With every release of Football Manager, there’s work to be done for folks like me that like to deconstruct the game, and attempt to really make sense of the big spreadsheet that FM can feel like at times. With the Editor available, I got right to work and I have a gift for you all, as well as the first of the interactive projects I intend to run here at Strikerless.
The Test League
The Test League is the foundation for all tests to come this year. While it’s not terribly useful on its own, a lot of work went into making it a good jumping off point for me, or you, or anyone that wants to knock out some tests quickly. It contains 36 teams. Those 36 teams are identical aside from name. They are each populated with 22 outfield players and 3 keepers, all of whom have identical stats of 10 in all categories. The outfield players have 20 in ability to play every position, and the keepers have 20 in keeper. The teams each have a Chairman, Head Coach, Assistant, Goalkeeping Coach, Fitness Coach, 6 Standard Coaches, a Head Physio, and another Physio, all of whom are clones of the same person with 10s in all categories. Additionally, all 36 teams are under transfer embargo for the next 100 years. They play in The Test League, which happens on the island of Bermuda. Each team plays each other once.
I pulled out every trick I had to get this league built quickly so others can begin to conduct their own experiments. It is available for download at Steam Workshop. Comment either here or there if you have any questions or notice any bugs.
Major League Statistics
This is the first interactive experiment that I want to run, and it’s simple enough in its setup. It begins with The Test League above, and then each of the 36 teams has all players set to 20 in a particular attribute. So there’s a team with all players having 20 in Corners, a team with all 20 in Determination, and so on, for each of the visible stats that a player has.
I ran this last year ten times, and made some interesting discoveries, but ultimately I threw the data out as unreliable because I did not make a perfectly clean Test League as I have this year; each player actually existed in the real world and though their height, weight, age and attributes were all set identically, I neglected to remove player traits, and couldn’t be sure they weren’t having undue influence on the final results.
This year, I’d like you all to take part as well. Doing so is quite simple:
- Download Major League Statistics 2018 from Steam Workshop.
- Load the data file and begin a new game in Bermuda (I strongly recommend unloading every other nation to make it run quickly.)
- Go on holiday until May 13th 2018.
- Go to the experiment page at Bearpuncher Labs. (Why Bearpuncher Labs? Long story, and one still to come.)
- Punch in the requested data in the submission form, as much or as little as you care to (although those of you that fill in everything are my favorites).
- Restart the save and repeat steps 3-5. Don’t continue after May 2018 as then we’re measuring other things and reducing the value of the data.
I plan to run this until I’ve got enough data that I’m really confident in the results. Some results will be interesting to you as soon as you try the simulation, and some may be misleading until it’s run a number of times and you can compare your results with others. We’ll be breaking down various interesting things from the data throughout the year. If you’re feeling far too lazy to punch in numbers, take a screenshot of your league table at the end of the season (please sort it alphabetically first!) and I’ll handle it.
Johnny Kollman · November 21, 2017 at 11:57 pm
Isn’t this set up also the best way to figure out which tactics are the best. Did you already tried it earlier or would it be an idea for future testing?
bluesoul · November 22, 2017 at 4:10 pm
Yes indeed! It’s in the cards for the future. Something like using all of the default formations, take the top 8 and do a league where each team plays each other 4 times for a 28 game season. Might repeat it again with top 4 after that.
fmnatic · November 30, 2017 at 9:43 am
I fail to understand the point of this test, all you are doing is identifying which attributes work best without the help of other attributes.
For eg. heading, is linked to jumping reach. In this setup all other attributes being equal, i expect the heading team to have a boost to both attack and defending and do well.
The team with 20 decisions is probably going to do poorly , as they have no edge over their opponents to use , so its like sending a team of armchair geniuses out.
Similarly i expect 20 dribbling to do well, but i would think a 15 to dribbling+agility+balance, would be better, as these attributes work together.
fmnatic · November 30, 2017 at 9:48 am
Also has SI ever stated that simulated results use the whole match engine behavior we see during matches. As a programmer i suspect they use a faster heuristic to simulate games , the user is not a participant in.