Service level agreements (SLAs), or contracts, have an important role in web services. They define the obligations and rights between the provider of a web service and its client, about the function and the Quality of the service (QoS). For composite services like orchestrations, contracts are deduced by a process called QoS contract composition, based on contracts established between the orchestration and the called web services. Contracts are typically stated as hard guarantees (e.g., response time always less than 5 msec). Using hard bounds is not realistic, however, and more statistical approaches are needed. In this paper we propose using soft probabilistic contracts instead, which consist of a probability distribution for the considered QoS parameter--in this paper, we focus on timing. We show how to compose such contracts, to yield a global probabilistic contract for the orchestration. Our approach is implemented by the TOrQuE tool. Experiments on TOrQuE show that overly pessimistic contracts can be avoided and significant room for safe overbooking exists. An essential component of SLA management is then the continuous monitoring of the performance of called web services, to check for violations of the SLA. We propose a statistical technique for run-time monitoring of soft contracts.