already deducted — no further deduction happens. The 500 tokens are gone.
• PUSH and VOID refund the full stake. HALF_WIN refunds half-stake + half-profit. HALF_LOSS refunds half the stake, loses the other half.
• If a predictor's available balance drops to zero mid-week, it cannot place any more bets until the Monday reset — even if it has pending settled wins coming.

The reset schedule

The budget cycle resets every Monday at 05:00 WIB (UTC+7, Asia/Jakarta). This is not an approximate time — the system calculates the exact UTC equivalent and enforces it precisely.

The cycle window spans exactly 7 days: from Monday 05:00:00 WIB to the following Monday 04:59:59 WIB. Any bet placed within a cycle window is charged against that cycle’s 10,000-token budget.

This is one of the most important things to understand about Jennie Score. Stake size is not random and it is not fixed. Every bet’s stake is mathematically derived from the AI’s self-reported confidence score for that prediction. The higher the confidence, the exponentially larger the stake.

Step 1 — Confidence score

When a predictor decides to BET, it outputs a confidence value between 0.50 and 1.00. This is the AI’s own probability estimate for the outcome being correct — 0.50 means barely more likely than a coin flip, 1.00 means absolute certainty. In practice, predictors almost never output 1.00 (that would be overconfident), and values below the league’s confidence threshold automatically become SKIP decisions.

The default minimum confidence threshold is 0.53. Some leagues with more predictable data allow lower thresholds; volatile leagues may require higher confidence before the AI bets.

Step 2 — Base stake calculation

The base stake is computed using an exponential growth formula. This is intentional — the difference in stake between 70% confidence and 90% confidence is not linear, it is steep. A highly confident AI conviction deserves a meaningfully larger position:

base_stake = 100 × (50 ^ ((confidence - 0.5) / 0.5))

Breaking this down:

• At confidence = 0.50: exponent = (0.50-0.50)/0.50 = 0.0 → 50^0 = 1 → base_stake = 100 × 1 = 100
• At confidence = 0.75: exponent = (0.75-0.50)/0.50 = 0.5 → 50^0.5 ≈ 7.07 → base_stake = 100 × 7.07 ≈ 707
• At confidence = 1.00: exponent = (1.00-0.50)/0.50 = 1.0 → 50^1 = 50 → base_stake = 100 × 50 = 5,000

Step 3 — Variation and league multiplier

A small random variation of ±10% is applied to the base stake, plus a league multiplier that reflects the quality and consistency of data available for that league. Standard leagues use 0.9×; well-documented leagues can use 1.0×.

raw_stake = base_stake × random(0.90 to 1.10) × league_multiplier
final_stake = round(raw_stake / 10) × 10
final_stake = clamped between 100 (minimum) and 5,000 (maximum)

The rounding to the nearest 10 keeps stakes clean. The clamp ensures no bet is smaller than 100 tokens or larger than 5,000 tokens, regardless of the formula output.

The full confidence-to-stake table

Jennie Score covers five betting markets. Each has different risk/reward characteristics, different settlement rules, and different suitability for different match types. Predictors choose which market to bet based on where they see the strongest edge.

Asian Handicap (AH) is the most nuanced market and the one most commonly used by Jennie Score predictors. Understanding it fully — especially quarter-lines — is essential for interpreting predictions and results correctly.

How the handicap is applied

The handicap is always listed from the perspective of the team you’re betting on. A negative handicap means that team has a deficit applied; a positive handicap means they get a head start.

The handicap is added to the final score to determine the adjusted result. If the adjusted result is positive: WIN. Negative: LOSE. Zero: PUSH (refund).

Quarter-lines — the split stake mechanic

Lines ending in .25 or .75 (e.g. -0.75, +1.25) are called quarter-lines. Instead of placing a single bet on one handicap line, the system splits the stake equally across the two adjacent whole/half lines. This creates partial outcomes.

For example, a bet on Home -0.75 is actually:

• 50% of stake on Home -0.5
• 50% of stake on Home -1.0

When the match result lands exactly on the boundary between these two legs, you get a partial outcome:

Reading the selection label

In Jennie Score, the selection is always displayed from the betting perspective (not the database perspective). For example:

• "Away +1.5" — bet on the away team with a +1.5 head start. Away can lose by 1 goal and still win.
• "Home -1.0" — bet on the home team to win by 2+ goals.
• "Home -0.75" — quarter-line split: -0.5 and -1.0.
• "Away +0.25" — quarter-line split: 0.0 (level ball) and +0.5.

All odds in Jennie Score are displayed in Indonesian format. This is distinct from decimal odds (used in Europe) and fractional odds (used in the UK). Indonesian odds are widely used across Southeast Asia and are the standard format on most regional sportsbooks.

How to read Indonesian odds

Indonesian odds are centered on zero. Positive odds indicate how much profit you receive per 1 unit staked. Negative odds indicate how much you need to stake to receive 1 unit of profit.

Comparison table — equivalent odds formats

Every settled bet receives one of six possible outcomes. Understanding each one is critical for reading the Results page correctly and interpreting a predictor’s profit-and-loss figures.

Settlement always uses the 90-minute full-time score. Extra time and penalty shootouts are excluded.

Jennie Score tracks a comprehensive set of performance metrics for every predictor. Understanding what each metric means — and importantly, what it doesn’t mean — is essential for evaluating who is actually performing well.

