TRAPPIST-1 also has sizable starspots, judging from its light curve in Figure 2 of the improved-masses paper. The star rotates with a period of around 3 days. Figure 3 in that paper shows how orbit fits were improved by adding the Kepler "K2" observations. It shows observed TTV's and calculated TTV curves from a large number of randomly-generated orbits. That random generation was a result of Markov-Chain Monte Carlo (MCMC) fitting, something that seems much like simulated annealing. Randomly change the parameters, and if they improve the fit, accept them, but if they don't, then accept them with probability exp(-(Enew - Eold)/T), where the E's are error values and T is a sort of temperature. For b to g, the new curves are well inside the old curves, meaning that the with-K2 mass estimates are both smaller and with smaller error bars than the without-K2 ones. My estimated amplitudes: b: 2 min, c: 2 min, d: 25 min?, e: 10 min?, f: 40 min, g: 30 min. Planet h has three distinct sets of TTV fits, with a few outlying fits. The most populous set having an amplitude of about 100 min. From the seven-planets announcement paper, transit durations are b: 36.40+-0.17 min, c: 42.37+-0.22 min, d: 49.13+-0.65 min, e: 57.21+-0.71 min, f: 62.60+=0.60 min, g: 68.40+-0.66 min, h: 76.7+2.7-2.0 min This may explain the error bars and scatter of the b and c TTV measurements. The scatter is much less for the outer ones.