Where the Tiers Break

A Tier Is a Gap You Can Measure

A tier break is a distance on the board, measured in projected points like everything else the model publishes. Sort a position by projected median, subtract each player's median from the one directly above him, and the column of differences is your tier map. Most of those differences are small and close to one another. A break is a difference several times larger than the ones on either side of it.

Hand-drawn tiers come out tidy — five names, five names, six names — because tidy is how people group things. Projection spacing is lumpy. A position can run eleven deep with no gap worth a line, while the position next to it stacks two walls inside six picks.

The practical difference shows up at your pick. A tier drawn by feel tells you a player belongs with a group; a break measured in points tells you what the next name down actually costs. That cost is the number you trade against when you decide whether to reach.

Two Measurements Decide Whether a Gap Is a Break

Local spacing is the first measurement. Take the median adjacent gap across the ten or so players surrounding the candidate break; that median is the normal distance between neighbors in that neighborhood of the board. A gap running two and a half to three times that local median is worth a line. That multiple is a judgment call, not a backtested cutoff; it marks the point where a gap stops reading as ordinary spacing.

Band width is the second, and it is the measurement that keeps the method honest. Every projection ships with a P10 to P90 range, and season-long ranges run wide enough that the two players straddling almost any break overlap across most of their distributions. How those bands get built has its own page in the model methodology.

The honest reading of a break is distributional. Across the whole set of picks you make, taking the higher side of the lines wins more often than it loses, while any single pair straddling a line is close to a coin flip. A tier map is a rule for spending many picks, and it pays at that scale.

Watching the Arithmetic Happen

Take an invented stretch of receiver board — numbers made up to show the shape, with no relationship to any live run. Six consecutive projected medians come off it at 191.4, 189.0, 186.8, 184.9, 176.2, and 174.4.

The adjacent gaps are 2.4, 2.2, 1.9, 8.7, and 1.8. Median local spacing is 2.2, which puts the 8.7 a shade under four times normal while every other gap sits within half a point of the median. The break falls under the fourth name, and nothing about those six players' reputations entered the calculation.

Band width sets how much to make of it. Give the fourth name an invented P10 to P90 of 149 to 221 and the fifth one of 142 to 212, and the 8.7-point gap is roughly an eighth of the width of either band. One pick across that line is a coin flip with a thumb on it. Forty picks taken on the right side of forty such lines is a different proposition, and that is the question a tier map answers.

The count is what turns the break into a decision. A third-slot manager in a twelve-team room picks at 27 and again at 46, so eighteen selections run between the turns. Say two of the four names above the break are already gone at pick 27. Give the invented room a draft rhythm to match — one receiver leaving every five picks through this stretch, made up like the medians — and the window prices at three or four more of them.

Two survivors against three or four expected selections: the tier does not reach the next turn. Take one of the two now, or write the tier off and plan to pick below the line where the run is flat. Change one input, five survivors in place of two, and the same arithmetic frees the current pick entirely, because the line still stands eighteen picks later.

Breaks Are Where Runs Start

The room is reading roughly the same cliffs you are, off consensus ranks and off ADP. That shared read is why position runs bunch immediately above a break: put the invented room on the clock, and half its managers are staring at the same two survivors above the 8.7-point line — a tier that took two rounds to build clears in a handful of consecutive picks. Your survivor count is a forecast of that moment.

The cost of being early and the cost of being late are not the same size. One pick early on the last name in a tier costs the local spacing, 2.2 points of median in the invented stretch above. One pick late costs the whole break, 8.7. That asymmetry is the argument for taking the last member of a tier a few picks before it feels comfortable.

Below the break the logic inverts. A flat run means the second and the sixth name in it sit inside each other's noise, so the manager who waits collects the same projection at a later price. That machinery sits underneath most of what surfaces as ADP values and steals. Waiting through a flat stretch is close to free. Waiting through a break is what managers mean when they say the board fell apart on them.

What a break is worth also depends on where the position sits against replacement level, since a nine-point gap at a position you start three of moves more roster points than the same gap where you start one, and value-based drafting carries that conversion in full. The widest breaks tend to sit near the top of quarterback and tight end, where the pool flattens hard a few names down and stays flat for rounds. The case for letting both positions wait is made on the late-round QB and TE page.

Using Breaks Without Trusting Them Too Far

A break inherits every error in the medians underneath it. A player misplaced by twelve points drags the line with him, either manufacturing a gap that was never there or filling one that was. Tier lines carry the model's ranking error in full, at the exact spot where that error costs a pick the most.

Deep breaks are mostly artifacts. Past the depth where projected medians converge toward replacement level, adjacent spacing shrinks toward zero and a multiple-of-local-spacing test starts firing on rounding. Run the test through the range where medians still sit meaningfully above replacement, and read the tail as one flat block.

The lines move under you as well. The model reruns on demand and ADP updates daily, so a tier map screenshotted in August describes August; count the breaks on the live PPR draft board at the hour you draft, with the picks that have already happened removed.

Tier breaks are the timing layer over a build you have already chosen. The board says where the cliffs land; whether a cliff matters is your build's call, and the count says whether it survives to your next turn. Walk in with the two or three breaks nearest your turns already measured, recount them once real picks start coming off, and let the flat stretches wait. Every call this model makes is graded on the public accuracy page, hits and misses in the same table; read that record, then decide how hard to lean on any of this, including this piece.