One way of yielding the same effects without even factoring in search inputs would be to assume a probabilistic relationship between the words on an origin page (e.g. a Google search result page, or a bog standard web page) and the destination URL the user clicks through to. This seems like a pretty reasonable design parameter for the Bing Bar to learn "Suggested Sites" and if you're collecting that information anyway why wouldn't you add all those associations into your search engine to help it rank those tricky obscure search queries?
For synthetic words like "hiybbprqag" that rarely/never show up on the internet [except on Google search result pages generated by Google engineers systematically searching for it], the basic probability algorithm would weight heavily towards assuming an association between "hiybbprqag" and the destination URLs viewed by people immediately after looking at pages referencing "hiybbprqag". Since probably the only people looking at pages referencing "hiybbprqag" were Google testers searching for it, who had been instructed to always click on the "synthetic" Google result, the probability of someone viewing a page referencing "hiybbprqag" subsequently going to http://www.teamonetickets....wiltern-map.html* would be close to 1 - suggesting a pretty strong association between the terms.
If you incorporated these associations into the Bing search engine in any way, it would be perfectly reasonable for a search engine to assume that that page is the most relevant result for hiybbprqag*, given the lack of any alternative data on what to show.
Obviously this isn't as simple as the other solution (and Google could have done more sophisticated tests which rule out this kind of algorithm as being behind the results), but the competitive end of search isn't simple.
For synthetic words like "hiybbprqag" that rarely/never show up on the internet [except on Google search result pages generated by Google engineers systematically searching for it], the basic probability algorithm would weight heavily towards assuming an association between "hiybbprqag" and the destination URLs viewed by people immediately after looking at pages referencing "hiybbprqag". Since probably the only people looking at pages referencing "hiybbprqag" were Google testers searching for it, who had been instructed to always click on the "synthetic" Google result, the probability of someone viewing a page referencing "hiybbprqag" subsequently going to http://www.teamonetickets....wiltern-map.html* would be close to 1 - suggesting a pretty strong association between the terms.
If you incorporated these associations into the Bing search engine in any way, it would be perfectly reasonable for a search engine to assume that that page is the most relevant result for hiybbprqag*, given the lack of any alternative data on what to show.
Obviously this isn't as simple as the other solution (and Google could have done more sophisticated tests which rule out this kind of algorithm as being behind the results), but the competitive end of search isn't simple.