The so-called "luck" is only a measure of past success and always averages out. Next week this thing you call luck will be 20% and you will be wondering where your luck went and who jinxed the pool, etc.
No, bitcoin luck is computed as a factor of the probability of finding a block given your (the pool's) hashing power and the network hashing power, given that the number of blocks that can be released by the network in the effective timeframe is known.
In our case (and this is a conspiracy theory of some magnitude!), I am thinking the new hardware that hit the market of late, ahem... S5 ... have not been deployed en-masse on slush, but elsewhere, and that is reducing their luck thus we pick up the pieces where we were beaten by a matter of seconds before! Of-course, that theory assumes there is something wrong with the S5's (even the S3 variants and other bitmain rigs updated with the Jan firmware!).
One small (but important) correction: Pool luck is dependent on the pool hashrate and current difficulty, not network hashrate. In this case it would matter because adding hashing power to the rest of the network (no matter what kind it is) doesn't affect Slush's luck. The ~10PH/s on Slush will still find X blocks/day whether the network is 20PH/s or 2000PH/s, at least until the next difficulty retarget.
There is also no limit on the number of blocks that can be released on the network in a given timeframe. Difficulty will adjust to try and get a block every 10 minutes, but you could have all 2016 blocks of one difficulty period found in one day as opposed to two weeks.