Soccer Predictions 

 

	For more than 30 years I have been using this system to
predict the result of games between two teams and the
probability that the predicted result will happen. First, I
calculate a rating for each team, based on all games
previously played that season. The rating depends on

     points scored minus points allowed in each game
     home advantage
     strength (rating) of each opponent

The rating at the end of each season carries over to the
start of each new season, except that in soccer each promoted team
moves to the bottom of each new division while each
relegated team moves to the top of each new division. This
is accomplished by giving each promoted or relegated team
the average rating of the teams replaced.

The expected point difference for the next game depends on

     weighted difference of the opponent ratings
     home advantage

The probability that a team will win, draw or lose the next
game depends on

     weighted difference of the opponent ratings
     home advantage
     consistency of the two opponents
     decision threshold for a home win, draw or away win

Before 1994 I focused on sports with very few draws, so
I only calculated the probability that the selected
team would win. In 13209 games, I predicted the actual
winner in 70% of the games in five sports as in the next
table

Sport                    Years          Seasons Games Actual
-----------------------------------------------------------
USA College Football     1970/71 to 1980/81  11   6137 0.72
USA Junior Coll. Foot.   1977 to 1979         3    536 0.70
USA Professional Foot.   1970/71 to 1980/81  11   2148 0.66
USA College Basketball   1972/73 to 1973/74   2   1926 0.69
Australian Rules Foot.   1980 to 1994        15   2282 0.68
------------------------------------------------------------
All Games                                        13209 0.70

The next table shows the accuracy of the predicted
probability that the selected team would win for 2282
Australian Rules Football games. For example, for Australian
Rules Football, when the selected team had a predicted
probability between 0.90 and 0.99 that it would win, the
selected team actually won 0.98 (98%) of the games.

Predicted Probability    Actual Result
-------------------------------------------
0.50 to 0.59             0.55
0.60 to 0.69             0.64
0.70 to 0.79             0.76
0.80 to 0.89             0.86
0.90 to 0.99             0.98

Notice how accurate the past predictions have been. I now
provide predictions for soccer football. In that football, 
about 28% of the games are drawn, so it is necessary to predict 
the probability of a home win, draw or away win and not just of a win. 

	I have been offering predictions for the English Premier 
League and for Italian Serie A. Few if any prediction services 
tell you just how accurate their predictions actually are. In this 
website you will find complete disclosure of the accuracy of my 
predictions. I have tabulated prediction results 
below for those two nations as well as predictions I formerly 
made for Germany and Norway. My web page for each nation 
(see link below) provides complete information about the 
predictions for each nation. I believe these predictions 
will enhance your enjoyment of the game and perhaps provide 
some help in gambling. Your suggestions are always welcomed at:
  • Raystefani@aol.com 
    

  •   Predicting Probabilities 
     The next table shows predicted and actual probabilities
for 4876 matches in England and Italy since I improved my 
system in 1998, including data from the 1998-99 season 
through the 2004-05 season. Data are organized into five ranges
each for home win, draw and away win predictions. The
smallest probability error tends to occur for the ranges
with the largest number of games, indicating that the
predictions are converging to the correct values and the
predictions are very accurate. For example, there were 2465
home wins predicted to be between 0.40 and 0.59 likely. The
average prediction was 0.49 likely while actually the home
team won in 0.50 of the games. There were 4455 draws
predicted to be between 0.20 and 0.39 likely. The average
prediction was 0.28 likely while actually there were draws
in the same 0.28 of the games. There were 2683 away wins
predicted to be between 0.20 and 0.39 likely. The average
prediction was 0.29 likely while actually the away team won
in 0.27 of the games. The average error for the 
fifteen ranges is only 1%.

(G = Games, Pr = Predicted, Ac = Actual)

Range        |  Home Win    |      Draw    |   Away Win
----------------------------------------------------------
             |  G    Pr   Ac|   G   Pr   Ac|   G   Pr   Ac
----------------------------------------------------------
0.00 to 0.19 | 117  .15  .18|  343 .17  .23| 1615 .14  .15
0.20 to 0.39 |1466  .33  .34| 4455 .28  .28| 2683 .29  .27
0.40 to 0.59 |2465  .49  .50|   78 .44  .31|  539 .46  .46
0.60 to 0.79 | 787  .68  .67|    0 .00  .00|   39 .65  .59
0.80 to 0.99 |  41  .86  .68|    0 .00  .00|    0 .00  .00
----------------------------------------------------------
             |4876  .47  .47| 4876 .28  .28| 4876 .25  .25


    • 

  •   Predicting Goal Difference 

     For each game, I also predict goal difference which is the
predicted difference between home team goals and away team
goals. There are gambling schemes for which goal difference 
predictions are used for money line betting 
(see the information on gambling below) or you may 
find the information to be of general
interest. The table below shows scoring tendencies for 
all of the 9908 games played 1994-95 through 2004-05. 
The home advantage is the average goal difference between 
the home team and the away team. The total number of goals 
scored is also shown. Notice that the average error between 
my predicted goal difference and the actual goal difference 
is less than 1/100 of a goal for each nation, my goal 
difference is within 0.05 goals per game.

Country | Games |  Home   |Total  | Goal Difference 
        |       |Advantage| Goals |   Error
-----------------------------------------------
England | 4262  |  0.42   |  2.62 | -0.02
Germany |  918  |  0.50   |  2.87 |  0.05
Italy   | 3440  |  0.50   |  2.63 |  0.01
Norway  | 1288  |  0.58   |  3.41 |  0.02
-----------------------------------------------
        | 9908  |  0.47   |  2.75 |  0.00


    • 

  •   Predicting Home Wins, Draws and Away Wins 

	It is also interesting to see how often my choice
with the highest probability actually happens. That is,
if a HW, D or AW is most likely, how often does that actually
happen? The table below provides an answer to that question.
If someone were to guess randomly among the three choices of
HW, D and AW, the person would be correct in 33 games of 100.
Notice that I am correct in 50 games of 100 which is 17% better
than random chance. My second highest probability choice is correct
in 26% of the games so my first or second highest probability 
choice indicates the correct outcome in 76% of the matches.

Country | Games |Highest Prob.
        |       |Choice Correct
-------------------------------
England | 4262  |  0.49   
Germany |  918  |  0.48   
Italy   | 3440  |  0.50   
Norway  | 1288  |  0.53   
-------------------------------
        | 9908  |  0.50