Not to question the model, but by looking at the graph it seems to underestimate the value of current results. By this I mean your percentages were so low 3 weeks ago even when his current trends at that time for the season were putting him on pace to hit these numbers. Obviously no model is perfect, but if you're seeing massive spikes in data over a short period it means that it could probably use some fine tuning (and yes, I realize that 13 goals in 10 games and all the recent points play a big role in that too).