How the Picks Are Made

Every recommendation on the board comes from the same stats model. Here is the logic behind each type of bet, in plain English.

The foundation: SP+ power ratings

Every team carries an SP+ rating for its offense and its defense — a points-per-game measure of how good each unit is. The model turns those into an expected margin for the game:

projected margin = (home rating − away rating) + 2.4 home-field points
    (home-field = 0 at a neutral site)

That single projection drives everything: it becomes a projected spread, and the four offense/defense ratings combine into a projected total. Each of those is then compared with Saracen’s posted number to find the edge.

Early-season caution: in Weeks 1–2 the ratings still lean on preseason data, so the model deliberately trusts itself less — it demands a wider gap before it flags anything. That fade eases off and reaches full weight by about Week 5.

Spread

Against the spread (ATS)

the default pick

A spread bet is on the margin — a favorite has to win by more than the number, an underdog can lose by less than it (or win outright). The model compares Saracen’s spread to its own projected spread; the difference is the edge.

RuleRecommend a side when the edge is at least 2.0 points and the spread is 24 or fewer. Huge blowout spreads are skipped — there the number is about garbage time, not a real matchup.
  • Model likes the favorite more than the line → bet the favorite.
  • Model has the game closer than the line → take the underdog + points.

Confidence starts at 50% and rises with the size of the edge, but is held back on big spreads (over 14 points), where blowouts are noisier. It lands in roughly the 51–66% range. If the line is also Market Soft (below), that’s flagged as extra value.

Total

Over / Under (totals)

the Total tab

A totals bet is on the combined points both teams score, versus a posted number. The model builds a projected total from the four SP+ offense and defense ratings, then compares it to Saracen’s line.

RuleLean a side when the projection differs from the line by at least 2.0 points. Projection higher than the line → Over; lower → Under.

Confidence runs about 51–64% — the wider the gap between projection and line, the higher it goes. Totals are worth a look even on blowout spreads, where the winner is obvious but the scoring pace still creates an edge.

Moneyline

Moneyline (pick the winner)

when the spread is fair

A moneyline bet is simply on who wins, no spread involved. The model reaches for it when the spread looks efficient (no ATS edge) but it still likes one side to win the game.

It converts the projected margin into a win probability using a normal curve — that probability is the confidence you see.

Value ruleAny favorite priced −500 or worse is dropped as “No value bet” — you’d risk $500 to win $100, which isn’t worth it even when the team is a near-lock. Payable-but-heavy favorites are shown with a note that the payout is small.

Because heavy chalk gets filtered out, the moneyline column is usually quiet — only genuinely priced games show a play.

Market Soft

Saracen’s number is off the national market consensus (DraftKings, FanDuel, BetMGM, Caesars…). A small regional book moves slower, so a gap means Saracen is behind where the sharp market has settled. Flagged at a 1.0-point gap or more. Activates once the national-odds feed is connected.

Model Soft

The line is off what the game’s stats say — the “emotional, not statistical” play. It comes from the SP+ projection above; marquee-team, recency, and rivalry angles are noted as context but don’t move the number. Flagged when line and model differ by a wider margin (about 3.75 points right now, under the early-season fade).

Situational factors

Separate from the model. These are context and flags only — they never move the projected margin or change a pick. They sit under the game name as small badges, and in full under Situational when you open a game’s analysis.

Line movement — every time lines are pulled, the numbers are logged. The first pull of a slate is the open; the last one before kickoff is the close. The board shows how far the spread and total have travelled and which way. A line moving toward the favorite usually means money is landing there.

Look-ahead — flags a team whose next opponent is a big step up in class, the classic spot where a team plays down to a lesser opponent this week.

Being added as the season banks games: travel distance · injuries · referee crew and penalty rate · red zone performance · turnover margin · yards per pass and rush, and allowed. A note of caution on two of those: red zone rate and turnover margin swing wildly week to week and predict next week poorly, so they will stay context rather than model inputs unless the record says otherwise.

Reading the board

Rank — games are sorted by the two gaps added together (Market gap + Model gap). The bigger the combined disagreement, the higher a game sits.

Confidence bands color-code how strong a pick is:

62%+  high55–61%  mediumunder 55%  low

The Results Tracker logs every recommendation above 55% and grades it against the final score, so the model’s record is kept honest over time.

Research layer only — you place every bet yourself in the Saracen app. Bet only what you can afford to lose.