After careful consideration, I have decided to discontinue public access to the TorCAST platform effective August 31st, 2026.
Moving forward, TorCAST will be used exclusively by the Storm Trooper Network as an internal severe-weather intelligence and operational decision-support tool.
TorCAST was built to provide earlier awareness of severe-weather potential, help identify areas requiring closer attention, and support informed operational decisions. That mission will continue, but within a more focused environment where the platform can directly support Storm Trooper Network operations, field teams, weather-camera deployments, and severe-weather coverage.
Public utilization of the platform has remained very limited, while the ongoing costs associated with data ingestion, model processing, infrastructure, and system maintenance have continued to grow. Given that imbalance, maintaining TorCAST as a public-facing service is no longer practical.
This change will allow me to concentrate development resources on improving the tools that are most useful to our organization without the additional costs and demands associated with supporting a public platform.
I sincerely appreciate everyone who registered, tested the system, provided feedback, or followed TorCAST's development. The technology is not going away—it is simply transitioning from a public service into a private operational platform.
The same rules every time, no cherry-picking — the engine scores itself against the public record.
A hit is when TorCAST showed elevated potential for a location and time, and verified severe weather then occurred in that area within the window the score covered. "Elevated" means a TorCAST score at or above our published threshold — not just any non-zero reading. If the atmosphere delivered where we said it could, that snapshot is credited as a hit.
A bust is the opposite: TorCAST showed elevated potential, but no qualifying severe report or warning was confirmed for that area and window. Busts are counted honestly into the bust rate — a high score that didn't pan out still counts against us. Quiet locations that we never elevated are not "busts"; we only grade the calls we actually made.
We grade against the official public record — the same ground truth forecasters are judged on:
Storm reports arrive late — LSRs trickle in for hours, and damage surveys can land a day or more after the storm. We don't grade a snapshot the instant the sky clears. Each TorCAST snapshot stays open for verification through a settling window, and the public figures back-fill automatically as new confirmed reports and survey results are published. A snapshot's grade can move from "unverified" to "hit" days later when the report it was waiting on finally posts.
Every TorCAST score is stamped with where (its grid location) and when (the engine cycle and the window it was forecasting). When a confirmed report or warning is published, we match it back to the snapshots whose location and time window contain it — the snapshot has to have been on the record before the storm, so there's no hindsight. That time-and-place match is what turns a months-old snapshot into a graded hit, miss, or bust.
TorCAST is a strategic, hours-ahead layer — it grades how well it flagged the potential, not whether it replaced the warning. It does not tell you the chance a storm will happen; it tells you how serious the atmosphere could become if storms develop near you. We hand off to the official NWS warning for the final go/no-go.
TorCAST turns severe-weather potential into one clear number for your exact location — a stadium, a campus, a job site — so you know when to prepare before warnings are issued.
The single highest active potential across TorCAST's grid — one calibrated number per location, refreshed every engine cycle.
National outlooks paint a whole region with one category, once or twice a day. TorCAST scores your exact location on a continuous 0–10 scale, every engine cycle — so you act on what's coming to you, earlier.
If severe weather forces a decision on your day, TorCAST is built for you — no meteorology degree required.
Delay, shelter, or clear the stands — with a defensible record of when you knew.
Shelter calls and dismissal timing, briefed before the buses roll.
Stand up the EOC and pre-stage assets with hours, not minutes, of warning.
Festivals, golf, races — a 20-minute shelter plan triggered on a clear number.
Pre-stage crews and curtail wind/solar assets ahead of the damaging-wind window.
Secure equipment, move livestock, stand down cranes — before the hail and gusts.
Drop your venue, campus, or site — by GPS or address. Existing accounts watch two.
A single 0–10 score, the dominant threat, confidence, the peak window, and a clear, do-this-now action — updated every engine cycle.
Escalate your own plan hours ahead, while the threat is still building — not when it's overhead.
Every TorCAST score is scored back against confirmed storm reports and warnings. These figures update automatically — they're the engine's own running record, not a marketing claim.
Did severe weather actually occur in the places where TorCAST showed elevated potential? Higher is better.
How many confirmed reports & warnings the engine has checked itself against to date.
How often TorCAST showed elevated potential where verified severe weather never happened. Lower is better.
What counts as a hit, what counts as a bust, which reports we use, and how we match an old TorCAST snapshot to the storms that came later.
Type any U.S. town — get its live TorCAST reading instantly, straight from the current engine cycle. No account needed to look.
TorCAST is no longer accepting new signups. Existing accounts remain available until August 31st, 2026, after which the platform becomes an internal Storm Trooper Network tool.
Public access ends August 31st, 2026. New signups are closed; existing accounts remain available until that date.
Sign in to your account →TorCAST is decision-support intelligence, not an official warning service. Always follow National Weather Service warnings and local emergency officials.