MoSSBODS and MatterOfStats' Funds for 2016

I mentioned in the previous blog here on Wagers & Tips that the MoSSBODS Team Rating System (V2.0) will be used to power a Head-to-Head, a Margin and a Probability forecaster in 2016. 

All three of those forecasters will, I expect, perform well, so much so that I've decided to use MoSSBODS' opinions to inform, exclusively, all three of the Funds that will operate in 2016: a Head-to-Head Fund, a Line Fund, and an Unders/Overs Fund. ('Eggs - meet Basket. What's that? Yes, there is only one ...')

This decision means that, for the first time in MoS' history, bookmaker data is not being used as an input into the algorithms whose outputs are, in turn, used to make wagers. Put another way, we can decide the terms on which we'd be willing to wager in advance of the posting of any of betting market. For me that is, in roughly equal parts, a source of pride and a source of terror.  

In this blog we'll look at how each of the MoS Funds will operate and at the supplementary information I'll be providing about MoSSBODS forecasts each week. Firstly, to the Funds.

HEAD-TO-HEAD FUND (Weighting 40%)

The Head-to-Head Fund will place a head-to-head bet equal to 2.5% of the original Fund whenever the price on offer for the Home team is sufficient to imply at least a 1% edge given MoSSBODS' assessment of the Home team's victory chances. The 1% threshold provides some minimal buffer against slight inaccuracies in MoSSBODS' probability estimations. 

(So, for example, if the Home team was assessed as a 72% chance, then the minimum acceptable price would be 1.01/0.72 or about $1.40 since 72% x 1.4 is about 1.01.)

I considered also imposing a maximum price ceiling as a way of mitigating the risk that MoSSBODS might significantly overrate a Home team's chances due to circumstances of which it is unaware, such as the absence of a  number of key players. In the end though, since we're only level-stake wagering, and since history suggests that such a cap would be profit-reducing, I opted to run without a cap. 

More significantly, I also investigated removing the traditional MoS constraint that prohibits Head-to-Head Funds from wagering on Away teams. The evidence in favour of such a change was equivocal, with profitability heavily dependent on the particular Bookmaker data used and on the seasons considered. What eventually persuaded me to maintain the prohibition was the modelled loss-making from wagering with the TAB on Away teams in each of the seasons 2012 to 2014. There was, however, a healthy profit in TAB Away team wagering in 2015, and more modest profits in recent seasons when using averaged data from other bookmakers.

My current hypothesis is that 2015 saw the demise of the systematic underpricing of Home teams, especially on the TAB, possibly from an influx of more-informed bettors. If that's right, and MoSSBODS proves capable of earning money from Away teams too in 2016, MoS Head-to-Head Funds will start wagering on Away teams from next year. (For the record, had I wagered on Away teams in 2016, I would have used the same 1% threshold that I'm using for Home teams. In the background, I'll be tracking the return to this form of wagering.)

Using the last few seasons as a rough guide, I'd expect this Fund to wager on 35 to 40% of the games, on which basis the expected turn will be about 2. Across the previous 10 seasons, the Fund averaged about a 17% ROI and made losses in only one season, 2006. It also, however, made its smallest profit in 2015, with an ROI of only about 6%. As always, please treat this historical performance narrative with the in-sample skepticism it merits. I wouldn't be writing about it if it mirrored the score from "Red Ink: The Musical".

LINE FUND (WEIGHTING 40%)

The Line Fund will place a line bet on the Home team proportional to the size of its measured 'advantage', whenever the handicap offered by the TAB bookmaker is assessed by MoSSBODS as being in error, in the Home team's favour, by more than 2 points. This small threshold, much as the 1% edge used for the Head-to-Head Fund, provides some protection against miscalibration errors.

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

Bet Size = max(5%, (Assessed Margin + Handicap Offered)/600) 
provided Assessed Margin + Handicap Offered > 2

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)/600 = 2.1% of initial Funds.

Allowing the Line Fund to wager on Away teams was an even nearer-run thing than it was for the Head-to-Head Fund, with modelling revealing a profit even on the TAB in 6 of the past 10 seasons, including an imposing 21% ROI in 2015, exceeding that estimated from Home team wagering in the same season. Modelling using the averaged data from a range of other bookmakers over the past three seasons also suggested that Away teams might be wagered on profitably within that group. Of concern though was the large loss registered when betting on Away teams with the TAB in 2014. Again, I wonder if 2014 marked the end of the systematic Home team mispricing era.

(For the record, had I wagered on Away teams in 2016, I would have used the same 2-point threshold that I'm using for Home teams and the same methodology for translating estimated advantage into bet size. In the background, I'll be tracking the return to this form of wagering too with a view to Away team line wagering in 2017.)

