Torrent Gear

How to optimize traffic when playing a mobile game

1) Why optimize traffic

Fewer delays → more stable session and higher hold.

Data savings → lower user costs and the risk of a "cut tariff."

Fast start → more runs of games with fluffs/ads.

Reliability on a weak network (3G/cafe-Wi-Fi/roaming).

2) Metrics that are really worth watching

First Contentful Paint (FCP )/Largest Contentful Paint (LCP): when the player "saw" and when "can play."

INP/TBT Interface Response.

Traffic/session (MB) and peak bit rate.

RTT/jitter/losses (especially for live games/streams).

Cache Hits - Percentage of requests from the application/CDN cache.

3) Network stack: basic hygiene

Enable HTTP/2/HTTP/3 (QUIC) for multiplexing and more robust packet loss operation.

TLS session resumption and 0-RTT (for H3) - less hand chats.

DNS-prefetch/Preconnect to CDN and game providers.

Competent cache policy: 'Cache-Control', 'ETag', versioning assets.

4) CDN and geography

Put static and media on the CDN with PoP closer to the user.

Enable image resizing/' accept '-based negotiation on CDN (WebP/AVIF).

For live video - multi-bitrate profiles on the edge (HLS/DASH).

5) Compression and formats (which actually saves tens of percent)

Images: WebP/AVIF + 'srcset/sizes', sprites and SVG icons.

Fonts: WOFF2, subset for the desired glyphs, 'font-display: swap'.

Video: H.264/HEVC/AV1 (where available), poster instead of autoplay.

Text/JSON: Brotli (br)> Gzip, enable on CDN/server.

JS/CSS: minification, tree-shaking, code-split.

6) Gaming canvas: slots, minigames, canvas/WebGL

Render for adaptive DPR: 'devicePixelRatio' limit to 1. 5-2 on mobile - sharpness persists, traffic/CPU drops.

Texture atlases and texture compression (ASTC/ETC/BC, where supported) → fewer downloads.

Lazy swapping of assets across levels/screens, not "all at once."

Remove "heavy" shadows/filters, limit the frequency of animations to 30-45 fps on weak devices.

For iframe slots: negotiate with providers about light assets and batch preloading only critical resources.

7) Live games and streams: save megabytes without pain

Adaptive bitrate (ABR) with thresholds of 360p/480p/720p; profile selection by width/RTT.

Low-Latency HLS/DASH only where needed; do not enable LLC for everyone.

Audio bitrate 64-96 kbps for speech is often enough.

Off lobby autoplay: show poster/animated GIF/webm preview.

8) Real-time communication

WebSocket: binary protocols, pack messages, heartbeat every 25-30 seconds.

WebRTC-data - only for narrow cases; avoid "unnecessary" NAT bypass if it's not about media.

Compress the payload (protocol buffers/MessagePack), do not drive the "fat" JSON.

9) PWA/Service Worker: Traffic Board on Mobile

App Shell: we cache the header/navigation and skeleton - instant first screen.

Runtime caching: 'Stale-While-Revalidate' for pictures, 'Network First' for API with TTL.

Background sync: deferred telemetry/logging, without blocking interaction.

Offline fallback: understandable screens instead of emptiness (saving retrays and unnecessary requests).

10) Smart downloads and priorities

Critical CSS inline, the rest on request.

'defer/async'for scripts, import () for later screens.

Lazy-load game lists (20-30 cards per pack), 'IntersectionObserver'.

Prefetch by intention: when the user lingered on the card → pull up the game assets.

11) Billing and checkout: traffic is important too

Use system payment dialogs (Apple/Google Pay) - they are more economical and sustainable.

Minimize redirects and extra analytics pixels on payment steps.

In the crypto module, do not load all networks/icons - only the selected network/coin.

12) Telemetry and A/B without "gluttony"

Collect only the necessary events, batch and send once every N seconds/by size.

Turn off the debug logs in the prod, truncate the fields in the events.

A/B flags - via easy remote-config, do not pull megabyte schemes.

13) Practices for players (quick traffic wins)

On iOS/Android, enable Data Saver/Traffic Saving.

If possible, play over Wi-Fi 5/6; on a mobile network, avoid "1-2 sticks" - higher loss.

Disable autoplay video/preview in the settings.

In Telegram and the browser, clear the cache every couple of weeks - but not before playing often (the cache helps).

Keep track of the application/PWA update - new versions are often more economical.

14) Developer/Product Checklist (one page)

1. HTTP/2/3, TLS 1. 3, preconnect to CDN/game domains.

2. CDN with picture resize, AVIF/WebP, Brotli per text.

3. App Shell + SW: offline-fallback, runtime-кеш, background-sync.

4. Lazy loading of assets, code-split, critical CSS inline.

5. Dynamic DPR (≤2), compressed textures, 30-45 fps on weak.

6. ABR video width/RTT, off autoplay in the lobby.

7. WebSocket with packaging; compressed protocol for data.

8. Telemetry by batchami; disabled prod-debag.

9. Cashier without unnecessary redirects; System payment dialogs

10. Monitoring: LCP/INP/traffic/session, cache hits, RTT/loss.

15) Frequent mistakes and how to fix them

Replace the video/stream autoplay in the → lists with a poster/preview.

We pull 3 × assets on all devices → use 'srcset '/DPR profiles.

Giant JS bundles → route separation, removal of unused deps.

Zero Cache Control → Configure TTL/ETag and versioning.

Chat/telemetry spam → batch, increase the heartbeat interval.

Everything in one WebSocket channel (game + chat + analytics) → divide by criticality.

16) Mini patterns that 'make the weather'

Button "Reduce video quality" in live tables with a bad network.

Placeholder covers for games before loading retina.

Last Session Save (State Cache) - Fewer retries.

Deeplink to the last running table/slot - minus two screens and a pack of assets.

17) FAQ

Will traffic optimization degrade quality?

If you do it adaptively (DPR/ABR/' srcset ') - no: you give the best balance of quality/speed for the device and network.

Do all users need to enable Low-Latency mode?

No, it isn't. It is more expensive in traffic and sensitive to losses. Leave for tournament/live cases.

PWA instead of native client - traffic below?

Often yes: less SDK and background threads, plus SW cache. But depends on implementation.

How much does AVIF/WebP save?

An average of 25-45% versus JPEG/PNG with no discernible loss of quality.

Should we always lower DPR?

Reduce dynamically on weak devices/low network; on flagships with Wi-Fi 6, you can keep 2. 0.

Traffic optimization is not about "cutting everything," but adapting quality and volume to the device, network and scenario. Combine fast network stack (HTTP/3, CDN, cache), smart assets (WebP/AVIF, textures, ABR), neat canvas and PWA cache, cut telemetry noise - and get fast downloads, stable gameplay and tangible data savings. Players fall less often due to the network, more often return, and the product wins in both retention and infrastructure costs.