A ranking is a z-score in disguise

Two players sitting 20 places apart on a preseason ranking can be worth the same amount to your team. Two players sitting one place apart can be worth wildly different amounts. The number next to a name does not tell you which situation you are in, and that is the single most useful thing to understand about rankings before you use one.

The reason is in how the list gets built.

This piece is about the method rather than the players. If you came for the ranked list, it is published separately as the nine-cat top 50, computed from last season's averages using exactly the arithmetic described below. Reading this one first is what makes that one arguable rather than just a list.

What the number actually is

Almost every category ranking is a sum of z-scores.

A z-score answers one question per category: how far above or below the pool average is this player, measured in standard deviations. A standard deviation is just the typical spread of the pool. If the average drafted player gets 1.1 steals and the typical distance from that average is 0.4 steals, then a player at 1.5 steals is one standard deviation up, a z-score of +1.0.

Do that for all nine categories, add them up, and sort. That sum is the rank.

Written out, the compression is obvious:

Category	Player X	Player Y
Points	+1.2	+0.1
Rebounds	-0.4	+1.0
Assists	+0.9	-0.6
Steals	+0.3	+1.4
Blocks	-0.5	+1.6
Threes made	+1.1	-0.7
Field goal %	+0.2	+0.5
Free throw %	+0.4	-0.9
Turnovers	-0.8	+0.5
Nine-cat total	+2.4	+2.9

Player Y ranks higher. A list would show him a round earlier, and a manager reading only the rank would take him a round earlier.

Now drop turnovers, which is what an eight-cat league does. Player X goes to +3.2 and Player Y goes to +2.4. The order reverses, and neither player changed.

That is not a flaw in the method. It is the method telling you something the single number could not: these two are close in total and nothing alike in shape, and which one is better depends entirely on the league you are in.

The pool decides the average

A z-score is measured against a pool, and whoever built the list chose that pool.

Rank every player who appeared in an NBA game and the averages come out low, because the pool is full of players at 12 minutes a night. Every rotation regular then looks above average, and the list compresses at the top.

Rank only the top 150, roughly the players who will actually be rostered in a 12-team league, and the averages rise. Marginal players go negative, and the gaps between the genuinely good ones open up.

The second list is the one you want, and it is the harder one to find. When two published rankings disagree sharply about a mid-round player, the pool is usually the reason before anything about the player is.

The same logic applies to your own league. A 10-team league with eight bench spots has a shallower pool than a 14-team league with three, and the deeper league makes scarce categories more valuable, because there is less of everything to go around.

Percentages are the part rankings get wrong

Points, rebounds and steals are counting stats. You get what you accumulate, and a z-score handles them cleanly.

Field goal percentage and free throw percentage are not counting stats, and treating them as though they were is the most common error in a published ranking.

A player who shoots 90% from the line on two attempts a game and a player who shoots 90% on eight attempts do not do the same thing to your team. The second moves your team percentage four times as hard, because your team percentage is made of makes and attempts, not of other people's percentages.

A ranking that does this properly weights each percentage by volume, sometimes called impact or adjusted percentage. It asks how much the player moves the team number rather than how good his own number is.

You can check a list in about 30 seconds. Find a low-volume, high-percentage free throw shooter, usually a guard who rarely gets to the line, and see whether the list treats him as a free throw asset. If it does, the percentages are unweighted and the whole list is tilted.

Why specialists rank higher than they feel

Standard deviation is not the same across categories, and that is where most of the surprise in a ranking comes from.

Points have a large average and a large spread. Blocks have a small average and a small spread. A player who gets 1.8 blocks in a pool averaging 0.6 is a long way up in standard deviations, while a player who scores 22 in a pool averaging 16 is a much shorter distance up.

So the shot-blocking centre with an ordinary scoring line ranks above the 22-point guard, and the ranking is correct about that in a vacuum. Scarce categories are where a unit of production buys the most.

Two things follow, and they pull in opposite directions.

Specialists are undervalued by instinct and correctly valued by the list. People draft points because points are visible. The list is right and the room is wrong.

Specialists are overvalued by the list once several of them are on one roster. The list scores every player against the pool, independently. It has no idea you already have three shot blockers, and the fourth block per game you add to a category you are already winning 1-0 every week is worth nothing at all.

That second point is the boundary of what a ranking can do. It prices players. It cannot price your roster.

Rank is not draft position

A preseason rank is an estimate of value. Average draft position is an observation of what rooms actually do. They are different quantities and the gap between them is the only free money in a draft.

The gap comes from two places. Some of it is the room being wrong, which you exploit by taking the underpriced player. Some of it is uncertainty the rank does not show, and that one will cost you.

A ranking is a point estimate. A player ranked 18th with secure minutes and a player ranked 18th whose minutes depend on a training camp battle carry the same number and completely different risk. The second player belongs later, and the list cannot say so.

This connects directly to the method in the first column on this desk: the rank is the rate half of the projection wearing a suit. The minutes half is still your problem, and the rank quietly assumed an answer to it.

Draft tiers, not ranks

The last thing a ranking hides is how big the gaps are.

Ranks are ordinal. They tell you the order and nothing about the distance. Players 8 through 14 might be separated by a tenth of a z-score each, which is noise. Players 14 and 15 might be separated by a full point, which is a real drop.

So go through the list once and draw a line wherever the value falls off. What you get is six or eight tiers. Inside a tier, take the player whose categories your roster needs, or whose minutes you are most confident in. Between tiers, take the higher tier and do not think about it.

That one change fixes most draft-day mistakes, because it converts the question "who is the best player available" into "am I about to fall out of a tier", which is the question that actually has consequences.

What to do with a published list

Use it, and use it for what it is good at.

A category ranking is a genuinely good estimate of one thing: how much value a player produced per game against a pool, assuming an amount of playing time somebody else chose. That is worth having, and rebuilding it from scratch is not a good use of your week.

What it cannot do is know your league, your roster, your format or your tolerance for risk. Those are four inputs it never received, and every one of them moves a player more than the difference between rank 24 and rank 31.

The list ranks players. You are building a team, and a team is a set of categories you intend to win. Those are different problems, and the number next to the name was only ever an answer to the first one.