Super 14 Rugby Predictions
How It Works
I have adapted a system that I have been improving over
the last 30 years after more than 20,000 predictions in a
variety of sports to making predictions for Super 14
rugby. I start with an offensive, defensive and total
rating for each team from the previous season and then
adjust that rating using season data. Each offensive
rating is adjusted, based on
Points scored by the team
Opponent defensive rating
Home advantage
Each defensive rating is adjusted depending on
Points scored against the team
Opponent offensive rating
Home advantage
Total rating depends on offensive rating minus defensive rating.
To predict score difference for the next game, I use
Weighted difference of total ratings
Home advantage
To predict the home team score I use
Score difference
Home team offensive rating
Away team defensive rating
Home advantage
To predict the away teams score I use
Score difference
Away team offensive rating
Home team defensive rating
Home advantage
To predict total score I add the home team score to
the away team score.
I also predict the probability that each team will
win, using
Score difference
Score error standard deviation
Normal distribution
I show fair decimal odds for each game. Fair decimal odds
are found by taking 1/probability. For example, if the
probability of a home with is 0.5, the fair decimal odds
are 1/0.5 or 2.0 (a bet of $1 will return $2, the original
$1 and a profit of $1). If the odds of a win are 0.25, the
fair decimal odds are 1.0/0.25 or 4.0.
The fair decimal odds can be used for money line betting.
Score difference can be used for handicap betting and for
betting on the margin of victory. See the section below
titled Gambling Methods and Results.
Home Advantage in Super 12/14
Previous to the 2006 season, the 12 teams formed Super 12 competion.
Staring with the 2006 season, two teams were added forming Super 14
competition. Currently there are five teams from South Africa
(Bulls, Cats, Cheetahs, Sharks and Stormers), five teams from
New Zealand (Blues,Chiefs, Crusaders, Highlanders and Hurricanes)
and four teams from Australia (Brumbies, Reds, Waratahs
and Western Force). I use
two separate home advantages, one for domestic competition
and one for international competition. Here is a summary
of overall scoring and home advantage. The total score
tends to be higher for international competition as does
the home advantage (due to travel fatigue, as one of
several factors).
Scoring and Home Advantage (6 Seasons, 2001-2006)
| Home
Competition| Points| Advantage
-----------------------------------------
Domestic | 49.4 | 4.6
-----------------------------------------
Internat. | 53.6 | 6.6
----------------------------------------
All | 52.3 | 6.0
Accuracy of Picking the Winner
If you refer to my website
Other Predictions
You'll find that for 13,209 predictions of sports
having few draws (ties) I picked the winner 0.70
of the time. I applied my prediction method to the
2001-2006 seasons of Super 12/14 and got the following
accuracy.
2001-2006 Accuracy of Picking the Winner
Season |Games |Right |Wrong |Accuracy
---------------------------------------
2001 | 69 | 50 | 19 | 0.725
---------------------------------------
2002 | 69 | 47 | 22 | 0.681
---------------------------------------
2003 | 69 | 44 | 25 | 0.638
---------------------------------------
2004 | 69 | 40 | 29 | 0.580
---------------------------------------
2005 | 69 | 46.5| 22.5| 0.674
---------------------------------------
2006 | 94 | 61 | 23 | 0.649
---------------------------------------
Total | 439 | 288.5| 150.5| 0.657
Accuracy of Picking the Probability
Below you'll find probability data for picking the
favorite for the 2001-2006 seasons of Super 12/14.
Of course, if you want the non-favorite probabilities,
just take one minus probabilities for the favorite. I
show each probability range, the number of games, the
predicted probability and the actual accuracy (how
often the favorite won). There is close agreement.
As more games are played, the predicted and actual
figures should come even closer together
2001-2006 Accuracy of the Predicted Probabilities
for the Favorite
|Predicted |Actual
Range |Games|Probability|Probability
--------------------------------------------
0.50 to 0.59| 133 | 0.550 | 0.478
--------------------------------------------
0.60 to 0.69| 114 | 0.650 | 0.610
--------------------------------------------
0.70 to 0.79| 103 | 0.746 | 0.762
--------------------------------------------
0.50 to 0.59| 71 | 0.841 | 0.831
--------------------------------------------
0.50 to 0.59| 18 | 0.932 | 1.000
--------------------------------------------
All | 439 | 0.685 | 0.657
Gambling Methods and Results
Each week, predictions are provided like the example that below,
taken from the 2004 Super 12 season.
***************************************************************
Predictions for Week 4 of 14 March 12-13 2004 6 Games
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Abbreviations: HW=Home Win, AW=Away Win
HS= Home Score, AS = Away Score
Score Difference= Predicted HS - Predicted AS
Fair odds = 1/Predicted Probability
A = Australia N = New Zealand S = South Africa
There are eight predictions below for each match.
