{"schema_version":"public-reports.v1","canonical_url":"https://footballquants.com/sources/reports/kicker-defense-draft-value","slug":"kicker-defense-draft-value","title":"The pick-110 question: what a kicker and D/ST are actually worth","date":"2026-09-04","teaser":"Seven seasons of preseason forecasts and four seasons of real waiver wires point to different draft strategies for kicker and defense—but the same final-roster point accounting as every other position. Defense is mostly a late, streamable slot; elite kicker retains some preseason signal.","tags":["draft","kicker","defense","streaming","projections"],"body":"\nOur draft simulator recently put a kicker near pick 100 and a defense near pick 110. The recommendations looked strange for a simple reason: a mandatory starter can score 100-plus points without being worth anything close to a 100th-overall pick. The question is not how many points the slot scores. It is how much of that total this particular player adds over the version you can draft later or pick up for free.\n\nWe tested that distinction across seven completed NFL seasons, then replayed actual waiver wires in hundreds of real leagues. The short answer is that **kicker and defense call for different draft strategies, not different roster-grade arithmetic**. Preseason defense ranks contain too little stable separation to support an early pick, and a weekly projection can find more value on the wire. Kicker is noisier than its tidy draft board suggests too, but the very top of the preseason kicker board did retain a measurable edge. That evidence should influence which unit to draft, when to consider it, and how readily to stream—not invent a separate point formula for the position.\n\n## The experiment\n\nThe season-level test uses final-week-zero [FantasyPros projection data](https://www.fantasypros.com/api-data/) from 2019 through 2025 and realized full NFL regular-season totals. That means sixteen team games in 2019-2020 and seventeen across NFL Weeks 1-18 from 2021 onward. We compared QB, RB, WR, TE, K, and DEF inside realistic starter pools: twelve at each one-starter position and twenty-four at RB and WR. Kicker uses a flat three points per field goal plus one per extra point. Defense uses common sack, turnover, touchdown, safety, blocked-kick, and points-allowed scoring.\n\nThe second test reconstructs weekly ownership in 345 ordinary Sleeper redraft leagues from 2022 through 2025. Every policy sees only the players or defenses actually available in that league that week, keeps the manager's current unit as an option, follows that league's scoring, and stops before its fantasy playoffs. That produced 5,810 team-position seasons. Rules learned from older seasons were frozen before the 2025 reveal; the separate 2025 weekly-projection policy was fixed rather than tuned.\n\nThe real-league replay excludes best ball, superflex, and leagues that did not start the position, and stops at each league's own pre-playoff cutoff. It therefore does not use NFL Week 18; the season-level projection-accuracy test above does. The historical preseason and weekly archives were retrieved in 2026. We can identify their upstream week label, but we cannot independently prove the exact timestamp a manager would have seen on every draft night or waiver morning. That matters, so none of the timing results below is presented as a randomized causal estimate.\n\n## Small error is not the same as useful accuracy\n\nAfter removing each season's shared bias, the within-season standard deviation of actual-minus-projected performance was 81.1 points for quarterbacks, 67.6 for running backs, 50.9 for receivers, 35.3 for tight ends, 31.1 for kickers, and 23.0 for defenses. Read alone, that makes K and DEF look easier to forecast. They are not. Their projected differences are simply much smaller.\n\n| Position | Within-season residual SD | Rank correlation | Pairwise order | Noise / projected spread |\n| --- | ---: | ---: | ---: | ---: |\n| QB | 81.1 | 0.329 | 62.3% | 3.1x |\n| RB | 67.6 | 0.395 | 64.2% | 2.1x |\n| WR | 50.9 | 0.303 | 61.0% | 2.3x |\n| TE | 35.3 | 0.371 | 64.0% | 1.4x |\n| K | 31.1 | 0.275 | 59.9% | 6.4x |\n| DEF | 23.0 | 0.186 | 55.7% | 6.1x |\n\nRank correlation measures whether the preseason order survives. Pairwise order asks how often the higher-projected starter actually outscores the lower-projected one; 50% is a coin flip. The last column divides error by the within-season dispersion of the starter projections. It is the most revealing column. Kicker and defense errors are six times the amount of separation the projection board asks us to draft.