In-Season Projection Adjustments: How to Update Models Mid-Year

A preseason projection is a structured guess about a player's output over a full season, built from historical rates, depth charts, and educated inference. Once the season starts, real data replaces inference — and models that fail to incorporate it become less useful with every passing week. This page explains how in-season projection adjustments work mechanically, what triggers them, where projection systems disagree on approach, and what the common failure modes look like for both models and the analysis who use them.

• Definition and scope
• Core mechanics or structure
• Causal relationships or drivers
• Classification boundaries
• Tradeoffs and tensions
• Common misconceptions
• Checklist or steps (non-advisory framing)
• Reference table or matrix

Definition and scope

In-season projection adjustment is the process of revising a player's expected statistical output — for a single game, a remaining-season total, or a rest-of-season arc — based on information that was unavailable or unreliable at the time of the original projection. The adjustment is not a correction of error in the original model. It is a response to new data entering the system.

The scope covers two distinct timelines. Single-game adjustments incorporate information available in the days or hours before kickoff: injury reports, weather, confirmed inactives, and line movement. Rest-of-season adjustments incorporate slower-moving structural signals: usage-rate changes, role evolution following a teammate's injury, demonstrated performance trends, and changes in a team's offensive identity. These two adjustment types require different inputs and operate on different update cadences. Conflating them is a reliable way to either under-react to short-term volatility or over-react to noise that hasn't had time to stabilize.

The projection update schedule used by any given system determines how often these revisions are published, which in turn defines how stale any given projection might be on a given day.

Core mechanics or structure

Most projection systems begin with a preseason baseline — a rate-stat forecast grounded in multiple seasons of historical data, typically 3 to 5 years of weighted performance. In-season adjustment layers sit on top of that baseline, either modifying the underlying rate assumptions or adjusting the opportunity inputs that drive volume estimates.

The two principal mechanical levers are rate adjustments and volume adjustments.

A rate adjustment changes the expected efficiency of a player's output per opportunity. If a wide receiver is averaging 12.4 yards per reception through six weeks after a career rate closer to 9.8, a model must decide how much of that gap reflects real improvement versus sample-size variance. The regression to mean in fantasy framework handles this decision explicitly, typically applying weighted regression that shrinks toward the player's long-run rate as a function of sample size — with smaller samples receiving heavier regression.

A volume adjustment changes the projected number of opportunities: targets, carries, snaps, or plate appearances. These are often more structurally durable signals than rate changes. A running back who has absorbed 22 carries per game across 4 weeks following a starter's season-ending injury has a new role, not a hot streak. Usage rate adjustments in projections formalize this distinction, separating opportunity structure changes from efficiency fluctuations.

Mechanically, the update sequence runs: new data ingested → prior estimate updated → posterior estimate generated → game-environment adjustments applied → final projection published. Systems that skip the posterior step — going directly from raw observed performance to a new projected rate — tend to overcorrect.

Causal relationships or drivers

Four categories of events consistently drive meaningful in-season adjustments.

Injury and roster changes produce the largest single-week projection shifts. When a No. 1 receiver misses 4 weeks, the targets that were going to him don't disappear — they redistribute across the remaining receivers and tight ends. Quantifying that redistribution requires knowing each remaining player's target share under comparable historical conditions, which is why snap count and target share data is treated as a primary input rather than supplementary context.

Usage-rate evolution is the slower-moving driver. Offensive coordinators adjust schemes over a season. A rookie running back who entered the season projected for 8 carries per game may be at 16 by Week 8 as the team's offensive line stabilizes. The projection model must detect whether this is a new equilibrium or a peak that precedes regression.

Matchup shifts operate at the weekly level. A quarterback facing a defense that ranks 32nd in passer rating allowed gets an upward single-game adjustment; that same adjustment has no bearing on his rest-of-season projection unless the schedule clusters those opponents. Matchup-based projection adjustments document how these opponent quality factors are typically weighted.

Vegas line movement functions as an aggregated signal from a market incorporating information at scale. A team's implied team total — derivable from the spread and over/under — correlates meaningfully with projected fantasy scoring for that team's skill players, particularly quarterbacks and pass-catchers. Vegas lines and fantasy projections covers this relationship in detail.

Classification boundaries

Not all projection revisions are the same type of adjustment, and treating them as equivalent produces muddled analysis.

Structural adjustments reflect durable changes in a player's role, team context, or physical condition. These carry forward into rest-of-season projections and affect dynasty valuations. A receiver's target share climbing from 18% to 28% following a teammate's ACL tear is a structural shift — one that warrants a significant model update even if the underlying rate stats haven't yet confirmed the new volume.

