relaydle

For creators · 5 min read

Recommending other daily games: 22,491 clicks studied

clicks per 100 result-screen views
11.0
of clicks on the first three cards
67.1%
of clicks reached the next game's result screen
43.0%

When a daily game shows a row of other daily games on its result screen, players click it. In October 2026, 12 games on the relaydle rail showed it 204,613 times and got 22,491 clicks: 11.0 clicks per 100 views. 43.0% of those clicks ended with the player finishing the next game, and the first three cards took 67.1% of the clicks on cards the rail ranked.

relaydle runs the rail measured here, so this is our own server data about our own product. Every figure comes from the production ledger, and the queries are listed at the end.

What is an end-screen recommendation rail?

A daily game is over in two minutes, and the result screen is the one moment a player has nothing left to do. The rail is a row of cards on that screen, each one another daily game. A click sends the player to that game; in exchange, the game that showed the rail receives recommendations on the others. Membership is free. How the rail works covers the ordering and the privacy rules.

In October 2026, 12 games ran it on their result screens:

MeasureOctober 2026 (UTC)
Result-screen views of the rail204,613
Clicks22,491
Clicks per 100 views11.0
Clicks where the player reached the next game's result screen9,662 (43.0%)
Clicks made after an earlier rail click in the same session4,667 (20.8%)

The last line is the chain effect: one click in five came from a player who had already followed a recommendation and kept going.

How much does the click rate vary between games?

A lot. The same widget got 4.3 clicks per 100 views on one game and 56.0 on another:

Game showing the railViewsClicksClicks per 100 views
Geopolitix4,8882,73856.0
Devine le prix6,4802,33536.0
Wordzic2,36681834.6
Le Bon Rang8,5372,41928.3
Unicorner Football Game4,12684620.5
Who's More Famous35,6974,12811.6
DailyCountry49,5473,8927.9
DailyCity42,8962,9947.0
DailyMetro47,2652,0354.3

Games with fewer than 2,000 views are left out of the table. Clicks include the rail's later pages; views count the first page only, which is why a rate can be high when players scroll.

The three games with the most views have the lowest rates. We have not isolated why. Treat any single click-rate figure, including ours, as a property of one game and one audience, not of the widget.

Does the position of a card matter?

Yes. Of 22,308 clicks on cards the rail placed itself, 67.1% went to the first three:

Card position12345678910
Clicks6,5655,0723,3352,3081,6011,3451,006572289164

Part of this is the ranking itself: the rail puts the games it expects to be clicked first. The ranking model separates the two by learning one position curve per source game, which is why the order is fitted rather than fixed.

What should a recommendation card show?

Keep the labels. From October 24 to November 1 we split slates 50/50 between two cards, about 96,000 per side:

  • Classic: the game's art, its theme chip, a lime "Play" button and a dark box behind the title.
  • Focus: no chip and no button, more of the art visible, a small play icon and a larger title.

The focus card got 5% fewer clicks: 25.4 per 1,000 slates against 26.7. The whole loss came from the games where players click most: on Unicorner Football Game clicks fell from 101 to 76 per 1,000, on Le Bon Rang from 124 to 101. The big, low-click rails did not move. Clicks turned into players reaching the next game at the same 37% on both sides, so the plainer card lost clicks, not click quality. When players compare cards, the topic label and an explicit button help them pick. The classic card is the only one served since.

Before the live test, we asked a panel of simulated players to rank the games. Its ranking ran opposite to real clicks (Spearman -0.38): in our reading it rewarded novel art, while real players clicked topics and languages they knew. Test cards on real players.

Should you recommend games in another language?

Put games in the page's language first. On a French page, an English game is a fallback card. From October 21 to November 1, about 44,500 slates per side with a fallback card were split between two orders:

  • page-language games always first, fallbacks after;
  • every card ranked together, with fallbacks discounted by a learned language penalty.

The fixed order reached 13.2 new players per 1,000 slates, the mixed order 11.8. On Geopolitix the gap was 111 against 80. In this test, an English game's click rate did not carry over to French readers well enough to rank it among French games.

Do players who arrive from a recommendation come back?

Some do, and we only count the ones we can prove. A return is counted when the same player reaches the same game's result screen on a later day. Of 7,406 arrivals in October with that tracking, 515 (7.0%) came back within seven days; of 5,800 arrivals from October 1 to 24, 246 (4.2%) came back the next day.

These are floors, not rates. A player who comes back and quits before the result screen, or returns from another device or network, is not counted. Destinations that report arrivals from their own server get a direct landing count in their dashboard instead.

The destinations' own analytics give a fuller view. On seven of these games, first-time visitors sent by cross-promotion came back the next day at 8.8%, against 8.4% for visitors from Google (where daily game players come from has the method). The gap is within noise: rail arrivals returned about as often as visitors from Google.

Received clicks were concentrated: Geopolitix received 4,052 and Le Bon Rang 3,683, while the smallest member received 434. The rail gives more to games that are clicked when shown and that send players on, so a new member should not expect the leaders' numbers in its first month.

How do I add recommendations to my own daily game?

  1. Submit your game for a free directory listing.
  2. Apply to the rail on the creator page. Joining costs nothing.
  3. Paste the widget snippet from your dashboard into your result screen, below the score and share button.

Your game is recommended on the others once its rail has 30 views and 3 outbound clicks. If you build your own rail instead, the data above gives you three rules: put it on the result screen, label each card with its topic and a button, and keep the reader's language first.

Sources

  1. How the rail works
  2. Join the rail