\n\n![Projection noise relative to the draftable spread](/reports/special-projection-noise.svg)\n\nAcross the twelve projected starters, the average top-to-bottom span was only 14.2 points at kicker and 12.0 at defense. The actual miss around those numbers was roughly twice the advertised separation. Only half of the projected top-twelve kickers finished in the top twelve; defense retained 51%. A preseason DEF ordering was right in a head-to-head comparison just 55.7% of the time.\n\nThat does not make projections useless. It changes what they can support. A small edge is reasonable as a tiebreaker. Counting every projected point as if it were scarce roster value is not.\n\nKicker also has a useful two-part mechanism. Conditional on a kicker recording an active week, the projected scoring-rate error averaged only 0.84 points per game. A descriptive decomposition of the full-season miss assigns about 42% of residual variance to observed-week and job availability, about 50% to scoring-rate error, and the remainder to their covariance. That is not a causal attribution, but it explains the apparent contradiction: team scoring opportunity can be forecastable while the season-long identity of the kicker receiving every attempt remains fragile.\n\n## The attainable prize, not the hindsight winner\n\nIt is easy to prove that K1 and DEF1 matter after the season: sort by actual points and circle the winner. Nobody gets to draft that way. Our identifiable test instead takes the unit ranked first in August and compares its eventual score with the unit ranked thirteenth in August, the first nominal replacement in a twelve-team league.\n\n![Actual surplus of preseason rank one over preseason rank thirteen](/reports/special-attainable-edge.svg)\n\nThe projected number-one kicker beat the projected number thirteen by 40.3 actual points on average. With only seven independent seasons, the conservative season-level 95% interval is wide, from 1.6 to 78.9, but it remains above zero. The first three projected kickers also beat the rest of the projected starter tier by 9.6 points on average, although that interval crosses zero. Across the broader tiers, projected K1-K12 beat K13-K24 by 12.5 points, with a season-level interval from 4.0 to 20.9.\n\nDefense did not clear the same bar. Projected DEF1 beat projected DEF13 by 18.7 points on average, but the season-level interval runs from -38.8 to 76.2. The top-twelve tier beat the next twelve by 12.2 points, with an interval of -0.1 to 24.4. There may be a small tier edge, but these seasons do not show that August can reliably identify one premium defense.\n\nThis is the main split: **an elite kicker can carry identifiable preseason value; an elite defense has mostly been a hindsight label.** Even for kicker, the uncertainty is large enough that one great season should not dictate a permanent draft rule.\n\n## Streaming works, but only with a real forecast\n\nFor 2025 we replayed four recognizable choices against each manager's reconstructed waiver wire. Draft-and-hold keeps the originally drafted unit. Real managers follow the transactions they actually made. Trailing form always selects the available unit with the best prior four-week average. Weekly projection selects the best available upstream projection before that week's games.\n\n| Position | Draft and hold | Real managers | Trailing form | Weekly projection |\n| --- | ---: | ---: | ---: | ---: |\n| K | 102.8 | 115.0 | 115.3 | 118.3 |\n| DEF | 91.8 | 106.9 | 93.4 | 129.1 |\n\n![Streaming policies on the 2025 real waiver wire](/reports/special-streaming-2025.svg)\n\nAt kicker, the weekly-projection policy added 15.5 pre-playoff points over holding while making 8.4 changes. Charge an intentionally steep one fantasy point for every move and 7.1 points remain. At defense it added 37.3 points over hold with 10.9 changes, or 26.4 after the same charge. The 95% league-bootstrap intervals for the uncharged gains were 13.0 to 18.0 for K and 34.9 to 40.1 for DEF.\n\nThe negative result is just as useful: chasing the prior four weeks was not a good defense strategy. It added only 1.6 points before 10.5 transactions and lost 8.9 after the charge. Streaming is not synonymous with cycling through whatever scored last Sunday. It worked when the selection rule described the coming week.\n\nReal managers corroborate that pattern, with the usual selection-bias warning. After controlling for league and drafted market rank, managers who changed defense one to three times scored 16.6 more points than holders; frequent streamers scored 21.2 more. At kicker, occasional changers were ahead by 6.6, while the estimate for frequent changers was -1.0 and its interval crossed zero. Better managers may both work waivers and make better decisions elsewhere, so those are associations, not proof that a transaction caused the gain.