How self-loops on vertices affect quantum walks is an interesting issue, and
self-loops play important roles in quantum walk based algorithms. However, the
original model that adjusting the effect of self-loops by changing their number
has limitations. For example, the effect of self-loops cannot be adjusted
continuously, for their number must be an integer. In this paper, we proposed a
model of adjustable self-loop on discrete-time quantum walk, whose weight is
controlled by a real parameter in the coin operator. The proposed method not
only generalises the situations where the number of self-loops is an integer,
but also provides a way to adjust the weight of the self-loop continuously. It
enhances the potential of self-loops in applications. For instance, we improve
the success rate of the quantum walk based search on a $20\times20$
two-dimension lattice from $23.6\%$ to $97.2\%$ by the proposed method. And the
success rate of the improved search, which only scales as $O(1/\log{N})$ before
being improved, even increases slightly with the size of the lattice. To the
best of our knowledge, this is the first time that such an improvement is
achieved on the quantum walk based spatial search.