You Have Fifteen Rounds, Not Eighteen

I measured the distance between every pair of consecutive players on the Drafters board, looking for the tiers everyone talks about. The average gap between one ranked player and the next is 0.78 picks. That number is the baseline for everything that follows: when players are 0.78 apart, the board is telling you it can rank them. When they are five picks apart, it is telling you it cannot.

The board has three distinct stretches and one hard edge.

The top is smooth, and that is a finding

The largest gap anywhere in the first fifty picks is 2.6, Tim Stutzle to Clayton Keller at pick 47. That is 3.3 times the average, which sounds like something until you see what a real break looks like later.

There is no cliff in the first four rounds. The top of this board is a continuous slope, which means the tier you are waiting for in round two does not exist, and the player you "reach" for at 18 instead of 24 is six picks of gradual decline away from the alternative, not a category above it.

I went in expecting to find a wall near the end of the elite forwards and there isn't one. Whatever edge exists in this draft, it is not in the first fifty picks, and time spent agonising there is time spent on the flattest part of the board.

The middle is choppy

Between picks 53 and 108 the board breaks six separate times, with gaps of 3.2, 3.4, 3.4, 4.3, 4.3 and 5.2. In that stretch, being one pick late genuinely costs you a tier rather than a rounding error.

I checked whether those breaks line up with positions running out, because that would be a tidy story. They do not. The mix on either side of each wall is jumbled: forwards, defencemen and goalies in no pattern I can defend. The middle rounds are choppy, and I cannot tell you they are choppy for a reason.

The cliff is at 148

The break that matters sits between Jordan Kyrou at 146.9 and Oliver Bonk at 152.8: a gap of 5.9 picks, 7.6 times the board's average, and the largest single break anywhere on it.

It marks the edge of the board's knowledge. Past it, the ranking stops being a ranking:

Fifty-eight players (26% of everyone with a draft position) are stacked into the last sixteen picks of the board. They are an unsorted pile with decimal places attached, and the decimals come from averaging, not from a judgement anybody made.

You can see it in what happens to them. In a ten-team league there are 180 picks and 224 players carrying an ADP, so forty-four ranked players never get drafted at all. Every one of them sits between 165.7 and 175, inside that pile. The board is ranking players it does not expect anyone to take.

So count backwards from 148

Pick 148 is round 15. After it you have 32 picks, better than three full rounds, and every one of them draws from the heap.

The draft you are actually playing is fifteen rounds long. In any planning sense the last three are found money. You will take four players out of that pile and they will be roughly interchangeable with the four players you would have taken instead, because the board could not tell them apart and neither can you.

That changes what a mid-draft decision costs. When you are at pick 120 weighing a need against a player you like, the honest question is "can I get a defenceman in the next 28 picks," not "can I get one later." After those 28 picks of real board comes a lottery.

Plan your roster against 148 picks. Every slot you genuinely care about (your starters, the positions where a body is not interchangeable with any other body) gets bought inside those fifteen rounds. What you carry into rounds 16 through 18 should be nothing but upside swings you would be happy to cut.

The mistake this prevents is the common one: arriving at round 15 with a hole, and telling yourself there are three rounds left to fill it. There are not. There are three rounds left to gamble on it, which is a different sentence.

What would make me wrong

• If your league is not ten teams. 148 is a pick number, not a law. In a twelve-team draft the same cliff sits at a different pick and a different round. What transfers is the method: find the biggest gap on your board, measure it against the average gap, and treat that as your deadline.
• If you draft well out of the pile. I am claiming those last 58 players are interchangeable to the board. Someone with a genuine read on fourth-liners and rookie call-ups is not drawing at random there, and for them rounds 16 to 18 are real. I do not think most of us are that person, myself included.
• If ADP compression is an artefact rather than a signal. The pile-up may exist partly because ADP cannot exceed the length of the draft, which squashes everyone marginal into the same range. If so the players are more distinct than their numbers suggest, and the cliff is sharper on paper than in life. I think the 44 undrafted players argue against that, but it is the objection I would press if I were reading this.
• If your format rewards streaming or churn. All of this assumes a roster you draft and largely keep. If you are cycling players weekly, late picks are disposable by design and their quality matters much less than their turnover.

The short version

The board is smooth for fifty picks, choppy for the next sixty, and then it breaks at 148 and never recovers. A quarter of every ranked player is piled into the last sixteen picks, and forty-four of them go undrafted in a ten-team league.

So you do not have eighteen rounds. You have fifteen, and then three rounds of lottery tickets. Draft the roster you need inside the fifteen and take your swings after.