Context. Several competing scenarios for planetary-system formation and evolution seek to explain how hot Jupiters came to be so close to their parent stars. Most planetary parameters evolve with time, making it hard to distinguish between models. The obliquity of an orbit with respect to the stellar rotation axis is thought to be more stable than other parameters such as eccentricity. Most planets, to date, appear aligned with the stellar rotation axis; the few misaligned planets so far detected are massive (>2 M J ). Aims. Our goal is to measure the degree of alignment between planetary orbits and stellar spin axes, to search for potential correlations with eccentricity or other planetary parameters and to measure long term radial velocity variability indicating the presence of other bodies in the system. Methods. For transiting planets, the Rossiter-McLaughlin effect allows the measurement of the sky-projected angle β between the stellar rotation axis and a planet's orbital axis. Using the HARPS spectrograph, we observed the Rossiter-McLaughlin effect for six transiting hot Jupiters found by the WASP consortium. We combine these with long term radial velocity measurements obtained with CORALIE. We used a combined analysis of photometry and radial velocities, fitting model parameters with the Markov Chain Monte Carlo method. After obtaining β we attempt to statistically determine the distribution of the real spin-orbit angle ψ. Results. We found that three of our targets have β above 90 • : WASP-2b: β = 153 • +11 −15 , WASP-15b: β = 139.6 • +5.2 −4.3 and WASP17b: β = 148.5 • +5.1 −4.2 ; the other three (WASP-4b, WASP-5b and WASP-18b) have angles compatible with 0 • . We find no dependence between the misaligned angle and planet mass nor with any other planetary parameter. All six orbits are close to circular, with only one firm detection of eccentricity e = 0.00848 +0.00085 −0.00095 in WASP-18b. No long-term radial acceleration was detected for any of the targets. Combining all previous 20 measurements of β and our six and transforming them into a distribution of ψ we find that between about 45 and 85% of hot Jupiters have ψ > 30 • . Conclusions. Most hot Jupiters are misaligned, with a large variety of spin-orbit angles. We find observations and predictions using the Kozai mechanism match well. If these observational facts are confirmed in the future, we may then conclude that most hot Jupiters are formed from a dynamical and tidal origin without the necessity to use type I or II migration. At present, standard disc migration cannot explain the observations without invoking at least another additional process.