Guides

How a trade gets valued, and why the calculator disagrees with your gut

Here is a trade I would have accepted without reading anything.

I send Puka Nacua. I get back James Cook, Nico Collins and Kyren Williams. In a 12-team PPR league this site projects those three at 269.05, 248.82 and 237.94 points for the season, which is 755.81 against Nacua's 347.76. More than twice as much.

The Trade Analyzer put my side at 146.46 and the three-player side at 116.88, and then refused to name a winner.

Three ways of counting one trade. The first is a landslide, the second is close, the third goes the other way. Which of them is the question you were actually asking? It took me longer than I would like to work that out.

One thing none of them is: a price. There is no market anywhere in this number. The site is not reading what other managers paid for these players, it is reading four sets of season projections and solving one league's starting lineups. Its own methodology opens by saying the Trade Analyzer "does not use market prices". If somebody in your league quotes you a value off a trade chart, the two of you are not disagreeing about a player. You are holding up two different kinds of object.

Patrick Mahomes is worth 0.7 points

Patrick Mahomes projects 287.83 points in that league. Justin Herbert projects 287.13, and Herbert is the thirteenth quarterback on a board with twelve starting spots, which makes him roughly what you can have for nothing.

So Mahomes is worth 0.70 points. The site prints his value index as 0.4 out of 100 and I still do a double take at it every time.

Josh Allen projects 362.68 and scores 75.55 on the same measure. He is 74.85 points better than Mahomes in the projection and 74.85 better in the value, because both are being measured off the same floor. Move the floor and everybody moves. The gaps stay where they were.

That is value above replacement, and it is the only idea here worth carrying around. A player is worth what he gives you over the man you would otherwise have started. Not what he scores.

The second player is close to free

Back to my trade. The analyzer is not comparing two piles of points. It builds one reference team, the median starting roster in that league, and asks what each package adds to it. For this trade that team's flex seat held Christian Watson on 201.30 and its second receiver was Tee Higgins on 220.

Cook, Collins and Williams all want a seat on that team and there are not three good ones going. The response says three incumbents lost their place. So the third arrival is not worth his 237.94 points, and he is not worth the 47.91 he carries above replacement on his own either. He is worth whatever he beats, and by the time he gets there the two ahead of him have taken the seats worth taking.

Which makes the middle row of my table a cheat. Adding 79.02, 58.80 and 47.91 to reach 185.73 counts the same empty roster spot three times over.

Two for one is not a discount. It is a queue.

So why will it not pick a side?

Because the four independent sources behind those projections disagree, and the tool runs the comparison again at the edges of that disagreement: my side at the sources' low end against theirs at the high end, then the other way round. The central answer came out 29.58 points my way. The full range ran from 36.06 against me to 92.93 for me. It crosses zero, so the verdict was close and the stated reason was that "the sources do not separate these packages".

I think that is the right behaviour and it irritates me every single time.

There is a stricter version of the same refusal. 186 of the 591 players in tonight's pool are carried by a single source, which is not enough to bound anything. I put Jahmyr Gibbs against AJ Dillon deliberately to see it happen. The tool had a perfectly good number for my side, 156.18, and published the verdict as limited anyway: "1 player(s) are carried by fewer than 2 independent sources, so this comparison cannot be bounded". It would rather say nothing.

What it does not know about you

It does not know your roster. Every figure above was measured against a median team the model built, not against yours, and it assumes you do not already own anybody in the trade. If Kyren Williams is on your bench already, the seat he is competing for is one you have filled, and the 47.91 stops being about your team. There is no way for the tool to find that out.

It does not know you are 1-4 and that a fair trade is no use to you. It does not know your bench is one hamstring away from unplayable. The methodology page says it outright: this is "an estimate about a reference roster, not a prediction that any particular manager would accept the trade".

So here is how I use it, and I am not certain I am right. When the verdict is lopsided I go and find the thing I had not priced, and the row I read is the second one, not the first. When it comes back close I stop asking it, because it has genuinely not decided and nor has anyone else.

All of this was pulled on 16 August 2026 and the projections move most nights, so your numbers will not be my numbers. The shape of it will be.

← All guides