And the first Defense of the Ancients was originally a mod for a completely different game. The common theme is polishing gameplay. Team fortress existed and was popular, but between the release of TF classic, with the announcement of TF2, and the actual release there were almost 9 years and a complete rewrite between two radically different versions of the game. At one point people compared it with Duke nukem, claiming it was vaporware and would never release. Truth is, it was in development hell for a long while. They didn't like what the game was at that time. TF classic and TF2 only common thread is class based team death match. Everything else is different. The producers have said that TF2 was resurrected to perfect the netcode, lighting, facial animation rigging, particle system and shading tech for the source engine in anticipation of the visual and gameplay improvements they wanted for HL2ep1 and 2. All three games were produced by the same guy and Gabe noticed what he experimented with on TF2 was worth developing into a finished game. Specially because they dropped all the ideas they didn't like and stripped down the gameplay.

The other side of the coin being that Valve had learned the importance of visual packaging and marketing with Ricochet. With pure gameplay, although wildly acclaimed for being super fun, it didn't reach the mass appeal and cultural impact of half-life. It had great repayable value, but no eye candy or lore to hook people long term. So, when TF2 was a success with its character based marketing narrative, it became the test bed for a myriad of things we now take for granted. Matchmaking, micro transactions, cosmetics stores, etc. (All things that were made to develop the Steam store social features, which was produced by the other guy who made the TF mod originally) Valve only goes hard on things they think are innovative or interesting tech, or at least plain fun to do. If the internal sponsor of an idea get bored or loses support from colleagues, the project just halts.