Everything here is counted above replacement, not in raw points. Raw points reward
quantity. Kyle Filipowski scored 2,035 in 25-26, which looks like a haul and is not: he played
at 27.2 a game against a replacement level of 24.4, so almost all of it was available off
waivers for nothing. His real worth was 188. Count raw points and you can win any trade by
sending one good player and taking back three warm bodies.
Replacement level is the twelve players ranked just past the last roster spot, recomputed
each season. A player is worth (his rate minus that) times the games he played, which is the
same measure the draft board uses.
Timing is normalised. A player picked up in week 3 has nineteen weeks to produce and
one picked up in week 17 has five, so totals quietly rank early trades above late ones whatever
either side received. Every figure is also shown per remaining week, which is the number to
compare across different points in a season.
Context is attached, because a trade is made by someone in a situation. Each side
carries his record going in, how many of his projected starters had collapsed in the fortnight
before, and what he did afterwards. A manager sitting at 7–4 with five starter-weeks of
nothing is not shopping for value, he is shopping for bodies, and that changes what winning
even means.
Every trade gets a verdict, and every manager a grade. The input is the gap between
the two sides per remaining week. Across all 55 of them that gap has a median of
16 and a 90th
percentile of 70, so the bands come from what this league actually does rather than from a
round number: under 12 is Fair, 12 to 29 a Slight edge, 30 to 59
Lopsided, 60 to 99 a Fleece, and 100 or more Veto-worthy. 0 of
those exist.
The letters, A+ down to F, run off the same gap, with one adjustment: a manager who went
into a deal short-handed and came out of it with a bracket place has his grade floored at C-.
Trading value for bodies when your starters are producing nothing is not a mistake, and a
grader that calls it one is measuring the wrong thing.
Every trade is graded twice, because hindsight is cheating. On the day prices
each player on what anyone could have worked out that morning: his rate so far that season,
pulled toward last season's rate by fifteen games of doubt, times the three games a week a
healthy player gets. No injuries, no drops, no idea what is coming. After is what
actually turned up. Where there is no previous season on record, which means 23-24, the prior
is the average rostered player, because that is all anyone could honestly have claimed to
know.
Only 0 of the 55 keep the same verdict, and 24 flip
which manager won. That is not the model contradicting itself, it is the shape of trading: a
deal both sides agree to is usually close on paper, and the season decides it afterwards.
Dennis sending Siakam, Reid, Wembanyama to Kevin for Jokić, Barrett read as a slight edge to Dennis on the day, 163 a week against 150, which is roughly what the league said out loud at the time. Over the rest of the regular season Kevin won it by 75 a week. Then in the playoffs, which that number does not touch, what Dennis had taken was worth another 628 to him against 304 — which is the trade he was making.
Value stops at the end of the regular season. Only six of twelve teams play the
three weeks after it, so counting them judges an eliminated manager against a game he was not
in. Both sides are measured over the weeks both men were still playing.
That would quietly punish the other kind of deal, though. A manager already safe in the
bracket can take the worse immediate return on purpose, for a player who wins him the thing
everyone is playing for. So the playoff weeks are counted separately, for the manager who
actually got there and only while he still held the players, and shown beside his grade rather
than inside it. A grade has to compare like with like. The playoff line is often the whole
reason the trade was made.
Then two numbers, because one on its own still lies. Value received is what
the players a manager got were worth for the rest of that season, whoever ended up holding
them: it prices the assets that moved. Banked is what they were worth while he actually
kept them.
The gap between those two is the story of an active trader. A player taken, held
a week and flipped on reads as nothing banked and was never a bust; he was a chip moving
through. Judge a churner on banked points alone and you punish him for the thing he is good
at.
Even with the context, none of this knows why. A manager buying a position
he is thin at, selling a season already gone, or clearing a roster spot is not playing for raw
value. This is the scoreboard, not the verdict.