2014 - Round 1: Half A Round's Better Than No Round At All

/

Welcome to the Wagers & Tipping blog for Season 2014.

This season, MatterOfStats offers 28 Head-to-Head Tipsters, 17 Margin Predictors, 8 Head-to-Head Probability Predictors, 1 Line Probability Predictor and 3 Wagering Funds each operating in a different TAB market. With that many opinions surely I'll be able to marvel at how prescient at least one of them has been most weeks.

But, while we're knee-deep in opinion providers, we're starved for contests for them to pass judgement on. Only four games will be decided in this weekend's season entree, the remaining five courses having been held off for another week. Quite why the League sees fit to start each season with this progressive-dinner style opening round every year puzzles me - are they worried we're all lacking in spectatorial match-fitness and couldn't stand up to a full round of fixtures?

WAGERS

The TAB Bookmaker has shown no such concerns for our welfare and has posted a full set of markets for all nine contests - all, somewhat surprisingly given the time between their posting and their resolution, carrying much the same overround. So, in this season's first break from tradition, I've placed wagers, or at least assessed the wisdom in doing so, on all nine contests, a full 11 days in advance of the conclusion of the most chronologically distant of them.

Here's what that lot looks like:

We've nothing in the opening game of the season, the Pies being considered too short for a Head-to-Head wager, insufficiently handicapped for a Line bet, and not expected to win by the Predictor that matters and so ineligible for a SuperMargin bet. (For details on the three Funds that are operating this year, you can refer to this blog. The only difference this year compared to last is in the Fund weightings within the Recommended Portfolio that are now 30% Head-to-Head, 50% Line, and 20% Margin.)

The head-to-head wager on GWS at $13 could not be fairly characterised as anything other than speculative, and represents by far the largest upside for the round. Should the Giants prevail over the Swans, 3.3c will be added to the Portfolio. A Roos victory by 1 to 9 points next weekend offers the next-best return at about 1.6c, while favourable results in any of four other contests would nudge the Portfolio up by a little over 0.5c.

The GWS v Sydney game also carries the largest downside, defeat by the Giants by 49 points or more threatening a loss of a little under 1c. In five other games the downside is about one-third of a cent, and in a sixth game, Gold Coast v Richmond, our wager is so small that the upside and the downside essentially round to 0.

Visually, those upsides and downsides all appear in the following Ready Reckoner:

HEAD-TO-HEAD TIPS

There's been talk in the paper this week hailing the 2014 AFL season as likely to be one of the most competitive in recent memory. I don't think it will be this - unless recent memory extends little beyond the past few years - and, for the most part, neither do the MatterOfStats Head-to-Head Tipsters feel that way about much of the first round of the season. 

(To understand the Tipsters and Predictors on MatterOfStats you could refer in the first instance to this blog, which will point you to other relevant sources of information.)

In six of the contests there's unanimous or nearly unanimous support for the favourite, and in a seventh - the Roos v Dons clash - there are only eight dissenting voices. Only for the Carlton v Port Adelaide, and for the St Kilda v Melbourne matches do we find any serious debate evidenced by double-digit support for the minority selection.

Home Sweet Home has rapidly reassumed its role as the universal dissenter, its tips being different from that of a randomly chosen Tipster almost half the time. Combo_NN1 is the next-most contrarian, and our new Tipster, based on the ChiPS C-Marg predictor, is third-most different.

MARGIN PREDICTIONS

The patterns of disagreement, and its general absence, are similar amongst the Margin Predictors. Combo_NN1 finds itself the lone Predictor with a plus sign in front of its margin prediction for the Collingwood v Fremantle, and Gold Coast v Richmond games and is also the outlier in the Kangaroos v Essendon game. As such, it's not surprising to find that it has the second-largest mean absolute difference (MAD) from the all-Predictor average predictions across the nine games.

The largest MAD belongs to C-Marg, whose confident prediction of a near 7-goal win for the Saints is a good indicator of its independent perspective. It's the only Margin Predictor that completely ignores the opinion of the TAB Bookmaker.

Measured by the size of the standard deviation in the predicted margins, the GWS v Sydney game is the most difficult to predict. The St Kilda v Melbourne game has a standard deviation of a similar size though its is largely a result of the outlying C-Marg prediction.

The games most likely to be close-run affairs are Collingwood v Fremantle, where the average margin prediction is for a 4-point Fremantle victory even though all but Combo_NN1 foresee a Freo win, Carlton v Port Adelaide (on average tipped to be about a 5-point Blues win), and Kangaroos v Essendon (tipped, on average, to be a Roos win by just over 2 points).

PROBABILITY PREDICTORS

Three contests have at least one Head-to-Head Probability Predictor on the opposite side of the 50% mark from the majority. In the Carlton v Port Adelaide match, Pro-Pred and the H2H Predictors are sided with Port Adelaide, while in the Kangaroos v Essendon contest this trio is joined by WinPred, here in predicting a Dons victory.

In the St Kilda v Melbourne clash, C-Prob stands alone in defiant and confident support for the Saints.

C-Prob is another of the new forecasters from the ChiPS pack, which I introduced in this blog earlier this year. All of these forecasters are underpinned by an alternative team rating system, which I've dubbed VSRS and written about here. The attraction of the VSRS and of the ChiPS forecasters are their relative simplicity.

VSRS ratings are designed to measure the difference in team abilities in terms of points scoring, so the base expected difference in scoring when Team A plays Team B is equal to the difference in their VSRS ratings. To derive a predicted margin for an actual game between two teams we add to this difference in VSRS ratings:

  • an adjustment for the difference in the teams' recent form (as measured by the change in their team ratings over the two most-recent rounds in the current season)
  • an empirical estimate of the home ground advantage for the home team at the venue on which the game is being played
  • a fixed allowance for any game deemed to be an Interstate clash (where the home team is playing in its home state and the away team is not)

You can see how this works for Round 1 games in the top half of the following table:

In Friday's Pies v Dockers clash, the Pies enjoy a narrow ratings advantage (+0.9 points) and benefit from the interstate nature of the game (+5.6 points), but they suffer from having typically performed poorly as the home team at Docklands (-9.4 points) so much so that the predicted result is a narrow win by Fremantle.

You can see here that Chi-Marg's confidence in the Saints' chances stems from the substantial ratings differential it believes separates them from the Dees (+24.7 points), and from the substantial Home Ground Advantage enjoyed by the Saints when playing at home at Docklands (+16.3 points).

The lower half of the table describes C-Prob's probability predictions, which are derived from C-Marg's margin predictions by assuming that these margins are drawn from a Normal distribution with a zero mean and a standard deviation of just under 34 points.

I'll finish by pointing out that the Line Fund algorithm, whose line betting probability estimates appear at the base of the table before last, is tipping the home team to emerge victorious on line betting in only the GWS v Sydney clash. Accordingly, this is the only contest in which we've a line bet.