Why are your performance predictions so terrible?

 I'm curious as to why my Strava performance predictions are consistently terribly wrong. For example, I just ran a marathon on Sunday.  It was totally demoralizing to see at the end of my training as I was tapering that Strava was predicting a 4:43 marathon.  I knew this was wrong and way off with my training but it was still depressing to see. (maybe we need a function to hide it if we don’t want to see it (or maybe that exists)  Meanwhile, my Suunto watch was predicting a 3:20 marathon based on the exact same data. That is an insane difference based on the same data.  I knew Suunto's numbers were too fast since my PR is a 3:26 and I knew my training was not up to par to beat it.  In the end, I ran a 3:34 marathon which was well within my range of expectations. But your predictions were so dramatically off that I don't really understand it and it doesn't make me think your algorithms are reliable at all. And this is after repeatedly complaining on the function that I thought the predictions were too slow (so they don’t even get adjusted if you make that feedback).  Meanwhile, your 5k predictions (21:50) are pretty bang on while any longer distance is off. Even now, having just run a 3:34 marathon, you are only predicting my performance at 3:52. Given the data you have, something is off and needs fixing. Or maybe this is just a useless AI generated tool that is terrible (although some of the other AI generated info you have has been quite good).