Speed Benchmark: Production vs. CMS-free Lobby vs. Competitors
16 September 2026 — four sites, 40 Lighthouse runs, median of 5 per site per device. Both HighStack arms measured on the logged-in lobby; Shuffle and Stake on their public landing pages.
On mobile, the CMS-free lobby scores 70 — ahead of our own production lobby at 60, and roughly 2–3× the competition (Shuffle 36, Stake 22). It fills the screen more than twice as fast as production: Speed Index 4.1s vs 9.3s.
On desktop the order flips: production leads at 89, the CMS-free lobby follows at 85, then Stake 77 and Shuffle 66. Desktop is not where the CMS costs us.
Comparability notice. This run was measured against different targets from the automated weekly sweep — stake.com rather than stake.bet.br, and production rather than staging as the headline HighStack figure. The numbers below are sound for the sites named, but should not be read against the 2026-07-05 baseline. See the fine print.
The headlineGoogle Lighthouse performance score
Median of 5 runs, 0–100, higher is better. Both HighStack rows are the authenticated lobby — the real product, a far heavier page than a marketing landing page.
70CMS-free lobbymobile · best of the four
60Production lobbymobile · today's live site
36Shufflemobile · best rival
22Stakemobile
Site
Desktop median of 5
Mobile median of 5
Desktop runs
Mobile runs
HighStack — CMS-free lobbypr-1215 preview build
85
70best of all four
88, 85, 85, 88, 83
70, 70, 70, 70, 66
HighStack — productionhighstack.com, CMS-driven
89best of all four
60
88, 89, 89, 92, 87
54, 56, 60, 60, 60
Stakestake.com landing page
77
2218–75, see caveat
77, 77, 83, 90, 76
22, 18, 19, 75, 27
Shuffleshuffle.com landing page
66
36
66, 74, 66, 66, 66
39, 37, 36, 34, 36
Stake's mobile score is unstable. Four runs landed at 18–27 and one at 75. The median of 22 is the honest summary, but their landing page clearly serves materially different weight run to run. Treat their mobile figure as a range, not a point.
Under the hoodThe metrics behind the scores
Median of 5. Lower is better on every row here. Speed Index is the one that tracks how fast a page feels — it measures how quickly the screen actually fills in.
Mobile — Slow 4G, 4× CPU throttle
Site
First paint
Largest paint
Speed Index
Blocking time
Layout shift
Server response
Requests
CMS-free lobby
2,572ms
7,350ms
4,090ms
69ms
0
695ms
632
Production
2,162ms
8,600ms
9,270ms
235ms
0
1,684ms
949
Stake
6,814ms
7,114ms
11,185ms
552ms
0.34
277ms
722
Shuffle
7,433ms
17,351ms
10,297ms
843ms
0
102ms
293
Desktop
Site
First paint
Largest paint
Speed Index
Blocking time
Layout shift
Server response
Requests
Production
502ms
1,719ms
2,122ms
0ms
0.01
1,373ms
970
CMS-free lobby
786ms
1,777ms
2,754ms
0ms
0.006
832ms
645
Stake
1,394ms
1,982ms
3,391ms
78ms
0.003
273ms
717
Shuffle
1,657ms
4,164ms
2,822ms
33ms
0
105ms
348
Side by sideHow fast the screen fills — mobile
Speed Index, median of 5. Shorter is better. This is the closest single number to what a player on a phone actually experiences.
CMS-free lobbyHighStack
4.1s
ProductionHighStack
9.3s
Shufflelanding page
10.3s
Stakelanding page
11.2s
Side by sideLighthouse score — both devices
Mobile
CMS-free lobby
70
Production
60
Shuffle
36
Stake
22
Desktop
Production
89
CMS-free lobby
85
Stake
77
Shuffle
66
What changedProduction vs. the CMS-free lobby
The comparison this benchmark was commissioned for.
Mobile
Production
CMS-free
Difference
Lighthouse score
60
70
+10 points
Speed Index
9,270ms
4,090ms
2.3× faster
Largest paint
8,600ms
7,350ms
−1,250ms
Blocking time
235ms
69ms
−166ms
First paint
2,162ms
2,572ms
+410ms
Requests
949
632
−317
Read this comparison with one caveat. Production runs on production infrastructure; the CMS-free lobby was measured on a preview deployment. So the gap between those two rows contains an infrastructure difference as well as the CMS difference, and it cuts both ways — production has the better CDN and caching, which is visible in its faster first paint and its desktop lead. An isolated test of the same two builds on the same infrastructure showed a smaller but same-direction mobile advantage for the CMS-free lobby.
Plain-English glossaryWhat the terms mean
Lighthouse score
Google's 0–100 summary of loading performance. It is the industry-standard ranking number and weights the metrics below.
First paint (FCP)
When the first text or image appears. The end of the blank screen.
Largest paint (LCP)
When the biggest element finishes loading — usually the hero image or main banner. Google treats this as "the page looks ready".
Speed Index
How quickly the screen visually fills in overall. The closest single number to how fast a page feels.
Blocking time (TBT)
How long the page is frozen to taps and clicks while scripts run. High blocking time is the "I tapped and nothing happened" feeling.
Layout shift (CLS)
How much content jumps around while loading. Lower is better; 0 means nothing moved.
Server response
How long the server took to send the first byte of the page.
The fine printHow we measured & what to keep in mind
Tool: Google Lighthouse 12.8.2 driving Chrome 152, every site under identical conditions, cold load. Mobile is Lighthouse's default profile (Slow 4G, 4× CPU slowdown); desktop is its desktop preset.
Scores are the median of 5 runs per site per device. All 40 runs completed; none failed. Full raw reports are in this report's lighthouse/ folder.
All four sites ran in the same browser session, alternating between sites within each round, so network conditions affected every site equally.
Every run was verified to have measured the intended page — both HighStack arms were checked to be on the authenticated lobby, not a redirect to the login screen.
HighStack is measured on the authenticated lobby; competitors on their public landing pages. This is the same methodology as previous reports and it is the tougher test for us — their landing page is a marketing page, ours is the full product with live wallet, game catalogue and bet feed.
All sites were measured in a headed browser. Both Stake and our own production site sit behind a bot-protection interstitial that never clears in a headless browser, so headless testing would have measured a challenge page instead of the real site.
Shuffle geo-redirects to its Portuguese site (/pt) from our test location. That is the page measured.
This run used different targets from the automated sweep, so it is not directly comparable to the 2026-07-05 baseline. The scripted pipeline (scripts/targets.mjs) measures stake.bet.br/bem-vindo, shuffle.com/pt?r=bonusregistration and staging.highstack.com as the headline HighStack figure. This run measured stake.com, shuffle.com and production instead, and did not include Roobet or CSGORoll. Stake in particular is a different property — stake.com is not the Brazilian site the archive has been tracking, so this Stake figure should not be read against July's.
File slugs differ from the archive convention. This run wrote hs-prod-cms-* and hs-nocms-*; the archive's contract is highstack-* / highstack-prod-*. Re-running through the pipeline would restore both the targets and the slugs.
Competitor server-response times look excellent (~100–280ms) because a landing page can be fully cached at the edge. Our lobby is per-player and cannot be. This is a structural difference, not a defect.
No video evidence in this report. Previous reports included screen recordings with the browser's own load timers overlaid; this run is Lighthouse only.
Lab, not field. These are controlled measurements on one machine and one connection profile. They predict real-user performance; they do not measure it.