For those of you who read our Round 4 wrap-up, you will know that we looked at our margin errors after 4 rounds for each teams, which are shown below:
Rather amazingly we have predicted the Gold Coast Titans to within 5 points on average for each of the 4 games, and this is reflected in the fact that they’ve won us 3 bets from 3 and are by far and away our best performing team.
But what about the other end? Where do we need to improve, and fast?
Well it’s those Rabbitohs. The team who are making us look like the bunnies!
The season to date
Our margins on the Rabbitohs have been as follows thus far:
So whilst we have 3 of the 4 games tipped correctly (that first game against the Roosters was, well wrong) we are not exactly close on the margin, and it would be good to figure out why. It’s not even the presence or otherwise of Sam Burgess, as we are just as bad when he plays compared to when he doesn’t.
So what is the TipBetPro approach?
As you may have read before, based on our experience from other sports, we wanted to create a “player based methodology” which essentially provides less weighting to more traditional “team based” tipping features (eg form and ladder position) and a higher weighting to the 17 players that actually run on to the ground.
Now in general the approach appears to be working, both in the NRL and in the AFL, where we have several years of proven success. But not so for the Rabbitohs.
A more in-depth look
So this round sees the Rabbitohs taking on the Sea Eagles, who are actually our 4th worst team for predicting margins (average 15 out per game). Now both teams have won 2 and lost 2, so have similar “form” and “ladder” positions. Manly playing at home sees us giving them a marginal “team based” advantage for the round of +4 points.
Now of course our approach is player based – and this is where we weight our predictions. TipBetPro has ranked each player, and each player contributes to the overall result. The contribution is based on two main factors:
Experience – the more games a player has the more likely they will have a positive impact on the result – although the impact tails off after 100 games in our model. Is this too few games – does a 200 game player contribute more than a 150 and more than a 100 game player?
Player Attributes – this is very similar to a Super Coach type factor and in essence measures a player’s impact on the game – for rookies we assume an equivalent super coach score of around 30. Does this seem reasonable?
So based on the teams named (taking the first 17 named for now) this is what we have for Round 5:
The difference between the two teams is the key, as the TipBetPro ratings on their own don’t mean too much. So the players match ups have a near +8 point advantage to Manly, which combined with the +4 team based factors, has us predicting a +12 point win for the Sea Eagles.
But the book makers have the Rabbitohs favourites (1.65) compared to Manly (2.25). So one of us is wrong – now we hope it’s the bookmakers but we are always willing to learn and to tweak. As we noted we are not very good at predicting the Rabbitohs and it could well be we are overlooking a key factor. Are the Manly backs really 9 points better than the Rabbitohs, and even factoring in Sam Burgess is the Rabbitoh pack 4 points worse?
Now clearly there are other factors which impact on a game of NRL, not least lady luck and the ref (North Queensland for example against Brisbane!). But in our experience the main ones are experience and player ratings.
What do you do? And have we got our ratings right? Is Sam Burgess the best player in the game? Are Greg Inglis and Brett Stewart evenly matched? Does Feleti Mateo have Kyle Turner covered to the tune of nearly 3 points?
We’d be really keen to hear your thoughts – we promise to share them when we tweak the model!