What I mean is that, the most frequently used objects, if accessed roughly with an even period of time, tend to be also the least frequently used objects, so if the accesses are very even LFU and LRU tend to be similar. For uneven access patterns, they are different, but usually LRU is used because access patterns are even so that it approximates LFU, since what we want to have in cache is, without other application-specific clues, the objects that we use more often. Knowing
exactly the access pattern one can use an application-assisted strategy which is optimal, but without clues LRU adapts well to different scenarios while LRU may fail in a catastrophic fashion.
Also note that your idea of LFU (in this context), from what you write, is perhaps one that does not adapt over time. Redis LFU employs a decay strategy so that the recent frequency is computed, not the whole-lifetime frequency of access, so even short lived items accessed very frequently for some time, will do well and will not get easily evicted.
Full story: http://antirez.com/news/109