\n\n## What about pick 110?\n\nThe 2025 timing result is more interesting than a slogan like “always draft special teams last.” In ten-team standard rooms, the median kicker and defense were each drafted in round 14. Pick 100-110 is roughly rounds 10-11, the early tail of the market.\n\nUnder the same weekly-projection streaming policy and the same reconstructed wire, teams that had drafted a kicker by pick 130 scored 25.8 more K points than teams that waited past 150 within the same league. Defense went the other direction: early drafters scored 7.3 fewer. That comparison is observational and does not charge the early drafter for passing on a skill player.\n\nWe added a fixed-board sensitivity for that cost. Among the 74 teams that selected K by pick 110, taking the kicker then rather than waiting one owner pick produced a 37.8-point special-teams upgrade in the 2025 outcome and gave up 4.6 points of realized skill-position value, a net of 33.3. Expand the cutoff to pick 130 and the average net shrinks to 3.5 points, with an interval from -5.4 to 13.1. For defense by pick 110, the comparable next-pick net was 0.7 points, with an interval from -9.4 to 10.4; by pick 130 it was -5.1.\n\nThe sharp kicker cutoff is a warning against turning one season into a universal round target. The same NFL player outcome repeats across many fantasy teams, the conditional league bootstrap does not represent a new season, and a fixed draft board cannot model all the roster options a live rollout sees. The defensible reading is narrower: 2025 rewarded the managers who secured that year's truly elite kicker, while providing no comparable evidence for early defense. It does not prove that K near 100 will pay again in 2026.\n\n## Why this overturns our earlier internal test\n\nOur earlier special-teams check looked encouraging: the best actual defense finished about 60 points above the thirteenth defense, and the best actual kicker about 41 above the thirteenth kicker. It then related those outcomes to market rank. That measured the prize visible after the games had been played, not whether an August projection could identify the winner. The hindsight-best defense is real fantasy value, but it is not a draftable asset unless a preseason signal can find it.\n\nA second validation replayed roster grades across 503 historical leagues and swept different K/DEF weights. We found a silent data gap during this audit: that canonical history had no historical kicker or defense projection rows because its projection table contained only the current season. Changing the factors therefore changed none of those old grades. The sweep was inert, so it could neither validate nor reject the live policy.\n\nThis study fixes both problems. It recovers official historical K and DEF projections, scores the units one could actually have selected by preseason rank, reconstructs the real weekly wire, and reruns special-position grade alternatives with genuine historical inputs. The prior result is not supporting replication; it is a useful example of how hindsight value and missing rows can make a draft rule look more certain than it is.\n\n## What a draft grade should count\n\nA completed roster grade should be literal: every selected starter contributes its projected season points, whether the label is QB, RB, WR, TE, K, or D/ST. An empty slot contributes zero. The rollout's job is to make the remaining picks before it compares final grades, so passing on a kicker in round ten does not mean finishing the draft without one.\n\nThe zero-to-full-season jump was a real *partial-search* problem, but our first correction put the answer in the wrong place. On August 31 we gave only an empty K/DST slot a virtual streaming baseline and graded a selected unit as baseline plus projected surplus. That stopped the search artifact, but it also made those positions obey a different terminal rule, floored a below-replacement selection, and forced the UI to subtract hidden points from in-progress scores.\n\nWe removed that special formula on September 4. The terminal grade is raw projected points again. A separate hidden completion estimate is now used only while ordering an unfinished search: it fills every open starter slot—not just K and D/ST—with the league's replacement projection. Once the lineup is full, the estimate and raw grade are the same. It never appears in a roster card or final score.