How the projections and edges are built

Every number on the board is the output of a repeatable process, not a hunch. This page walks through that process in detail so you can decide for yourself whether the projections are credible and how much weight to put on the edge grades.

Step 1 — The market total

We start with the current consensus market total across major sportsbooks. If books disagree — one book at 47, one at 47.5, one at 48 — we take the median. The market total is the anchor: without it, an edge grade is meaningless. Our projection is always shown alongside the market number so you can see the gap.

Step 2 — Pace and volume

Total points is essentially efficiency multiplied by volume. Volume in football is measured in plays per game, adjusted for game script. We use neutral-situation plays per game — plays run when the score margin is within one score — as the base pace number. Then we adjust for the opponent, because a fast-pace offense playing a slow-pace defense produces a game somewhere in the middle, weighted toward the offensive team's pace.

Step 3 — Efficiency

Efficiency is measured in Expected Points Added (EPA) per play, opponent-adjusted. EPA converts every play into the change in the offense's expected points on that drive, based on down, distance, and field position. Higher EPA per play means the offense creates more scoring opportunities per snap. Opponent adjustment strips out schedule strength — a team that has feasted on bad defenses does not get credit for being elite until the model sees them do it against real defenses.

We pull separate EPA numbers for passing and rushing, then reweight them by each team's projected pass/run split based on personnel, matchup, and script. A team playing from behind will pass more; a team with a lead will run more; the projected script feeds back into the volume math.

Step 4 — Defensive contribution

Every point projection has two teams' offenses and two teams' defenses. Defensive EPA per play allowed, opponent-adjusted, is the mirror image of the offensive number. When a top-10 offense plays a top-10 defense, the model does not simply average them — it applies a small "matchup drag" because good defenses disproportionately punish good offenses' mistakes.

Step 5 — Situational adjustments

• Rest: Short weeks (Thursday NFL, midweek CFB) subtract about 1.5 points from expectation.
• Travel: West-to-east 1 pm ET starts subtract about half a point. Cross-country trips for late kickoffs are roughly neutral.
• Altitude: Mountain West and Denver games add about a point due to fatigue and thin-air ball flight.
• Rivalry / familiarity: Third meeting between two teams in a calendar year (uncommon but happens in college conferences and NFL playoffs) subtracts about a point on average.
• Coach tendencies: Head coach pace and aggressiveness scores from the last 3 years feed a small adjustment when a coach has a distinctive style.

Step 6 — Weather

Weather is pulled from the most recent NOAA forecast for the venue's ZIP code at kickoff time. The three variables that matter are wind speed, precipitation, and temperature. Wind over 15 mph is the biggest lever, reducing passing efficiency and long-play frequency. Steady precipitation matters mostly through fumble rate. Cold under 20°F is a small factor. Dome games get a static +1.5 point bump versus their outdoor equivalents to reflect optimal playing conditions.

Step 7 — Injuries and inactives

Injury data is pulled continuously from official injury reports. Star player designations (quarterback, top wide receiver, top edge rusher, top corner) trigger larger adjustments; role-player designations trigger smaller ones. When we see multiple secondary injuries stacked on one team, we apply a compounding factor because the market tends to price them individually and miss the combined effect.

Step 8 — Edge and verdict

The final projection is the market total plus or minus the sum of every adjustment above. Edge equals projection minus market. Verdict is assigned by edge magnitude: HAMMER for edges of 4+ points, PLAY for 2.5–3.9, LEAN for 1.0–2.4, PASS for anything smaller. Direction is Over when the projection is above the market, Under when it is below.

What we do not do

We do not use "consensus" picks, tout services, or public betting percentages as inputs. We do not follow steam. We do not adjust projections based on which side the model has been "right" on lately. Recency bias is not a signal.

What could still be wrong

Every model is a simplification. Ours works well in normal conditions; it is less reliable in bowl season, in games with 20+ opt-outs, in weeks with fresh coaching changes, and in Week 1 of any season before the current-season data has stabilized. We flag those cases so you can adjust conviction.

How to use the numbers

Treat the projection as a best estimate with a plus-or-minus 2.5 point standard deviation. A 3-point edge is meaningful but not certain. A 6-point edge is rare and usually the result of the market not fully digesting news. Bankroll sizing should reflect edge magnitude, not confidence in a specific outcome.

Backtesting and honesty

We backtest the model against three prior seasons of NFL and college football closing lines. Historical hit rate on PLAY-and-above grades is roughly 54%, which is above the 52.4% breakeven at -110 pricing but not miraculous. Past performance does not guarantee future results; the market gets sharper every year, and our edge is not permanent. We will publish updated backtest numbers each offseason.

Data sources

Schedules and live game data: ESPN. Weather: NOAA. Injuries: official league reports. Market totals: aggregated from major North American sportsbooks. Nothing on the site is scraped from paid data vendors, and nothing is behind a hidden model that we cannot describe.

Feedback improves the model

If you spot a game where the projection looks obviously wrong given a factor we should have caught, send a note through the Contact link in the footer. Reader feedback has caught real bugs — most memorably a broken injury feed in November of a prior season — and every note gets read.

Bottom line

The model is transparent, testable, and improving. It will not make you rich by itself. Combined with disciplined bankroll management, patient line-shopping, and the willingness to pass on 80% of games, it can give you a small, real, sustainable edge on NFL and college football totals — the two markets where a totals-only bettor has the best chance of making the math work.