A rating is just a claim about who is better than whom. The fair thing to do with a claim is check it against someone else's. So I lined our boys numbers up against two independent sources who owe this site nothing: a statewide community poll run by the people who actually watch these teams, and MaxPreps' own computer ranking, going back to 2012. Here is where we agree, and, more usefully, where we do not.
We are not out on our own. On the current boys ladder, our order and the community poll's line up about as closely as two rankings of the same 250 teams ever will. All three sources, our ratings, the poll, and MaxPreps, open with the same team at number one. The interesting part is not the agreement, it is the handful of names where the model and the room part ways, and why.
The best outside source is not another algorithm, it is a person. A Tennessee high-school soccer account publishes a full postseason ranking of the boys field every year, an eye-test list built by someone watching results all season and ordering the entire state, not just a top ten. For 2026 that poll ranked 339 teams. I matched 250 of them to a team on this site and asked the plain question: do our rating and their eye rank teams in the same order?
They do, closely. The rank correlation across all 250 shared teams is 0.94, where 1.0 would be perfect agreement and 0 would be a coin flip. A pure results-based rating and a season of human watching are, for practical purposes, telling the same story about who is good in Tennessee. That is the single most reassuring number on this page, because the two methods share no inputs at all.
Here is the current boys top of the table three ways: our rating, where MaxPreps has the team, and where the poll has it. Read across each row to see who agrees with whom.
Mostly a lot of near-matches and a few honest splits. Everyone has Ravenwood first. We and MaxPreps, the two algorithms, sit within a place or two of each other on most of the list. The poll wanders a little further on a few names, which is exactly what you would expect from an eye test, and points straight at the disagreements worth explaining.
One number needs an asterisk before it misleads anyone. The 0.94 with the poll is measured across the whole 250-team ladder, where the range runs from the best team in the state to teams barely hanging on. Agreeing on that broad shape is easy and genuinely meaningful, but it flatters everybody. MaxPreps only publishes its top 25, so the only fair comparison to it is within that elite band, and inside the top 25 nobody looks tidy, because those teams are all good and bunched together. Shuffling two nearly identical teams by a spot or two costs a lot of correlation when there are only 25 of them.
Measured that way, on the top 25 alone, the two algorithms actually agree with each other a hair more than the human poll agrees with either of us: our order versus MaxPreps' lands around 0.68, while the poll's own top-25 order versus ours is about 0.59. That is the honest shape of it. Two rating engines, built by different people from different data, converge slightly harder on the exact pecking order than the eye test does, and the eye test is better at the broad ladder. The metric that travels best is not the correlation at all, it is simple overlap: of MaxPreps' 25 teams, 22 are in our 25 too. Everyone agrees on who belongs. They quibble on the order.
The disagreements are not random, and they are the most useful thing here. Line our current boys order up against each source and a pattern falls out: the model is cooler than the consensus on traditional name-brand programs, and warmer on this season's schedule-dominators. Toggle between the two sources; the same teams keep showing up on the same side.
The left column is teams that won a lot of games this year against whoever was in front of them. The right column is the blue bloods, Farragut, Maryville, Baylor, Brentwood, Oak Ridge, programs a human ranks partly on what they have always been. The model has no memory of reputation. It only scores the season it was shown, so a down year for a famous program drops it further on our list than a human is comfortable dropping it. Neither view is wrong. They are answering slightly different questions, one about this spring, one about the program.
Worth saying plainly: this cuts against us sometimes. On South-Doyle we are the outlier, ranking the Cherokees near our top ten while both the poll and MaxPreps have them at the back of the top 25. When one source sits alone, it is usually the one to doubt, and here that source is us. The most likely reading is a strong record against a schedule the model rates more highly than the room does. It is on the list to watch, not to defend.
The cleanest split on the board is one this site has written about before. Both computers, ours and MaxPreps, rank Grace Christian (Knoxville) third in the state. The human poll puts it sixth, behind McCallie. In other words, the two algorithms agree with each other and disagree with the crowd, in the exact place you would predict: an unbeaten Division II-A team whose rating is built on dominating everyone it actually played, ranked ahead of a Division II-AA power that tested itself against a brutal schedule and took a few honest losses doing it.
That is not a bug in either machine. It is the undefeated-mid-major problem, and it is the whole subject of a separate report. A rating scores the games that happened; the eye test prices in the games that did not. When two independent algorithms land on the same "wrong" answer and the humans land on the other one, that is not an error to fix, it is a known and honest fault line between two ways of ranking. The Private Ceiling walks the whole case.
The poll only exists for 2026, but MaxPreps has published a Tennessee boys ranking for years, so the historic question is answerable: has our agreement with an outside consensus held up across seasons, or is 2026 a lucky match? I pulled MaxPreps' top 25 for six past seasons back to 2012 and ranked our teams as they stood at each of those season-ends.
It holds. In every season the overlap sits between 16 and 22 of 25, and the order correlation stays in the same band as 2026, no drift, no season where the two methods fall apart. (I left out the 2020 spring, when the boys season was cancelled and no full ranking exists.) Whatever the ratings are doing, they have been doing it consistently, and an independent ranking built by other people has agreed to about the same degree for over a decade.
The ratings are not eccentric. A season of human watching correlates with our order at 0.94 across the whole ladder; a rival algorithm shares 22 of its top 25 with ours, and has for twelve years. On the big question, who is good in Tennessee, there is broad, independent agreement.
The disagreements are structural, not sloppy. We are cooler on name brands and warmer on this year's winners because the model has no reputation and no memory, only the season it was shown. That is a feature you can decide you like or dislike; it is not noise.
And when we stand alone, we say so. Grace Christian is a place two computers out-vote the crowd for a reason worth reading. South-Doyle is a place the crowd is probably right and we are the odd one out. Both are on this page on purpose. A rating you only ever hear about when it agrees with you is not worth much.