RPS 90+ · NUCLEAR

Strongest edges on the board. All three axes pulling the same direction, with the math axis showing meaningful divergence (typically 20+ points handle/bets gap), reverse line movement, or both. 1.0 unit standard. Capped at 2 per day under the v2.1 ruleset.

RPS 85–89 · ELITE

Premium plays. Two of three axes strongly aligned, third axis at minimum neutral. Often a sharp side without dramatic divergence — money flow tells you the side, fundamentals confirm it. 1.0 unit standard.

RPS 75–84 · STRONG

Solid edge. Edge is real but not extreme. Most board picks fall in this range over the course of a season. 1.0 unit standard, dropping to 0.5u when the line is +150 or longer (keeps variance controlled at long prices).

RPS 65–74 · LEAN

Thin edge — there's a signal but it's not strong enough to play full size. 0.5 unit, optional. If you skip LEAN picks entirely you're not leaving much on the table.

RPS <65 · PASS

No actionable edge. The game is logged on the board for transparency — you see what got considered and rejected. 0 units. Don't play it. A PASS that "looks good" is exactly the bias the engine exists to filter out.

Divergence — Handle vs. Bets

The most-cited metric on the board. Divergence = |Handle% - Bets%|. When 20% of tickets are holding 80% of the money on a side, that's big money on small ticket count — i.e. sharps. When the split is even, public and sharp money agree, and there's no extra signal.

• Under 10pt → no actionable signal
• 10–20pt → notable, contributes to score
• 20–30pt → strong, ELITE territory
• 30+pt → NUCLEAR territory

Reverse Line Movement (RLM)

The line moves against the side getting majority of the bets. Books only move lines for one of two reasons — risk management, or someone with size betting the other way. RLM after a high bet count on the favored side is the cleanest "smart money on dog" indicator the market gives you.

Today's example: 76ers @ Celtics opened BOS -9.5, drifted to BOS -7.5/-8.5 across the market, even with PHI taking 70% of bets on the spread. That's money flowing onto Philadelphia despite ticket count favoring them — classic RLM into the dog.

Steam

Coordinated rapid line movement across multiple books at the same time. Steam is the loudest market signal — it means a sharp group has hit the line at every book they have access to within minutes of each other. The board flags steam events when detected during the day.

Implied Probability vs. Model Probability

The MathRinger axis converts the line to an implied probability (de-juiced), then compares it against the engine's own win probability for that side. Kelly fraction is computed from the gap. If implied probability matches model probability, edge is zero. The bigger the model-implied gap, the bigger the math contribution to RPS.

NBA

Full engine — moneyline, spread, totals, key player props. Heavy weight on injury status (NBA games are most affected by single-player absences of any major sport). Rest advantage and back-to-back tracking factor into FES. Playoff series weighting separate from regular season — Game 7 home dog adjustments built in.

NHL

Full engine — moneyline, puck line, totals. Goaltender start is a heavy FES input (often the biggest single-game factor). Special teams differential and recent road/home splits weighted. Playoff series weighting built in.

MLB · v2.1 ruleset

Moneyline only. No run lines, no totals, no first-five — until a 30+ games-per-team baseline is locked. Standard tier sizing applies; STRONG at +150 or longer drops to 0.5u to keep variance controlled. Same general 65 RPS minimum and 2-NUCLEAR-per-day cap as the rest of the engine.

Horse racing

Not currently scored. The MathRinger module for thoroughbred markets doesn't exist yet. Major-stakes races may be referenced for context (Kentucky Derby, Preakness, Belmont) but no RingerBets pick will ever be issued on a horse race until the model exists. If we don't have edge, we don't pretend.

Not a guarantee. Variance exists. Even a 60% win rate on 1u plays loses money over a small sample of unlucky nights. The engine identifies edges; the dice still roll.

Not a "+EV lock club." No service can promise +EV across every pick. RPS is a confidence model, not a probability oracle.

Not for chasing. Bankroll discipline is on you. The board gives you sized plays. If you lay 3u on a NUCLEAR because you're stuck, that's not the engine.

Not a predictor of single-game outcomes. It's a long-run edge identifier. 50+ pick samples are the smallest meaningful unit of evaluation.