Realism Project

This evening when I have more time I'm going to see just how accuratly Football manager models real life football. As you know the demo contains the real fixture list. Im going to print off the scores for the opening weekend of football and see what percentage of them football manager gets right.

I'm not talking about the exact score but the correct result e.g. away win/loss, home win./loss or draw.

Tune in this evening to see how the game does but take a few guesses, we'll see how close it actually is

Nice idea. But if I did it - we'd both get a totally different set of results. So I'm not sure what any outcome will actually mean, in terms of usefulness

Nice idea. But if I did it - we'd both get a totally different set of results. So I'm not sure what any outcome will actually mean, in terms of usefulness

I totally agree, this is a game after all. Its like when you play a match and then shut down and replay the match and you get a completely different result.

Plus take into account any real life injuries that players had, so the ai teams could be playing completely different players, in different positions with a different formation.

Overall, it seems like a fun little experiment, but as expressed above - you can't use this as a model as it just doesn't work.

I think some people are mixing reality and realism when they speak about FM.

FM is realistic, but it will never reflect the reality, Thank GOD !

you'll need to disable transfer budgets in the first window.

How many reloads would it take for Colchester to win 7-1 at Norwich, the sort of result that if it hapens in FM people would be on here saying it's unrealistic.

It will be interesting to see the results. Every post has been negative in saying it will prove nothing etc but its not about proving anybody wrong its just an experiment. Obviously the results will be different and there are many different aspects to consider but go for it mate.

Its not about saying hey in FM it was 3-0 but in real life it was 0-3- how inaccurate is FM? Just be fun to see the results!!!

How many reloads would it take for Colchester to win 7-1 at Norwich, the sort of result that if it hapens in FM people would be on here saying it's unrealistic.

Lots of people are confusing results with scores, I'm not looking for exact scores, im looking for Colchester to beat Norwich for it to match real life. I agree you will get freak results but for the majority of games the better team will win.

Im just wondering out of curiosity how accuratly the programmed teams reflect what that team would do in real life.

Obviously there are going to be small differences like certain players will be injured when they wernt in real life but i think it'll be an interesting test on the whole and i think it will show how accurate the database is in terms of mirroring what happens in real life.

nice idea, will be interesting to see the results.

Plus take into account any real life injuries that players had, so the ai teams could be playing completely different players, in different positions with a different formation.

Overall, it seems like a fun little experiment, but as expressed above - you can't use this as a model as it just doesn't work.

Well injury proneness is a stat for players and also the managers have stats , So again this would be reflected by how close FM got to the real thing in my opinion, And of course some of it will be down to luck in real life and in FM.

Will look forward to the results.

THE RESULTS

Aim:

The longer the game goes the more AI and the user will change the game from how it pans out in real life. in theory, At the start of the game the teams are as closely matched to their position in real life than at any other point of the game. I wanted to see how accurate they were by seeing how differently/similarly they did against real results. (I mean results rather than exact score lines)

What i did:

Started unemployed and went on holiday till after the first weekend of football league fixtures. i wanted minimal time for AI to change things with signings and things and also minimal time for injuries to make availible squads radiacally different

The Results follow:

THE RESULTS

Aim:

The longer the game goes the more AI and the user will change the game from how it pans out in real life. in theory, At the start of the game the teams are as closely matched to their position in real life than at any other point of the game. I wanted to see how accurate they were by seeing how differently/similarly they did against real results. (I mean results rather than exact score lines)

What i did:

Started unemployed and went on holiday till after the first weekend of football league fixtures. i wanted minimal time for AI to change things with signings and things and also minimal time for injuries to make availible squads radiacally different

The Results follow:

You must be either bored or have too much time on your hands

There are no prmeiership games by the way as they kick off a week later meaning more time for AI to change the teams and make the test less realistic

In the championship my game got one result correct

Preston 1-1 bristol City (Real scoreline was 2-2)

League 1 was a little better, the game predicted 3 correct results

Gillingham 3-0 Swindon (5-0 real scoreline)

Southampton 0-0 Millwall (1-1 Real scoreline)

Yeovill 3-0 Tranmere (2-0 Real scoreline)

League 2 was similar to league 1

Rotherham 3-0 Accrington Stanley (3-2 Real Score)

Lincoln 3-0 Barnet (1-0 Real score)

Cheltenham 2-1 Grimsby (2-1 Real Score)

I thought the game would be a little better. It'd be interesting to see if the number of correct results stayed roughly the same each time you did it, might do it myself another time.

Interestingly there were a very large number of reversescorelines predicted correctly

I still maintain that the game is very accurate, there were no score lines were i thought WOAH! thats way off the mark and all were very close to the real scores, i dont think there was ever a differenceo f more than 2 goals in all the divisions.

Damn FM sucks !

I guess the moral is not to use FM to guide you for you ladbrokes accumilators

edit: other bookamkers are availibkle

I think this would be an interesting thing to do.

Thing is, you need something to compare the FM predictions to i.e. a benchmark.

A couple of suggestions...

- Any football tipster, either in the papers or maybe Mark Lawrenson for the BBC

- The Times' "Fink Tank" predictor, which is a tool using statistical evidence and a computational match engine to provide different % chances of different results (Links below)

http://www.timesonline.co.uk/tol/sport/football/fink_tank/

http://www.dectech.org/football_sites/football/

Also, I would just stick with providing predictions for home win, draw, away win. Expecting individual scores to be accurate is way too much.