We first show that the Traveling Salesman Problem in an n-vertex graph with average degree bounded by d can be solved in O ⋆ (2 (1−ε d )n ) time 1 and exponential space for a constant ε d depending only on d. Thus, we generalize the recent results of Björklund et al. [TALG 2012] on graphs of bounded degree.Then, we move to the problem of counting perfect matchings in a graph. We first present a simple algorithm for counting perfect matchings in an n-vertex graph in O ⋆ (2 n/2 ) time and polynomial space; our algorithm matches the complexity bounds of the algorithm of Björklund [SODA 2012], but relies on inclusion-exclusion principle instead of algebraic transformations. Building upon this result, we show that the number of perfect matchings in an n-vertex graph with average degree bounded by d can be computed in O ⋆ (2 (1−ε 2d )n/2 ) time and exponential space, where ε 2d is the constant obtained by us for the Traveling Salesman Problem in graphs of average degree at most 2d.Moreover we obtain a simple algorithm that counts the number of perfect matchings in an n-vertex bipartite graph of average degree at most [3], and breaking these time complexity barriers seems like a very challenging task.From a broader perspective, improving upon a trivial brute-force or a simple dynamic programming algorithm is one of the main goals the field of exponential-time algorithms. Although the last few years brought a number of positive results in that direction, most notably the O ⋆ (1.66 n ) randomized algorithm for finding a Hamiltonian cycle in an undirected graph [2], it is conjectured (the so-called Strong Exponential Time Hypothesis [8]) that the problem of satisfying a general CNF-SAT formulae does not admit any exponentially better algorithm than the trivial brute-force one. A number of lower bounds were proven using this assumption [6,10,11].In 2008 Björklund et al. [5] observed that the classical dynamic programming algorithm for TSP can be trimmed to running time O ⋆ (2 (1−ε∆)n ) in graphs of maximum degree ∆. The cost of this improvement is the use of exponential space, as we can no longer easily translate the dynamic programming algorithm into an inclusion-exclusion formula. The ideas from [5] were also applied to the Fast Subset Convolution algorithm, yielding a similar improvements for the problem of computing the chromatic number in graphs of bounded * Partially supported by NCN grant N206567140 and Foundation for Polish Science.