Using the last few seasons as a rough guide, I would expect the Fund to wager on 45 to 50% of the games with an average bet size of about 1.75% of the Fund, on which basis the expected turn will be about 1.8. Across the previous 10 seasons, the Fund averaged about a 19% ROI and made profits in every season. The same caveats apply.

UNDERS/OVERS FUND (20%)

This marks the first season in which MoS will wager in the Unders/Overs market, where bettors are asked to decide whether they expect the aggregate score in a game to be higher than ('over') or lower than ('under') some bookmaker-set total. Usually, as in the line market, even money will be offered for both the "over" and the "under" bet, and the total set by the bookmaker will include a half point to prevent ties.

The data available to me on this form of wagering extends only to the last three seasons, and is not data I've personally captured, so for reasons of conservatism I've weighted this Fund at only one-half that of the other two Funds (ie at 20%).

Analysing the available data alongside MoSSBODS' forecasts suggests that "unders" wagers are appropriate when MoSSBODS' forecast total is more than 12 points less than the total set by the Bookmaker, and that "overs" wagers are appropriate when MoSSBODS' forecast is more than 2 points greater than the set total. Despite the larger difference required to trigger an "unders" as compared to an "overs" wager, "unders" bets should be expected about 5 times as often as "overs" bets, possibly reflecting a tendency for MoSSBODS to be conservative in its totals forecasts, regressing them towards the mean.

Combined, we should expect this Fund to wager is about 55% of games, or around 110-120 games. Each of the Fund's wagers will be 2% of the initial Funds, so we can expect a turn of around 2.5. History suggests an ROI of about 8% is plausible.

In summary then, here are the facts about each of the three MoS Funds that will operate in 2016.

View fullsize

One other thing worth pointing out that is different from previous seasons is that all Funds will wager commencing in Round 1, and that bet-sizes will not vary based solely on the portion of the season in which they're made.

DETAILS ABOUT MoSSBODS PREDICTIONS

Since MoSSBODS will have such a central role in MoS for 2016, each week I'll be providing extra details about its forecasts in a form shown in the annotated diagram that follows.

View fullsize

Let's step through the data for the long-awaited Springfield v Shelbyville clash, which appears as the first row in the table. The first pair of numbers tells us that MoSSBODS rates Springfield as a 1.8 Scoring Shot (SS) better-than-average defensive team, and a 0.4 SS worse-than-average offensive team. In other words, MoSSBODS would expect Springfield to concede about 2 fewer SS than the all-team average for a game, and to register about 0.5 fewer SS than the all-team average (assuming, in both cases, that Springfield were playing an average team at a neutral venue).

Shelbyville is rated a considerably worse team by MoSSBODS, with a (2.4) SS defensive rating, and a (3.0) SS offensive rating. 

Both Springfield and Shelbyville will have Venue Performance ratings for the particular ground on which the contest is taking place. These ratings, netted here and provided as a Net Venue Effect, have Shelbyville enjoying a small, 0.3 SS advantage over Springfield on this ground.

The Expectations section in the diagram begins with the Expected Home Team Victory Margin in Scoring Shots, calculated as Home Team Def Rating + Home Team Off Rating - Away Team Def Rating - Away Team Off Rating + Net Venue Effect. We see then that MoSSBODS expects Springfield to register 6.5 more Scoring Shots than Shelbyville. This is converted to points (24) using the average Scoring Shot Conversion Rate from 2015, which was about 53.5%, at which Rate one SS is worth about 3.6 points.

MoSSBODS also provides expected scores for each of the two teams, here showing an expected 93-69 result. These team scores can be added to yield an expected aggregate score, here 162 points (note that there is rounding in all of these numbers, which is why they don't add.)

The expected Scoring Shot margin for the Home team can be used to derive a victory probability for them via the equation Probability = 1/(1+exp(-Expected Margin ^ 0.184)). For the Springfield game where the Expected Margin is 6.5 Scoring Shots, that makes the home team a 77% favourite.

Given that assessment, and imposing our requirement that an acceptable price provide at least a 1% edge means that we'd need a price of $1.32 or higher for Springfield to encourage the Head-to-Head Fund to wager. This number appears under the Min Home H2H Price heading.

The next column provides the mimimum start that the Home team would need to be receiving (or giving) in order for the Line Fund to bet. This minimum is equivalent to two points more than the negative of the expected Home team victory margin, which here is about -22 points. So, if Springfield were giving 21.5 points start or less, the Line Fund would be willing to wager.

At the far right of the table are the boundaries outside of which the Under/Over Fund would be willing to bet. If the total offered were 158.5 or less, the Fund would make an Overs bet; if it were 173.5 or more, it would make an Unders bet; if it were in the range 159.5 to 172.5, if would make no bet at all.

That's all for now on MoSSBODS and the MoS Funds for 2016. In a future post I'll explain how the forecasts of the other Head-to-Head, Margin and Probability Predictors will be presented each week.