Situational adjustments are game-week-specific and decay after the relevant game. A weather downgrade, a one-week suspension for an opposing cornerback, or a narrow injury designation that resolves by Sunday all fall here. Applying these situational adjustments to a rest-of-season model is a category error.

Statistical corrections occur when observed performance reveals that a baseline assumption was wrong — not that the player changed, but that the preseason estimate was miscalibrated. A pitcher who was projected at a 3.40 ERA based on 2022-2023 data but who has demonstrated consistent command improvements visible in StatCast spin metrics (Baseball Savant, MLB.com) may warrant a structural ERA revision rather than mean regression.

Tradeoffs and tensions

The central tension in in-season adjustment is between responsiveness and stability. A model that updates aggressively on new data catches real trend changes early but generates volatile, noisy projections that oscillate week to week. A model that updates conservatively remains stable but lags behind genuine role changes and can misprice emerging players for 3 to 4 weeks.

There is no setting that eliminates this tradeoff. Different users have different tolerances for it. A daily fantasy player optimizing a single slate needs the most aggressive possible incorporation of game-week information. A season-long manager making a waiver wire decision at Week 6 about a player's rest-of-season value needs a model that has allowed enough weeks of data to distinguish signal from noise. The sample size and projection reliability framework addresses exactly this problem — establishing minimum observation thresholds before rate-stat revisions are treated as credible.

A second tension exists between individual-level and team-level adjustments. When a team's total offensive output rises, the projection bump needs to be allocated across skill positions. That allocation is itself uncertain, and naive approaches that simply apply a proportional lift to all skill players miss the asymmetric way that offensive volume concentrates under real play-calling conditions.

Common misconceptions

Misconception 1: A player's strong early performance justifies a proportional projection increase.
Three games is not a sample. The projection confidence intervals literature is consistent here — rate statistics require 6 to 8 games at minimum before a shift from a player's established baseline reaches statistical significance. A receiver averaging 18 PPR points over 3 weeks is more likely to regress toward his 12-point baseline than to sustain the elevated rate, absent a structural explanation.

Misconception 2: Injury adjustments apply immediately and completely.
When a starter is ruled out, backups rarely absorb 100% of the departed player's opportunity share. Historical distributions from Pro Football Reference data show that opportunity absorption across positions typically ranges from 60% to 80% for the primary backup, with the remainder scattered across the remaining roster. Models that project a handcuff to fully inherit a starter's workload are overstating the adjustment.

Misconception 3: In-season adjustments are only relevant for weekly lineup decisions.
Rest-of-season projections for dynasty leagues, keeper decisions, and trade valuations are all products of in-season adjustment. Rest-of-season projections depend on updated baselines, not preseason ones, and the gap between an unadjusted preseason model and a properly updated model can represent 30% to 40% of projected scoring for players whose roles have shifted significantly.

Misconception 4: Weather adjustments matter for all positions equally.
Wind speed above 15 mph materially affects kicker and quarterback projections; its effect on running back projections is far smaller and in some precipitation scenarios slightly positive, as teams shift toward a ground-oriented game script. Weather impact on fantasy projections maps these relationships by position.

Checklist or steps (non-advisory framing)

The following sequence represents the standard update process for a weekly in-season projection adjustment cycle.

1. Injury report integration — Wednesday through Friday practice designations are ingested; players verified as Doubtful or Out trigger downstream opportunity redistribution calculations.
2. Usage-rate audit — Snap counts and target/carry shares from the previous week are compared against the rolling 4-week average and the preseason baseline; deviations exceeding 15 percentage points flag for structural review.
3. Vegas line incorporation — Thursday and early-Sunday lines are pulled; implied team totals are calculated and compared against the projections' embedded game-script assumptions.
4. Matchup overlay — Opponent defensive rankings by position are applied as a multiplier against the updated individual baseline.
5. Weather check — Game-time forecasts for outdoor stadiums are reviewed; wind and precipitation adjustments are applied to the relevant position groups.
6. Regression audit — Players whose rate-stat projections have drifted more than 20% from their established baseline receive a regression weight applied proportional to sample size.
7. Final output review — Projected totals are cross-checked against team-level implied scoring caps; individual projections that sum to more than 120% of the team's implied total are flagged for reallocation.

Reference table or matrix

The table below classifies common in-season triggers by adjustment type, time horizon, and position sensitivity.

For a broader orientation on how individual adjustment layers fit into the projection ecosystem as a whole, the Fantasy Projection Lab home provides context across all major sports and modeling approaches.