How to Interpret the Projections Page

The Projections page is where SharpSideBaseball turns every scheduled MLB game into a final-score prediction to the tenth of a run. It's the quantitative backbone of the site — the daily article, Best Bet, and the "Opportunity" flags all trace back to these numbers. This guide is written for the bettor who already knows how to read a line and wants to use the projections as one input in a wider process, alongside Live Signals, Heat scores, and Market Split warnings.

What each row shows

Every game gets one card with:

• Projected final score to the tenth (e.g. Yankees 4.6 – Red Sox 4.1). The higher-scoring team is our projected winner; the total is the sum.
• Projected side lean — ML, run line, and total, each compared to the current consensus market price. If our number crosses the market by enough, the pick is tagged as an Opportunity.
• Best Bet — at most one per slate, chosen from the strongest, cleanest edges. Best Bets are capped at −300 and are skipped when a game is flagged as conflicted (see below) or when a team is missing two-plus core hitters.
• Market Split ⚠ — an amber badge when our projection disagrees with where the sharps are pushing the line. Conflicted picks are sorted to the bottom of the list and are ineligible for Best Bet.
• Lock icon — appears once the game is inside 30 minutes of first pitch. The version shown when the lock hits is the version that gets graded on the Accuracy tab.

How the number is derived

The projection is a blended model, not a single formula. In rough order of weight:

1. Starting pitchers — season and last-30-day ERA, FIP, xFIP, K/9, BB/9, HR/9, xwOBA allowed, and hard-hit percentage. For pitchers who spend most of their starts at hitter-friendly parks (Colorado, Cincinnati, Fenway, etc.), we apply a road-split adjustment so their raw stats aren't over-punished.
2. Openers and bullpen games — when a confirmed opener is announced, we shorten his expected IP to ~1–2 and project the bulk innings from the team's bullpen composite. If a bulk reliever is publicly reported (via beat writers), the daily article names him.
3. Team offense — season wOBA and last-10 runs per game, with a weighting toward recent form.
4. Star-player absences — teams missing multiple core hitters (Judge, Stanton, Betts, etc.) get a run-scoring penalty and are blocked from Best Bet eligibility.
5. Ballpark and weather — park factors plus wind, temperature, and humidity from the live weather feed.
6. Market blending — the final number is anchored partly to the market consensus so we don't drift away from where the sharps are pricing the game.
7. Totals nudge — projections are nudged off half-run lines (8.5, 9.5) so we don't grade a coin-flip as a hit or miss on the wrong side.

Beyond these seven inputs, the model also draws on additional advanced analytics, proprietary systems, and sabermetric data — so the final number is the product of a broad toolkit, not any single stat.

The Over/Under lean is then simple: if we project ≥ 8.5 total runs it's an Over lean; < 8.5 is Under. The ML lean goes to whichever team we have scoring more runs.

When it updates — and when it locks

• The page refreshes throughout the morning and afternoon as lineups drop, starters get confirmed, weather sharpens, and market prices move.
• The daily article built on top of the projections publishes at 11:30 AM ET each day.
• Each individual game locks 30 minutes before first pitch. Whatever projection, lean, and Best Bet flag exist at that moment are frozen — that's the version graded on the Accuracy tab, and it's the version Market Split flags snapshot to (see next section).

If you refresh the page after lock, you'll still see the frozen values with a lock indicator. Live in-game score is not part of the projection — this page is a pre-game tool.

Reading Market Split ⚠ warnings

A Market Split flag fires when the projection says one thing and the sharps say another. Specifically:

• The public handle is heavily on one side, but the line has moved against that side (classic reverse line movement).
• A majority of our tracked books have repriced in the same direction across the last few snapshots.
• The disagreement is on a market row (ML, RL, or total) where we currently show a lean.

When you see the amber badge, treat that game as decision-worthy but not obvious. Our number likes one side; the money likes the other. Neither is automatically right — but the pick is disqualified from Best Bet, and a smart bettor should either pass, size down, or wait for another signal (a Live Signals Heat spike on the opposite side is usually the tiebreaker).

Once first pitch is thrown, the Market Split state freezes to whatever it was pre-game. In-play line moves do not retroactively change the flag.

Using it alongside Live Signals

Projections are your thesis. Live Signals are your timing and confirmation. A workflow that works:

1. Open Projections in the morning. Note the games where our lean has real distance from the market — especially anything tagged Opportunity or Best Bet.
2. Cross-check those games on Live Signals. If sharps are agreeing with the projection (Heat rising on our side, RLM in our direction), that's confirmation. If Live Signals show sharps on the other side, expect a Market Split flag to appear on the Projections card — and treat that game with caution.
3. Read the daily article at 11:30 AM ET. It narrates the highest-conviction game(s), calls out any pitcher-role uncertainty (opener vs. starter, bulk reliever coming in), and explicitly acknowledges any market disagreement instead of hiding it.
4. Use Heat Alerts to get pinged if a signal fires late on a game you're already tracking.

Common mistakes to avoid

• Chasing the projection alone. A 0.3-run edge on a game with a Market Split warning is a coin flip dressed up as a lean. Wait for market confirmation.
• Ignoring the lock. If you're comparing your grading of a pick to ours, compare against the locked version — mid-day snapshots will differ.
• Treating openers like starters. When the "SP" row is flagged as an opener, the bullpen composite is doing most of the projection work. Read the article's pitcher block for the story.
• Over-weighting yesterday. The model already blends last-10 offense and last-30 pitcher form. Recency is priced in; you don't need to double-count it.

In one sentence

The Projections page tells you what the model thinks — Live Signals, the Market Split flag, and the daily article tell you when to trust it. Use them together.