The exact math behind every alert

Multiplicative devig against Pinnacle

Pinnacle is the only sportsbook that welcomes sharp action and adjusts lines based on it. Their prices bake in the wisdom of professional bettors. Their hold is 2-3% on game lines and 4-6% on player props — the tightest in the industry.

For each market, we take Pinnacle’s two-way (or three-way for soccer) prices and strip out the vig using the multiplicative method:

# Two-way (totals, spreads, player props)
p_over  = implied_prob(over_odds)
p_under = implied_prob(under_odds)
overround = p_over + p_under
true_p_over  = p_over  / overround
true_p_under = p_under / overround

# Three-way (soccer moneyline w/ draw)
p_home, p_draw, p_away = implied_probs(...)
overround = p_home + p_draw + p_away
true_p_home = p_home / overround
# ... etc

We use multiplicative over the power method because Pinnacle’s hold is symmetric across sides — there’s no evidence favorites/dogs need asymmetric treatment on their book. Every serious quant devig library confirms this choice for symmetric-vig sharps.

12-book price scan

For every prop Pinnacle prices, we scan the same prop across 12 books:

EV per bet is computed as EV = true_prob × (decimal_odds - 1) - (1 - true_prob) per $1 stake. The book with the highest EV wins the alert.

ProphetXis a peer-to-peer exchange included because they don’t limit sharp bettors — the whole point.

Per-market EV cutoffs

A single global threshold is lazy. Different market types have different vig profiles and different variance, so we use per-class thresholds:

🔥 FIRE tier triggers at 8%+ EV on any market — separate visual label + routed to a dedicated Discord channel.

Outlier detection vs. soft-book consensus

Not all +EV alerts are equal. A soft book at +150 while every other US book is at −140 is a real outlier (probably slow to update). One at +150 while others are at +145 could just be book variance.

For every alert, we compute the best-book’s implied probability distance from the median of the other soft books (excluding Pinnacle), then classify:

Tier-aware fractional Kelly

Kelly criterion assumes your true_prob estimate is exactly right. Reality has estimation noise, so we use fractional Kelly plus a per-tier confidence multiplier:

base_fraction = 0.375        # 3/8 Kelly (conservative default)
max_bet_pct   = 5%           # never exceed this per bet

tier_multipliers = {
    "STRONG": 1.5,           # 0.563 × Kelly, cap = 7.5%
    "MIXED":  1.0,           # 0.375 × Kelly, cap = 5.0%
    "WEAK":   0.5,           # 0.188 × Kelly, cap = 2.5%
}

stake = bankroll * full_kelly * base_fraction * tier_mult
stake = min(stake, bankroll * max_bet_pct * tier_mult)

Stakes shown in units where 1 unit = 1% of bankroll, so any subscriber can apply the sizing to any bankroll.

Auto-hedge calculation on FIRE alerts

When an alert triggers at 8%+ EV on a 2-way market, we automatically check whether the opposite side is available at another book at odds that allow arbitrage. If yes, the alert includes the exact hedge stake to lock in a guaranteed profit regardless of outcome:

# Original bet: stake S at decimal odds d
# Hedge:        stake H at decimal odds e (opposite side)

H = S * d / e
guaranteed_profit = S * (d - 1) - H

No arb possible → the hedge suggestion is omitted rather than recommending a locked loss. Only shown when the math works.

CLV is the only real leading indicator

Win/loss over a small sample is variance, not skill. The real proof a betting system works is Closing Line Value: did you get better odds than the closing price? Sharp bettors evaluate systems by CLV because it converges to true EV faster than P/L does.

Every NovaLine alert has its closing Pinnacle price captured 5 minutes before first pitch/tip-off/opening bell. We track:

• Average CLV % — how much better our entry was vs. the close, averaged across all bets
• Beat-close rate — % of bets where our entry odds beat the closing odds
• 30-day CLV trend — live sparkline on the public tracker so you can see if CLV is drifting up or down over time

Sharp tracking apps like Pikkit report CLV in the same way. Our numbers align when subscribers sync alerts into their Pikkit accounts — verifiable third-party proof of edge.

What we can’t do

Every methodology page should include what the system does NOT do. Ours:

• Markets Pinnacle doesn’t price— batter hits, RBIs, runs scored, singles, doubles, triples aren’t exposed by Pinnacle via our data source. We can’t devig them, so we don’t alert on them.
• Futures / season-long markets — different market shape than daily props. Not currently supported.
• Same-game parlays — correlated outcomes require a completely different model. Not offered.
• Book limits— we can find edges but we can’t place bets for you or prevent DraftKings from capping your stake to $5. That’s why ProphetX (an exchange that doesn’t limit) is in our book list.
• Backtesting the specific strategy— we track live CLV going forward but don’t claim retrospective returns from a backtest. Real trading data only.