The Premiere League opens this weekend. In the weeks leading up to this event all of our attention has been focused on selecting the ultimate FPL squad that will provide a large point haul and shoot us to the top of our mini leagues.
Beyond this pursuit, there is another competition available to all registered site managers. It is the Prediction Game, which provides us an opportunity to compete against one another in a friendly atmosphere to determine who is the best prognosticator. The game awards points based on wins, draws, exact scores, etc, and is very competitive.
Let me lay out a little background about how all this came about for any newcomers to the site. I had been searching for online prediction sites to assist in making decisions concerning my entries into the Prediction Game. A few other managers were also into this, and we started to exchange site names that we visited. We were about 9 weeks into the season when ChorleyRocks asked if I would be interested in writing a weekly article, using statistics to consolidate the predictions from all the online sites available. I agreed and have had a blast learning how to put together a helpful article each week. It was extended to the MLS, once the EPL season was over. Something to keep busy over the summer. “An idle mind is the Devil’s workshop.”
The statistical analysis involved in this is not difficult; it is basic MEAN & MODE. The predictions used must be posted by Wednesday of each GW in order for the analysis and writing to be available on Thursday evening. This is the deadline as there will be matches on Fridays this season.
In addition to the site prediction data, I began late last season to experiment with Poisson Distribution Models. I am currently running 2 Models for MLS (1 using 2015 data & 1 using 2016 data). I plan on doing the same this year with EPL, but will probably expand it to 3 Models (1 using 2015/16 data, 1 using 2015/16 data adjusted for promoted teams & 1 using 2016/17 data once all teams have played a home & away match). The limited experiment last season yielded results very similar to the prediction sites. A benchmark of 40% correct win/draw predictions was established last year, and that appears to be what is expected from these sites. They also have a 10% exact score average.
For the next 12 weeks, the article will provide predictions and reviews of results, for both EPL & MLS. Now that I have bored most of you to death with the details, it is time to illustrate what these articles will present to you each week. The first Table will contain information that may be of interest to those playing the Prediction Game.
The statistical predictions for EPL GW1 are –
|Burnley||v||Swansea||1.1-0.9||1-0||1-1 or 1-0|
|Chelsea||v||West Ham||1.7-0.9||2-0||2-1 or 2-0|
|Crystal Palace||v||West Brom||1-0.6||1-1||1-1|
|Middlesbrough||v||Stoke||1-0.8||1-0 or 1-1||1-1 or 1-0|
A review of the sites posting predictions, as of this date, reveals the following –
- Man City is a unanimous pick to top Sunderland.
- Man United, Arsenal, Leicester & Southamptom are heavy favorites to win.
- Chelsea is favored over West Ham.
- Burnley could either win or draw with Swansea.
- Crystal Palace are leaning toward a draw, but could win.
- Tottenham are 50/50 to win or draw vs Everton.
- The Middlesbrough vs Stoke match favors a draw.
The next portion of the article deals with MLS.
The statistical predictions for MLS GW23 are –
|NY Red Bulls||v||Montreal||1.8-0.5||1-0||2-1 or 1-0|
|Columbus||v||NY City||1.5-1.4||2-2||2-1 or 2-2|
|Dallas||v||Sporting KC||1.8-0.1||1-0||2-0 or 1-0|
|Houston||v||Toronto||0.5-1.1||0-1||1-1 or 0-1|
|Los Angeles||v||Colorado||1.8-0.5||1-0 or 2-0||1-0 or 2-0|
|Chicago||v||Orlando||1.4-0.8||1-0||1-1 or 1-0|
|Seattle||v||Real Salt Lake||1-0.8||1-0||1-1 or 1-0|
A review of the sites posting predictions reveals the following –
- Dallas is a unanimous pick over Sporting KC.
- NY Red Bulls, New England & Los Angeles are heavy favorites to win.
- Vancouver may win vs San Jose, but could end in a draw.
- Toronto are a slight favorite over Houston.
- Chicago could win or draw vs Orlando.
- The DC vs Portland match leans toward a draw.
- Columbus vs NY City & Seattle vs Real Salt Lake provide no clear cut tendencies, the matches could result in wins for any team, or they could draw.
The top prediction sites from MLS GW22 were –
|# of Predictions||# correct wins/draws||# exact scores|
|Poisson Distribution Model 2||10||6||1|
|Poisson Distribution Model 1||10||5||0|
9 of the 10 sites met the 40% correct benchmark which was established previously. The number of exact scores seems to be better the last 2 weeks.
The predicted vs actual score results from MLS GW22 are –
|Home||Away||Predicted Score||Actual Score|
|Toronto||Real Salt Lake||2-1||1-0|
|San Jose||NY City||1-1||0-0|
|DC||Philadelphia||2-1 or 1-1||2-2|
|Montreal||Houston||2-1 or 3-0||1-0|
|Toronto||New England||2-0 or 3-0||4-1|
|Real Salt Lake||Chicago||1-0||3-1|
|Portland||Sporting KC||1-1 or 1-0||3-0|
|Orlando||Seattle||1-1 or 2-1||1-3|
|Los Angeles||NY Red Bulls||2-0||2-2|
As can be seen, the predicted results matched the actual outcome in 8 of the 10 matches. This is the best outcome experienced to date.
These predictions/results are posted for your review, and to assist you in your Prediction Game choices and possible lineup considerations in both of the Fantasy Football games. It is done with the understanding that these predictions may not outperform the processes that you currently use, but are only offered as another tool in your kit.
As stated each week, “You pays your money, and you takes your choice”. This old adage is so true, as it applies to predicting football score lines, and it is also a reminder to us that when we make plans, God laughs.
As always, I make no claims that the results presented here should replace what has worked for you in the past, but offer them to compare with your prediction choices and as an aid for those managers strapped for time. I hope all have fun playing the Prediction Game (open to all registered site regulars), and until next week, good luck with the numbers game.