You can now show Scanr.ai a chart pattern and have it find that pattern across the market. The Model Trainer is out of testing on the web and in the iOS and Android apps, for Super and Super Pro. This release also moves the AI features to paid plans and changes what a saved chart-pattern filter matches. If you have one of those, read that section before your next scan.

The Model Trainer

Open Model Trainer → Train, name a model and load a ticker. Slide the window over the stretch of chart that shows your pattern and add it as an example. If a saved scan already finds the setup, pick it as the source and take examples from its latest matches. You can start from any chart, too: Train a model on this sits in the ticker menu.

With three examples in, press Train model. The model scores every stock in the market against them, and the Candidates tab lists what it found. Accept the charts that fit, reject the ones that don't, and nudge a window first if the model framed it a few bars off. Each verdict becomes an example for the next Retrain. A threshold slider under the chart shows how many stocks match at each setting.

To scan with a model, add the Pattern model filter, pick your model and set a rank cut such as the top 0.5%. Sort by Pattern score to see the closest matches first. The filter works with the Time Machine, so you can see what your model would have found on a past date, and you can ask the AI terminal to scan with one of your models by name. If you'd rather not start from nothing, Model presets has official models you can scan with or copy and retrain.

Super keeps 5 models and Super Pro keeps 20, and one scan can hold up to three pattern-model filters. Models train on daily charts for now.

In the app, My Models takes the tab Performance used, and Performance moves into More. You can build and train a model on your phone, swipe through candidates to accept or reject them, and edit a scan's pattern-model filter. Tap Train on a ticker, a scan result or a watchlist row to start a model from that chart.

AI moves to Super and Super Pro

The AI terminal, natural-language scans and Refine now need a Super or Super Pro plan. Free and expired plans used to get a small daily allowance; they now get an upgrade prompt where the AI input was. The API add-on and MCP keep their own credits and work as before.

Chart-pattern filters get a State

Chart-pattern filters now take a State: Active, Forming, Confirmed breakout, Invalidated, Target reached or Any state. Active, the default, covers forming and confirmed patterns, so a pattern that failed or already hit its target drops out.

The lookback changed as well. It used to count calendar time back from the pattern's last swing. It now counts trading bars of the filter's timeframe, back from the bar where the pattern became visible. Hover the field for a reminder.

Your saved scans move to the new rules. Every saved chart-pattern scan uses the Active default, and its lookback covers a different stretch of time: shorter on daily charts, longer on hourly and 5-minute ones. Expect fewer matches. Set State to Any state to include finished patterns again.

Scanner pages show a dated run

The public pages under scanr.ai/scanners now show a run from one day to two weeks back, depending on how fast each setup goes stale, and print that run's date. Earnings and split calendars stay current. In the app, any scanner your plan can run shows its latest results. On a free account, a paid scanner shows a delayed sample of five tickers, or none when fewer than 25 matched.

Also in this release

New accounts get a short tour of the app on first sign-in. Take the tour in the sidebar footer plays it again.
Long AI terminal answers no longer end cut off or blank. A turn now gets 150 seconds, and one that runs out of time tells you.
When the AI refines your scan and runs it in the same step, it runs the refined version.
A saved scan's background refresh and alerts follow the filters you last saved. Before, a lost sync could leave them running on your old filters.
A theme's Top Tickers list names its return period, so a week's gain no longer reads as today's move.

Tell us what you train, or anything you run into, at feedback.scanr.ai or [email protected].

— Eoin & The Scanr.ai team