\n\nThat architecture keeps two questions separate. Final-roster accounting asks how many points the selected lineup projects to score. Pick selection asks what this player adds relative to the alternatives likely to remain, including the skill player forgone now and the K/DST tier available later. For defense, the rank test and real-wire replay still argue for drafting late and being willing to stream from Week 1. For kicker, zeroing the position would discard positive top-of-board evidence. Neither conclusion requires phantom points or a K/DST-only grading rule.\n\nWe did test numeric shrinkage with the same historical gate used for other production coefficients. Both defense-projection proxies selected 25% of K surplus and 0% of DEF surplus on the 165 development leagues from 2022-24. On the 81 untouched 2025 leagues, that challenger reduced mean within-league rank correlation by 0.0066 to 0.0072 and failed a major format-family guard under each proxy. It was rejected, and no K/DST-specific surplus coefficient is served.\n\nAfter the uniform correction and a fresh projection pull, the exact historical CLBH slot-10/pick-110 rollout completed in 7.53 seconds. K, D/ST, QB, TE, WR, and RB candidates all landed within 2.8 projected grade points, with overlapping rollout uncertainty. That near tie is plausible; a 100-point phantom K/DST bonus is not. The important check is that every terminal score came from an actually completed roster and every partial slot used the same search rule.\n\nPractically: a defense recommendation around pick 110 is not supported by this study. A kicker in that neighborhood is not automatically wrong if the candidate is the clear projection leader and the skill-player alternatives are flat, but it demands a much stronger case than “the K slot is empty.” Most K and DEF selections still belong in the final few rounds.\n\n## How this compares with prior work\n\nThe direction agrees with a [2022 study in Judgment and Decision Making](https://www.cambridge.org/core/journals/judgment-and-decision-making/article/drafting-strategies-in-fantasy-football-a-study-of-competitive-sequential-human-decision-making/2AB841B3F446833348D784C0FC54DAD2), which found shallow marginal point gaps at K and defense in 1,350 Sleeper leagues and recommended deprioritizing both, while finding more useful calibration for kicker than defense. Older [Becker and Sun prediction-error research](https://doi.org/10.1515/jqas-2013-0009) likewise reported smaller raw errors for K and DEF, but raw scale alone does not answer whether one draftable option can be distinguished from another.\n\nPublic strategy research has reached similar conclusions from different data. [PFF's kicker study](https://www.pff.com/news/fantasy-football-metrics-that-matter-kickers) found little relationship between kicker ADP and finish and pointed streamers toward favored teams with high implied totals. [Footballguys' multi-year Rent-a-Kicker results](https://www.footballguys.com/article/2025-rent-a-kicker-week-01) show that a weekly process can produce usable scoring while remaining noisy year to year. Our contribution is to make availability league-specific, score every league under its actual rules, separate an identifiable preseason edge from a hindsight winner, and place season-level uncertainty ahead of thousands of repeated fantasy-team observations.\n\n## Boundary of the result\n\nThis report applies to ordinary one-K, one-DEF redraft leagues. Distance bonuses, field-goal miss penalties, return-yardage scoring, multiple special-teams starters, best ball, unusually shallow wires, locked waivers, or a league that awards unusually large defense points can change the economics. The projection archive is late-preseason rather than a snapshot from each real draft date, and the 2025 weekly archive is treated as pregame-frozen but was not independently timestamp-verified. Closing betting lines were useful in exploratory defense rules but are not presented as a deployable waiver-time result.\n\nThe historical roster-grade gate also inherits nonrandom gaps in its skill-position projection base: top-150 coverage was about 61-62% in 2022, 74-75% in 2023, and 88% in 2024, while the retained 2025 snapshot was complete. The paired K/DEF policy comparisons are less exposed than absolute grade correlations because every arm shares that same skill base, but development selection is not immune to the limitation.\n\nMost importantly, seven NFL seasons are seven independent seasons, not thousands. The wide defense interval is not a technicality to average away. It is the result: the apparent precision on a preseason D/ST board is much greater than the evidence underneath it.\n","limitations":["This is a dated research snapshot, not current player advice. Conclusions apply to the populations and test periods described."]}