The data are organized in two lines per match.
Home Team Away Team Prob. Prob Score
HW Aw Difference
Fair Fair HS AS Total
Odds Odds Score
HW AW
----------------------------------------------
Highlanders(N) Sharks(S) 0.67 0.33 8
1.49 3.06 21 13 34
Hurricanes(N) Cats(S) 0.78 0.22 14
1.27 4.64 31 18 49
Reds(A) Blues(N) 0.57 0.43 3
1.74 2.35 31 28 60
Bulls(S) Brumbies(A) 0.51 0.49 1
1.94 2.06 35 34 69
Stormers(S) Waratahs(A) 0.43 0.57 -3
2.31 1.76 31 34 64
**************************************************************
I believe that money line betting (on the winner) and betting
on margin of victory each have a good chance for making a profit
against published odds. For money line betting, here were the
offerings from TAB New Zealand for two of the matches above.
Team Payoff to win (decimal odds)
Sharks 4.00
Reds 2.65
Let my predicted probability of a team winning be called p,
Let my fair decimal odds (1/p) be called Rf while the bookie's
decimal payoff odds are called R. Put another way, my predicted
probability of winning is 1/Rf. When the payoff odds are 4.00,
as for the payoff odds for the Sharks to win, a successful bet
of 1 unit returns the original 1 unit bet plus 3 units of
profit. The integer odds would be 3:1. You could accept a bet
whenever the payoff odds are more than my fair odds. My fair
odds for the Sharks to win were 3.06 while the payoff odds
were 4.00, suggesting that a bet should be placed supporting
the Sharks. The expected overlay was R/Rf - 1 or
4/3.06 - 1 = 0.307, an expected profit or overlay of 30.7%.
As to the Reds, the overlay is found by taking
2.65/1.74 -1 = 0.523, an expected overlay of 52.3%.
How much should be bet? One method is to decide upon a comfort
level and bet that same amount on each overlay. Another method,
the Kelly method, allocates more money for bets that are more
likely to be won and less money for bets that are less likely
to be won. The equation for "full Kelly" is that (R/Rf-1)/(R-1)
(the expected overlay divided by the realized overlay) is the
fraction of the bankroll to bet on an overlay when (R/Rf-1)
is positive. The Kelly method was created to provide the best
exponential growth of the starting bankroll when each bet is
settled one at a time. In league competition, many bets may be
made simultaneously before the first bet is settled so using
1/2 or less of the Kelly formula may be a good idea for
conservative betting. If "1/2 Kelly" is used for an overlay
supporting the Sharks with a bankroll of 1000, the bet could be
1000*.5*(4/3.06-1)/(4-1) or 51. For the Reds one could bet
1000*.5*(2.65/1.74-1)/(2.65-1) or 159. Both bets actually won.
An agency such as TAB New Zealand offers odds on the margin of
victory. For TAB New Zealand, 7 odds are posted, including the
favorite to win by 13 or more, the favorite to win by 1-12, a
draw, the underdog to win by 1-12 and the underdog to win by 13
or more. I provide a table converting each predicted score
difference into the 7 odds for that match. When the offered
payoff odds are more than my fair odds, a bet could be placed.
Among the matches above, I predicted the Reds to win by 3 over
the Blues. That corresponded to fair odds Rf = 3.54 for the
Reds to win by 13 or more compared to corresponding payoff
odds of 7.00. That bet has an expected overlay of 7/3.54 -1
or 97.7%. The "1/2 Kelly" bet using a bankroll of 1000 would be
1000*.5*(7/3.54-1)/(7-1) or 81. The bet was successful returning
7*81 or 567 for a profit of 486 or nearly 1/2 of the bankroll.
Below is a summary of 119 bets placed against the
TAB New Zealand offerings over two seasons using 1/2 Kelly
to determine the amount of each bet. The average return on
the gross bet is 17-211%.
2003 2004 2004
Money line Money line Win Margin
Start Bankroll 1000 1000 1000
Bets Made 29 32 35
Bets Won 17 17 10
Amount Bet 3454 5188 4288
Amount Won 4119 6248 5151
End Bankroll 1665 2060 1863
Profit/Amount Bet 19% 21% 20%
2006
Money line
Start Bankroll 1000
Bets Made 23
Bets Won 11
Amount Bet 1882
Amount Won 2200
End Bankroll 1318
Profit/Amount Bet 17%
The link below is to a spread sheet showing each of the
96 bets summarized above. In the spread sheet
Sheet 1 2003 Money line
Sheet 2 2004 Money line
Sheet 3 2004 Margin of Victory
Sheet 4 2006 Money line
Super 12 Betting Results
Guinness Premiership Predictions
The Future
I look forward to providing predictions for the 2007
season of the Super 14. Your comments are always
welcomed. To receive predictions, go to
How To Get Predictions