Twenty years of trying to give players one identity

Every few years someone decides that a player should have one identity that follows them between games. It is an obviously good idea. It has been tried, in public, with real money, at least a dozen times.

This is what happened to those attempts, and why exactly one of them is still standing.

The freshest grave is six months old

Ready Player Me was the most successful cross-game avatar platform ever built. One avatar, usable in more than 10,000 apps, games and virtual worlds. It raised a 56 million dollar Series B led by Andreessen Horowitz in August 2022.

Netflix acquired it in 2025. Public services ended on 31 January 2026.

Netflix kept the technology to power cross-game avatars inside its own games. The open system became a closed one, and ten thousand integrations became ten thousand things to rip out.

That is not a story about a failed product. Ready Player Me won its category and was bought for winning. The acquisition is what ended it, which is the same sentence we had to write about OpenFeint, bought by GREE for 104 million dollars and switched off with under thirty days notice, and about Scoreloop, bought by BlackBerry and closed in 2014.

Three of the most successful attempts at shared player identity died the same way, and none of them died of failure.

The blockchain wave

Between roughly 2021 and 2024 the answer was going to be a chain. Your identity, your items and your reputation would live somewhere nobody could switch off, and every game would read from it.

About 93 percent of GameFi projects have ceased operations.

The post-mortems mostly blame token economics, and that is fair, but it hides the more useful failure. Interoperability was always described as a technical problem, and it never was one. Moving an item between two games is trivial to encode and nearly impossible to make meaningful: a sword balanced for one game is either overpowered or useless in the next, and no chain fixes that. The projects that survived, Enjin and Loom among them, survived by becoming ordinary developer tooling.

A shared identity has the same shape. The hard part is not storing it. The hard part is that it has to mean something in a game whose designer you have never met.

The one that lived

Gravatar launched in 2006, built by Tom Preston-Werner, who went on to co-found GitHub. Automattic bought it in 2007, when it had 115,000 users and WordPress.com's own avatar system had around a million.

By 2013 it had passed a billion profiles. Today it serves more than eight billion avatar requests a day.

Twenty years. Still running. Nobody talks about it because it never needed talking about.

So why did the small one survive while the ten-thousand-integration one did not?

Gravatar is not a service you integrate. It is a URL you construct. Hash the email address, put the hash in a path, get an image. The site showing the avatar needs no account, no API key, no SDK, no callback, no contract and no relationship with Gravatar at all. If Gravatar disappeared tomorrow, every integration would degrade to a broken image, not a broken build.

Compare the failure modes. Ready Player Me shutting down breaks ten thousand products. Gravatar shutting down breaks ten thousand pictures.

And the second reason, which matters as much: Automattic bought it because WordPress needed it. It was infrastructure for the acquirer's core product, not a promising asset in a portfolio. That is the difference between being bought and being absorbed.

The integration surface decides how long a shared identity lives. An SDK is a dependency. A URL is a convention. Conventions outlive companies.

The elephant nobody mentions

Any honest version of this article has to say the following out loud.

Epic Online Services already exists, it is free, and it does most of this.

Cross-platform accounts, login, friends, presence, achievements, stats and leaderboards. Free for all developers, all engines, all stores, on Windows, macOS, Linux, PlayStation, Xbox, Switch, iOS, Android and web. Epic describes it as reaching a cross-platform network of nearly a billion players, and it is battle-tested because Fortnite runs on it.

If you want one player identity with leaderboards across games, that is available today at no cost from a company that is not going anywhere.

It also carries three conditions. It needs an SDK, which puts it on the wrong side of the sentence in bold above. It assumes a game substantial enough to justify an integration, which a browser game built in an afternoon is not. And it is owned by one company, which is the same structural risk as everything in the graveyard, only with a much larger company holding the switch.

Those are real limits. They are not a reason to pretend EOS does not exist.

Where avatars actually got standardised

The interesting counter-example is not a service at all.

VRM is a 3D avatar file format from Japan, released in 2018. It became the de facto standard there across VirtualCast, VRoid Studio and cluster. In 2026 the VRM Consortium and the Khronos Group are folding VRM functionality into official glTF extensions, and the Metaverse Standards Forum has an Interoperable Characters and Avatars working group with a VRM Consortium co-chair.

Note what that is. Not a platform, not a login, not a company. A file format, going into an existing open standard, through a standards body.

Meanwhile the biggest avatar ecosystems by headcount, Roblox and Fortnite, remain completely siloed, and will stay that way, because an avatar is a cosmetics business and cosmetics do not travel.

The pattern across twenty years is consistent: formats spread and survive, services concentrate and die. A format costs nobody anything to keep alive. A service costs somebody money every single day, and one day that somebody sells.

What this means for a universal high score protocol

We are building one, so here is the reading of our own homework.

Design it so it survives us. If the thing only works while our server answers, we have built a company, not a protocol. That is the OpenFeint shape and it ends the same way.

Make the integration a convention, not a dependency. The closer it is to constructing a URL, and the further from installing an SDK, the longer it lives and the more games can afford it.

Do not pretend EOS is not there. It covers the games big enough to integrate an SDK. The gap is everything below that line: jam games, browser games, things made in an afternoon that will never justify a build step. That gap is real, it is where the growth is, and it is not served.

Expect the acquisition question early. Every project on this page that got big got an offer. The answer to "what happens to the data when you sell" has to exist before anyone embeds a line of your code, not after.

The protocol itself, what a score record has to contain and what it deliberately leaves out, is a separate piece.

---

Sources

• Ready Player Me: 56 million dollar Series B led by Andreessen Horowitz, August 2022, per Road to VR and Business Wire; support for more than 10,000 apps and games; Netflix acquisition 2025 per Variety; public services ended 31 January 2026
• OpenFeint: acquired by GREE for 104 million dollars, 2011; shutdown 14 December 2012 with under thirty days notice. Scoreloop: acquired by BlackBerry 2011, closed 1 December 2014
• GameFi failure rate of approximately 93 percent, as reported in 2025 and 2026 crypto gaming reviews; Enjin and Loom continuing as developer tooling
• Gravatar: founded 2006 by Tom Preston-Werner, acquired by Automattic 2007 at 115,000 users; over one billion profiles by 2013; more than eight billion requests per day today
• Epic Online Services: free cross-platform accounts, leaderboards, achievements and stats across nine platforms, per Epic's own documentation and coverage in PCGamesN and MCV
• VRM: released in Japan 2018, de facto avatar standard there; VRM Consortium and Khronos Group liaison to propose VRM functionality as glTF extensions; Metaverse Standards Forum Interoperable Characters and Avatars working group

Retrieved 4 August 2026.