True Grind logo
TRUE GRIND
Sign in with Steam
public roadmap

Community feedback, tracked in public

Who asked for what,
what shipped,
and what still needs research.

This page credits community suggestions by name where they were given publicly, keeps their requested features visible, and shows the current product status without pretending unfinished work is already solved.

View tracked requestsOpen the App
Tracked requests
3

Named public suggestions currently represented on this page.

Remaining items
0

Entries that are still not fully shipped and need more product or implementation work.

Shipped entries
3

Roadmap items with a completion date attached.

Feedback entries

Current community requests

Status stays conservative on purpose. If something is not production-ready, it is not marked as shipped.

Trust and safety

Community report layer for suspicious players

Shipped
Requested by
MrGlueru/Mrgluer
Source
Reddit conversation with ZigmundTrash
Logged from
March 2026
Requested outcome

Add a community-facing report flow so players can flag profiles directly on the site.

Status note

Signed-in profile pages now support one report per reporter/profile pair, cautious wording, and an edit cooldown to suppress duplicate spam. Reports stay separate from the suspicious-play model score.

Shipped on March 26, 2026

Demo intelligence

Suspicious-play confidence score from demo data

Shipped
Requested by
MrGlueru/Mrgluer
Source
Reddit conversation with ZigmundTrash
Logged from
March 2026
Requested outcome

Use parsed demo data to estimate how suspicious a player's behavior looks without claiming proof.

Status note

The parser now ships a suspicious-play score with 4 bands, factor-by-factor breakdowns, sample-strength scaling, and signed-in match/profile analytics. The product still avoids calling it a literal cheating probability.

Shipped on March 26, 2026

Match forecasting

Win-loss probability calculator based on Glicko-2

Shipped
Requested by
Discord community
Source
True Grind Discord
Logged from
March 2026
Requested outcome

Estimate likely match outcomes from player or team ratings instead of only showing raw historic results.

Status note

A signed-in Glicko-2 calculator is now available in the app. It runs client-side, accepts per-player team inputs, and computes team win probabilities without adding server load or requiring stored rating history.

Shipped on March 26, 2026

Methodology Notes

Suspicious-play score notes

The suspicious-play feature is live, but the modeling caveats still matter. The current parser supports a strong review signal, not an honest cheating percentage.

available now

Signals we can compute from the current parser output

These are derivable today from replay frames, kill events, weapon fires, grenade events, flash events, and per-player summary stats already written by the parser.

Headshot, wallbang, through-smoke, and noscope rates

Per-player kill events already carry these flags, so unusually high lethal-event mixes can be tracked across one match or many demos.

Weapon-specific kill profile

The parser stores weapon kill counts, which makes it possible to compare suspicious event rates by weapon instead of only on overall K/D.

Weapon-fire cadence and burst patterns

Shot timestamps and player IDs make it possible to measure burst size, shot spacing, and how often lethal results happen immediately after firing starts.

View-angle snap approximation

Sampled player positions plus view yaw and pitch give enough signal to estimate abrupt aim corrections around engagements, even if the sampling is still coarse.

Flash and utility conversion context

Player blind events, grenade trajectories, and smoke timing make it possible to separate utility-assisted kills from isolated dry peeks.

Cross-demo consistency

Once the same Steam profile has several parsed demos, the site can compare whether suspicious patterns are rare outliers or repeat over time.

needs parser changes

Signals that need parser or data-model upgrades first

These would materially improve score quality, but the current artifact shape does not preserve enough attribution or precision yet.

Attacker attribution for non-lethal damage

Current hurt events keep the victim and hitgroup but not the attacker, so non-kill aim quality cannot be tied back to one suspect reliably.

Denser fight snapshots

Replay frames are sampled every 4 ticks. That is enough for broad movement and angle trends, but too coarse for tight reaction-time claims.

Scoped, recoil, and movement-state context

A stronger model would want scoped state, velocity, crouch state, and recoil or punch-angle data before judging whether aim behavior is abnormal.

Visibility and line-of-sight truth

Positions alone do not prove whether an enemy was actually visible through map geometry. A reliable model needs true LOS or map-geometry-aware visibility checks.

Calibrated probability labels

A real percentage requires a labeled dataset of confirmed cheaters and confirmed clean players. Without that, the product should show a suspicious-play score instead of a cheating probability.

guardrails

Guardrails before anything public-facing goes live

The trust and safety side matters as much as the model itself, otherwise the feature becomes a brigading tool.

Reports should be weighted, not counted flat

Community reports need reputation, recency, and duplicate suppression so one streamer clip cannot flood a profile with identical accusations.

Use cautious wording

The UI should say suspicious-play signals or confidence bands, never proof, verdict, or ban recommendation.

Separate model score from manual reports

If both exist, the system should show them as two different inputs instead of blending them into one opaque number.

Next step

The safe product direction is a suspicious-play score plus guarded community reports, not a one-click accusation feed.

That keeps the feature useful for pattern discovery while reducing abuse risk and false certainty.

Go to App HomeContinue the discussion