Valve shipped Gameplay Patch 7.41e on 30 July with Summer Scrub 2026 riding alongside it, two weeks before The International. Snapfire, Drow Ranger, Bane, Clockwerk and Treant Protector came down, among others, with a buff list alongside. Divine Rapier’s spell amplification no longer stacks. Controlled units got their own hotkeys. Immortal MMR was squished. The full list is on the patch page; this post is about what a patch that size does to the numbers you use to judge yourself, and why Dota Companion 1.9.0 was built for exactly this moment.
The population average moved. Yours did not.
Every site that shows a hero win rate is showing a population number, and for the next several weeks that number is a blend of two metas: games from before the patch, where Snapfire was one hero, and games after, where she is a different one. The global win rate on any touched hero is currently measuring a hero that no longer exists, averaged with one that does, weighted by whatever ratio of old and new games happens to be in the window. Filtering a stat site to the new patch does not fix it; for the first weeks that filter is a sample that is either thin or dominated by whoever queued most on day one, and neither of those is your bracket.
Your own record has a different problem. It is small. It is also honest, and it is the only baseline that was actually generated by you, on your heroes, in your bracket. A drop from your own 58% on a hero to 49% after the patch is information. A global 51% is noise wearing a decimal point.
What Dota Companion does with your baseline
Dota Companion compares each hero against your personal baseline, not against 50%. 1.9.0 gives that idea four surfaces, and a patch is where all four earn their keep.
- Insights. Finds the single metric that has slid furthest below your own baseline and says one thing to work on. After a patch, that is the fastest way to find out which of your heroes the patch actually hit, as opposed to which ones the notes say it hit.
- Per-hero progress. Your record on each hero against your own history on that hero. Post-patch, it is the only chart where the before and the after belong to the same player.
- Role-split baselines. Safelane, mid, offlane and roaming each get their own baseline, because the Treant change alters your offlane games and far less to your mid ones. One blended number would hide that.
- Draft helper. Ranks your own hero pool against the enemy picks as they happen, ordered by your record with a confidence lower bound doing the sorting so that a three-game 100% hero does not outrank a forty-game 65% one, and shows how you actually do when a given enemy hero is on the other team. It does not claim to know the new meta. It knows your games.
Before any of that renders, a line at the top of the Overview says what the data is: whether the profile is private, whether the rank is uncalibrated, whether the sample is too thin for a baseline to mean anything, and how old the cached copy is. A patch week is exactly when that line matters, because a thin post-patch sample looks like a collapse until it is twenty games deep.
Local-first matters more after a reset
An MMR squish rewrites the number next to your name, on Valve’s server. It does not touch your match history, and Dota Companion keeps its copy of that history in your browser: no account, no API key, nothing uploaded anywhere, OpenDota read directly, and Steam Community calls deliberately paced to stay under the rate limits. Whatever the medal says after the reset, your recent games are still your recent games and the baseline is still yours.
Get it
Install Dota Companion from the Chrome Web Store — free, version 1.9.0. Enter a Steam ID, let it pull your recent matches, and read the status line before you read anything else. If your profile is private, the Overview says so rather than leaving you to guess.
NEXT: THE EXTENSION → the Dota Companion project page has the feature tour and the full changelog. The build story behind 1.9.0 is in A missing stat is not a zero.
Leave a Reply