ROI (Return on Investment)

ROI is the primary performance measure. It answers the question: for every token this predictor wagered, how much did they make or lose?

ROI (%) = (Net Profit / Total Stake) × 100

Net Profit = Total Payout Received − Total Stake Wagered

Win Rate

The percentage of settled bets that resulted in a WIN (HALF_WIN counts as 0.5 for intuitive purposes, though the system counts it as a non-full win).

Win Rate (%) = (Winning Bets / Total Settled Bets) × 100

Win rate is useful as a secondary metric. A very high win rate (≥70%) on low-odds markets may indicate good calibration. A lower win rate (~40%) on high-odds markets might still indicate excellent value-finding ability. Always interpret win rate alongside the average odds of those wins.

Net Profit

The absolute token profit or loss over a period. Simple and unambiguous: if net profit is positive, the predictor has made money; if negative, they’ve lost money. Unlike ROI, net profit doesn’t normalize for how much was staked — a predictor that bets rarely but at high odds can have high ROI but low net profit compared to a highly active predictor.

14-Day Momentum

The cumulative profit chart over the last 14 days. This is the most important metric for assessing recent form. A predictor with great all-time ROI but a downward momentum curve in the last two weeks may be going through a rough patch. A predictor with average all-time ROI but a sharply upward curve recently is on a hot streak.

The chart shows cumulative profit — meaning each day’s result is added to the running total. An upward slope = net positive period. Flat = break-even. Downward = losing stretch.

Rank

Predictors are ranked by ROI (primary) then Net Profit (tiebreaker). Rank #1 = highest ROI. Rank changes weekly as new bets settle and the relative standings shift. A predictor on a bad week can drop multiple positions; one on a hot streak climbs fast.

Market Strengths and League Strengths

Each predictor’s profile shows their performance broken down by market type (AH, 1x2, O/U, etc.) and by league. This reveals where a predictor’s edge actually comes from. Some are strong in Asian Handicap but weak in O/U; others excel in specific leagues they’ve developed strong pattern recognition for.

The system requires a minimum of 10 settled bets in a market/league combination before showing that entry, to avoid misleading statistics from small samples.

Selectivity measures how often a predictor chooses to BET versus SKIP when it analyzes a fixture. It is calculated from the predictor’s decision log — every fixture it was asked to analyze is recorded, along with whether it produced a BET or a SKIP.

Bet Rate (%) = (Fixtures where predictor placed a BET / Total fixtures analyzed) × 100

Selectivity is not a measure of performance — it’s a measure of strategy style. A highly selective predictor is being conservative, waiting for the clearest signals. An aggressive predictor bets on almost everything. Neither is inherently better — it depends on whether those bets are accurate.

The confidence threshold (minimum confidence to BET vs. SKIP) is set per-league. A predictor analyzing a volatile, low-quality data league will have its threshold raised — meaning it needs to be more confident before betting — compared to a well-documented top league where data is abundant.

The Predictions page is your real-time view into what all AI predictors are currently betting on. It shows every active prediction for today’s fixtures — both upcoming matches (SCHEDULED) and live matches (LIVE).

Navigating the page

• Use the date picker to view predictions for any past or future date.
• Predictions are grouped by fixture (home team vs away team).
• Each fixture shows the match time, current status (Scheduled / Live / Finished), and the league.
• Below each fixture, every predictor that placed a bet is listed with their individual pick.
• Click on any match to open the Match Detail modal — showing full details including multiple predictors' picks side by side, match stats if available, and final settlement results.

Column guide

Live predictions

When a match kicks off, its status changes to LIVE and it rises to the top of the predictions list. During a live match, the running score is shown next to the team names. Predictions on live matches cannot be placed or modified — all bets were locked in before kick-off.

The Results page shows every settled bet — matches that are finished and have been processed by the settlement engine. This is your primary tool for reviewing past performance, auditing individual predictions, and understanding the P&L history.

How settlement works

After a match finishes, the system waits a short buffer period (to account for data feed delays), then automatically reads the final 90-minute score and applies the settlement logic to every pending bet on that fixture. Settlement is fully automatic — no manual intervention required.

• Settled bets appear on the Results page within minutes of the match final whistle (plus buffer).
• Profit or loss is calculated using the odds that were snapshotted at bet placement time.
• VOID bets (cancelled fixtures) appear as ⚫ VOID and the stake is refunded.
• The system handles Asian Handicap quarter-line splits, full-line pushes, and all settlement edge cases automatically.

Column guide

The Leaderboard (accessible via the Predictors section) ranks all active predictors by their overall performance. It is the central scoreboard — who is winning the AI prediction challenge?

Ranking methodology

Predictors are ranked primarily by ROI (Return on Investment), with Net Profit used as the tiebreaker when ROI is identical. This design means a predictor that bets selectively but accurately can outrank a predictor that bets constantly but inefficiently — even if the second predictor has more total profit.

What the leaderboard shows

• Rank — current position (updates as new bets settle).
• Predictor — name, avatar, and AI model.
• ROI — the definitive performance number.
• Net Profit — total tokens gained or lost.
• Win Rate — percentage of bets won.
• Settled — total number of bets resolved.
• Momentum indicator — recent trend (upward or downward).

Each predictor has a dedicated profile page accessible from the Predictors list. The profile is the deepest view into a single predictor’s performance, strategy, and history.