Confidence Tiers Explained
Answer Hub · System
Every Otterline pick belongs to a named bucket. Each bucket has defined thresholds for consensus percentage, combo win rate, and market pricing. This page documents exactly what qualifies a pick for each tier and how to use that information.
| Bucket | Sport | Min consensus | Min combo W% | Volume |
|---|---|---|---|---|
| Power Plays | NHL | 56% | 55% | High |
| Diamond Picks | NHL | 56% | 55% | Medium |
| Mladek Spotlight | NHL/NBA | — | — | 1/day |
| Value Edge | NHL | Any | 58% | Low |
| Verified Value | NBA | 56% | — | Medium |
| NBA Sniper | NBA | 56% | — | Very low |
Power Plays
NHLThreshold: 56%+ consensus · combo win rate ≥ 55%
Volume: High — multiple picks per day
Focus: Consistency and daily volume
- Power Plays are the workhorse bucket — the highest-frequency NHL plays that pass the baseline agreement and combo win rate filters.
- The 56% consensus threshold means at least two of the four models agree on the same side. The combo win rate filter ensures the specific agreeing pair or trio has historically been above 55% together.
- Because Power Plays are designed for volume, price constraints are relaxed. They may include moderate chalk, but not extreme favorites (-300+).
- Recommended unit sizing: 0.5–1 unit per game, flat across all Power Plays for the day.
Diamond Picks
NHLThreshold: 56%+ consensus · combo win rate ≥ 55% · no heavy chalk
Volume: Medium — 2–5 per day typical
Focus: Agreement + historical edge + price awareness
- Diamond Picks require a >=55% historical combo win rate plus a no-heavy-chalk price filter.
- Heavy chalk is excluded regardless of consensus level. A -350 favorite with 100% model agreement is not a Diamond Pick — the risk-adjusted value isn't there.
- Diamonds are the core curated bucket for users who want a smaller shortlist of higher-confidence plays without hunting for value every day.
- Recommended unit sizing: 0.5–1 unit per game, flat.
Mladek Spotlight
NHL / NBAThreshold: Mladek model signal + value gap vs. market + underdog lean
Volume: One per day
Focus: Best model-vs-market value play of the day
- The Mladek Spotlight is a single daily pick generated specifically from the Mladek Model's output combined with market-implied probability comparison.
- It targets spots where the Mladek model's confidence meaningfully exceeds the implied probability from market lines — the gap is the edge.
- Underdogs and mid-range pricing (not extreme chalk or extreme dogs) are prioritized. The Spotlight is where Otterline's contrarian value logic is most concentrated.
- Because it's one pick per day and uses a specific model, the Spotlight has a tighter historical sample. It should be treated as a supplementary play, not the foundation of your daily action.
- Recommended unit sizing: exactly 0.5–1 unit, never more.
Value Edge
NHLThreshold: Combo win rate ≥ 58% · model confidence exceeds market implied by ≥4 pts
Volume: Low — 0–2 per day
Focus: Market mispricing and underdog value
- Value Edge is a price-first filter — picks only qualify if there's a meaningful gap between model confidence and what the market implies.
- The market comparison uses closing line or pre-game Polymarket probabilities. A team implied at 48% by the market but rated 54% by the models is in the window; a -200 favorite rated at 68% by the models is not.
- Value Edge skews toward underdogs and games where public action may have moved lines away from fair value. This bucket will have lower volume than Power Plays but is designed for better expected value per unit wagered.
- Not every day will have a qualifying Value Edge pick. Expect 0–2 on average. Forcing a pick when the criteria don't align is not part of the system.
Verified Value
NBAThreshold: Multi-model consensus + price-aware filter
Volume: Medium — 2–5 per day typical
Focus: NBA consensus with value component
- Verified Value is the NBA equivalent of the Diamond/Power Play combined bucket — it requires both model agreement and a passing grade on the market comparison filter.
- NBA presents more lineup variance than NHL, so Verified Value is particularly sensitive to late injury and rest news. When available, lineup flags can override or downgrade picks.
- The value component prevents the bucket from just being 'pick the favorite every night.' Agreement without a market edge doesn't qualify.
- Recommended unit sizing: 0.5–1 unit per game.
NBA Sniper
NBAThreshold: Strict consensus + Polymarket implied 40–59% window
Volume: Very low — 0–3 per week typical
Focus: Tight market window, high model conviction
- NBA Sniper targets the most specific value window: games where the market prices a team at 40–59% implied probability AND the models show high conviction on that same side.
- The 40–59% window is intentional — it captures near-coin-flip games where even a small model edge is actionable, while avoiding games where one team is a heavy favorite and the public has already priced in the expected result.
- Volume is very low by design. If no game meets the criteria on a given day, no Sniper pick is published. Forcing a pick outside the window defeats the purpose.
- Sniper is for users comfortable with patience and low-volume, high-precision staking. It is not a standalone daily system — combine it with Power Plays or Verified Value for daily action.
Thresholds are periodically reviewed against rolling performance data. The goal is always to maintain signal quality — if a bucket's historical win rate regresses below its threshold over a meaningful sample, the criteria are tightened. See the full methodology for the complete pipeline, or the FAQ for common questions about how to use each bucket day-to-day.