Their data should have included RWC matches or tiers or both. Certain teams look quite good on paper; Japan have won about half their games (ever) but only won

Their data should have included RWC matches or tiers or both. Certain teams look quite good on paper; Japan have won about half their games (ever) but only won one match in the RWC (ever) before their "biggest-ever upset" victory against South Africa. I guess this proves the importance of making a good estimate. If something is obviously off with your result, it's worth checking your model.