Matter of Stats

View Original

The MoS Tipsters, Predictors and Funds for 2017

If history chooses to repeat one of the themes of 2016, please let it be the largely unembarrassing performances of MoS' bookmaker-input free Tipsters and Predictors, and the resultant profitability of the MoS Funds. (I'd wish for something grander on the global stage, but I suspect that's just courting disappointment.)

On the Tipster and Predictor front, this year sees the retirement of ENS_Linear_MoSSBODS and ENS_Greedy_MoSSBODS, both of which last year failed to provide anything much different from MoSSBODS itself. That's perhaps not entirely surprising given that their only input was MoSSBODS' predictions, but I had hoped for a little more divergence in their opinions nonetheless. I didn't get it - at season's end, less than 6 goals separated them both from MoSSBODS in terms of aggregate absolute errors.

ENS_Greedy and ENS_Linear suffered from a similar shortcoming and finished the season separated by less than a goal, so only one of them will return for the new season. That'll be ENS_Linear, and it's been rebuilt entirely using only the data for seasons 2006 to 2017 and just three base learners:

Each learner regresses home team margin on home team bookmaker probability, home team handicap, home and away team prices, MARS ratings, team venue experience values, team form (measured by recent MARS rating changes), and the games interstate status.

Three new Tipsters and Predictors will spring from the newly-created MoSHBODS Team Rating System, which will provide a Head-to-Head Tipster, as well as a Margin and a Probability Predictor in 2017. Its margin prediction will be the difference between its home and away team score predictions plus the 2-point bias adjustment, and its head-to-head tip will be the home team if this adjusted margin is positive, and the away team otherwise.

Its probability prediction will be derived from its adjusted margin prediction using the following logistic formula:

Estimated Home Probability = 1/(1+exp(-(Expected Adjusted Margin * 0.0428774)))

The other Team Rating Systems, ChIPS and MoSSBODS will both continue to provide head-to-head, margin and probability forecasts in 2017.

ChiPS has been reoptimised for the new season, based on the mean absolute error (MAE) over the period 2007 to 2016, while MoSSBODS will run with the same parameter values that it used in 2016 (excepting, of course, the Venue Performance Values, which update after every game in the manner described in the blog post just linked).

In total, there will be nine Head-to-Head Tipsters, 10 Margin Predictors, and six Head-to-Head Probability Predictors for the 2017, as listed in the table below.

(You can find out a little more about some of the forecasters in this blog from last year.

As well, if you're new to MoS, this post from 2015 will give you some idea about how it all works here on MoS, although many of the statistical models discussed there are no longer in use.) 

So, how will the MoS Funds operate this season?

Well, again there'll be a Head-to-Head, a Line and an Over/Under Fund, more details of which appear below.

HEAD-TO-HEAD FUND (WEIGHTING 20%)

Last year, head-to-head wagering was lucrative for about one-third of the rounds, roughly break-even for another third, and diabolical for the remainder. The final result for the Head-to-Head Fund was a loss of just over 9% on a 2.3 turn. Only 42 of 93 bets were collects.

In retrospect, I think that was partly due to my calibrating MoSSBODS margin to probability formula using data from the entire V/AFL history rather than just from the modern era.

This year the Fund's wagers will be guided by a new algorithm, MoSHBODS, whose margin to probability formula is calibrated in exactly that manner. As this is MoSHBODS' first time in the role, and because I'm not entirely certain that the value in the head-to-head market hasn't disappeared forever, the Fund will carry only a 20% weighting.

Also, despite lingering nervousness on my part, the Fund will be free to wager on home and on away teams during the course of the season. Hopefully that will increase the likelihood of finding the more limited opportunities for value that do remain.

The Fund will fractional Kelly bet a proportion of initial funds equal to:

1/7 * (Price x Estimated Probability - 1)/(Price-1)

provided that Price x Estimated Probability > 105%. In other words, we require a minimum 5% overlay before we'll wager.

I expect that the Fund will make about 130 bets across the season (a little over half of them on home teams), with an average size of about 1.9%, which gives a turn of about 2.6. A reasonable target for this Fund would be a 5% ROI, which would provide a 13% return on funds.

LINE FUND (WEIGHTING 40%)

The Line Fund will again be informed by MoSSBODS this year, and it too will be allowed to wager on home and on away teams during the course of the season.

It will be more conservative than last year, however, and will place a line bet on a team proportional to the size of its measured 'advantage', only when the handicap set by the TAB bookmaker is assessed by MoSSBODS as being in error by more than 8 points. 

Bet size, as a percentage of initial Funds, will be determined as:

Bet Size = (Assessed Margin + Handicap Offered)/300
provided Assessed Margin + Handicap Offered > 8

So, for example, if the Home team is giving 12.5 start and MoSSBODS assesses them as winning by 25 points, the bet would be (25 - 12.5)/300 = 4.2% of initial Funds.

Instead, if the Away team is receiving 26.5 start and MoSSBODS assesses them as losing by only 12 points, the bet would be (26.5 - 12)/300 = 4.8% of initial Funds.

Using the last few seasons as a rough guide, I would expect the Fund to wager on about 50% of the games (a little over half the time on home teams) with an average bet size of about 2.4% of the Fund, on which basis the expected turn would be about 2.5.  An ROI of 5-7% would seem to be a reasonable target and would provide a 13-18% return on funds.

OVERS/UNDERS FUND (40%)

There are two changes for the Overs/Unders Fund this season: a doubling of its weighting to 40% and the use of a 6 point minimum overlay in both directions.

It will continue to use MoSSBODS for its wagering advice.

We should expect this Fund to wager in about 60% of games, or around 120-130 games across the season. Roughly three-quarters of those wagers are likely to be unders bets. Each wager will be 2% of initial Funds, so we can expect a turn of around 2.4 to 2.6. History suggests a target ROI of about 4% is reasonable, which would yield about a 10% return on funds.

For now, that's all. I'm still investigating online wagering options for the Funds, though, whichever I choose, I will continue to use TAB prices for model inputs since that is the data those models have been trained on.