Our recent work on using “conservation of information” as a foundational principle of computation requires a highly symmetrical computational model. In particular, the type structure of such language includes duals to conventional sum and product types, the so-called negative and fractional types. Although we have a reasonable operational understanding of computations involving such types, previous attempts to find an adequate semantic model have failed. A very recent investigation shows, however, that the theory of games founded by Conway in the 70s may have the right ingredients for our desired model. The current status of this investigation is speculative at best, so this will a rather informal talk that aims to introduce the basic definitions of our computational model and of Conway games and argue that there is a good match without providing any concrete results.