Well after all my carping, it didn’t turn out to be a bad set of predictions at all. All my bets came through and eight out ten matches correctly called, and the two upsets (Racing Metro and Leicester) were in what were predicted to be close games.
More worrying maybe was the crush that the model seems to have on Leinster – anyone could see ahead of the game that 96% was too high an estimate for them to win, even if they did do so in the end. Leinster have been struggling to find their top game this season, while Wasps had just produced an impressive performance to smash Bath at home the week before. This points to two things. Firstly, although I have already tried to introduce changes like this, the model still needs a little more sensitivity to form – at the very least to introduce uncertainty. Secondly, and probably more importantly, is that even the best statistical model (which ain’t this one) will still need a human at the other end to interpret, adjust and correct the outputs for things the computer doesn’t know about.
So what did this weekend’s results mean for the knockout stages? As per usual, I’ve run simulations of all the remaining game, to get likely final tables for each pool (as well as probabilities for teams finishing in different positions). Unlike the Pro 12, one match is quite a significant proportion of any given pool, and the pools are also a lot closer than the normal league, so there is a lot more volatility as the games start to come through.
In Pool 1, Saracens are now expected to top the group ahead of Clermont, while Munster’s last minute heroics haven’t had much impact on their expected third place finish.
For Pool 2, it’s very much as you were , with much the same predicted table as last time round, with Leinster on top and Castres the whipping boys of the pool.
Pool 3 is also very similar to last week, the main change being Toulon expected to finish a point lower, largely due to a slightly weaker than expected result against Scarlets.
Pool 4, the most closely matched in terms of the teams, sees the most volatility. A very solid performance by Glasgow in putting away Bath catapults them into an expected second place finish in the group, though still missing out on the quarter finals by a very slight margin. Bath on the other hand are now expected to finish last in the group.
For Pool 5, Ospreys and Saints trade places at the top of the pool, while Racing Metro are still expected to come third despite a good win over Northampton.
If these results were to come true, the quarter finals would look like:
Toulon v Harlequins
Saracens v Saints
Leinster v Clermont
Ospreys v Toulouse
With still only one Irish team making the cut, has Bruce Craig got his dastardly way?