Vantage Points™

Last updated 13 October 2026

A team's VP estimates the points they personally add to an alliance score. Everyone starts at the same value at the beginning of a season, and the number moves after every match a team plays.

What it takes into account

Your partners. Qualification partners are assigned, not chosen. Scoring well with a weak partner is strong evidence about you; scoring well with a strong partner is less. An alliance-score average splits every score evenly and cannot tell the two apart — VP can.
Your opponents, and their defence. A score is judged against what that alliance should have managed against those opponents. Holding an opponent below what they should have scored counts in your favour, so defence is part of the number.
How strong the event was. A high-scoring event is partly strong teams and partly a strong field. VP accounts for the event's level so a team is not inflated just by where it played.
How much evidence there is. A new team's rating moves quickly; an established one moves slowly, so one bad match does not erase twenty good ones. Qualification matches count fully; elimination matches, played with a chosen partner, count for less.

What the board shows

The board ranks teams by how much of the field has actually tested them. A team that has only ever played one small region has not met the field a world board claims to rank it against, so its rating is shown with that caution built in. Early in a season everyone sits close together, which is the honest picture before teams have met each other.

Events that only played themselves — a long league drawn from a handful of schools whose teams play nowhere else — are left out. Without that, one team finished the 2025–26 season ranked first in the world on the strength of a two-month league against the same few teams, each lopsided win making the next look more impressive. Every championship is kept.

How it tested

Walk-forward across every V5RC match of the completed 2025–26 season: 107,194 matches across 12,377 teams from 1,513 events. Each match was predicted using only matches that started before it. Every model — ours and all three alternatives — was fitted on the first 60% of the season and graded on the last 40%, which it never saw.

Model	Brier	Partner luck	Finals won by top 20
Vantage Points	0.1559	4.6%	152
Elo	0.1586	14.7%	188
Average alliance score	0.1612	32.7%	139
TrueSkill	0.1632	14.1%	209

Brier: how good the win predictions were — 0 is perfect, 0.25 is a coin flip
Partner luck: how much of a team's ranking came from the partners it happened to be scheduled with — lower is fairer

VP predicts best and is the least affected by partner luck — and it does not have the best top twenty. TrueSkill's top twenty won 209 finals against our 152. That column comes from tournament results rather than from anybody's rating, which is exactly why it is here, and it does not flatter us.

How we know it is real

Defence is measured, not assumed. After both alliances' scoring is accounted for, the defending alliance's rating still explains how far a score fell short — and the same numbers shuffled between teams explain nothing.
The margin over Elo holds up. Resampled over events 2,000 times, the advantage stays above zero in 99% of them.
No knife-edge. Nearby settings perform about the same, so the result does not depend on one lucky choice.
No lookahead. No match is ever used to judge itself.

Predictions and projections

At a live event the app predicts each match still to be played and projects where every team will finish qualifying — a likely rank, the range it will probably land in, and its chance of a top-8 finish. Both update after every match, using that day’s results. Tested on Override matches the method never saw, with 95% intervals over events:

Measure	Result
Favourite won	68.5% [65.9, 71.2]; Elo on the same matches 67.1%
Brier score	0.202, against Elo’s 0.215; the 95% interval of the gap does not reach zero
“Strong favourite” won	81%
Top 8 named correctly, halfway through qualifying	67%, against 62% from the standings at that moment
“Chance of top 8” calibration	teams given 62% made it 60% of the time

A match is not predicted when a team in it has not played yet this season. Replayed match by match through a complete event, no prediction used a result from later in the event.

Method version

The method itself is not published. What is published is its identity: every board the app and this site show carries the version of the method that produced it and a SHA-256 fingerprint of that version’s exact code and settings. The method cannot change without the fingerprint changing, in public, and anyone later shown the method can confirm it is the one that produced these numbers.

Version: 2026.10b
Fingerprint: 2a0d85b041b02ef2aba3a75164c2e84d4e0c524d0eb709671445d8f8f2ca4a88

Change log

13 October 2026 · 2026.10b. The same model now also produces match predictions and rank projections. No change to how ratings are computed.
13 October 2026 · 2026.10. Version and fingerprint published. No change to how ratings are computed.
13 September 2026. Long events drawn from only a handful of schools are no longer counted, after one such league ranked a team first in the world.
30 August 2026. Corrected how an event’s scoring level enters the rating, and refit the season.
29 August 2026. Defence added: holding opponents below their expected score now counts.
27 August 2026. Vantage Points introduced.

What VP cannot see

The prediction claim rests on the completed season, not the current one. Override is young: across its first 5,885 matches VP and Elo predict equally well (about 65% of matches), a statistical tie. Prediction sharpens as teams play more. What the current season already shows is the fairness side.
Cross-region comparison. Teams compete mostly locally, and regions rarely play each other. VP accounts for event strength; it cannot make every region directly comparable.
One complete season. Every number above comes from 2025–26. A second finished season is the next real test.
Official standings. Those are WP / AP / SP. Nothing here changes them.

Vantage is not affiliated with, endorsed by, or sponsored by VEX Robotics, Innovation First, or the REC Foundation.