close
Please login using this form
Registered users can
  • Post comments and rate articles and downloads
  • Submit articles, downloads and graphics
  • Post in the forums!
If you would like to publish an article or become a member of our staff, we would love to have you on board! Contact us
Login

FM2009 players

Text Links

Users online

We have 97 guests and 1 member online

Featured discussion

Which are the best DCs in the world of Football Manager 2009? Take part in the discussion
Would you like to share your Football Manager story/article with us? Submit it here

The Mechanics of Manager Reputation: Hypotheses and experiments
(19 votes)
This is where I list my experiments and working hypotheses. Please note that I do not update what I have written about each experiment even though new material that contradicts my initial statements are revealed, so If my thoughts here contradicts what I have written else were this post is out of date. Starting variables and applying for jobs I have done some experiments to try to discover how the different manager reputation variables work. My main working hypothesis has been that when you apply for a job the game compares your reputation to that of the club, and if your reputation is equal or better you would get the job. However, it's not as simple as Manager reputation > Club Reputation = You get the job. The most straightforward variable seems to be "Home reputation". If you home reputation is grater than the clubs reputation you would usually get the job, given club has same nation. If this variable only applies to clubs of your nationality it could explain why people report that Nationality has an effect on reputation, even though my data proves that nationality does not change the variables in any way. It doesn't seem quite as easy as "Home rep" applies to certain clubs and "World rep" applies to the rest though. I have not found any data to contradict my theory of the "Home reputation" variable, but I have found plenty of data to contradict my assumption about "World rep". In this experiment I created 4 Norwegian managers in the same game at the same point in time, one for each of the difficulty levels except Automatic (I will be ignoring this difficulty level from now on, as my testing has proven that the only thing that changes it is what your least reputable playable league is). In contradiction to my "World reputation" hypothesis I found that my manager with 8000 home rep and 4000 world rep got the available Coventry job, while Coventry has a reputation of 4850. 4850 > 4000 obviously, so it's not as simple as I originally thought. It could be that Home reputation applies to all countries in the same continent as your home country, but my second manager with 6000, 3000 was rejected for the Southampton job, while Southampton's reputation is 5500. 6000 > 5500 so this hypothesis doesn't seem to work either. To further confuse things my 8000, 4000 character got the Ukrainian National job, reputation 6000, and my 3250, 1625 character got the Nigerian national job, reputation 5500!. Seems like national teams follow their own set of rules. Hopefully we will be able to make sense of these thing as more data turns up. Reputation changes from friendly matches For this experiment I played a friendly match at home with Modum (Rep 2185) against the third best Norwegian team Brann (Rep 5554), the match odds were 9-1 (home), 4-1 (even) and 1-4 (away) in favor of Brann. I found that the following results gave me the following changes in reputation afterwards (Sorted by goal difference): Modum 5 - 1 Brann, +7 Home rep, +7 Current rep, no change for world rep. Modum 2 - 2 Brann, +2 Home rep, +2 Current rep, no change for world rep. Modum 2 - 3 Brann, No change Modum 0 - 2 Brann, No change Modum 0 - 4 Brann, -1 Home rep, -1 current rep, -1 World rep. Modum 1 - 6 Brann -2 Home rep, -2 Current rep, -1 World rep. Modum 0 - 6 Brann, -3 Home rep, -3 Current rep, -1 World rep. Modum 0 - 10 Brann, -7 Home rep, -7 Current rep, -1 World rep. From these data I draw that the game predicts the favorite to win by a certain goal difference and sets a threshold there. If the favorite wins by this threshold or less nothing happens, if the difference is greater the manager of the favorite team gains reputation points at the cost of the other manager. Each goal the result differs from the predicted threshold represents either 1 point gained or one point lost for each of the managers (unless the favorite wins, which seems to prevent gain for the other manager). In this case the threshold seems to be between 2 or 3 goals in favor of Brann. Why two thresholds and not one? It's tempting to guess that the game calculates the difference by dividing the reputation of the favorite club with the oppositions reputation. In this case that would mean 5554/2185 which equals 2.54, as you can see quite in the middle of 2 or 3 goals. This equation does not incorporate home advantage though, so the equation might have more to do with the odds, but then again you would usually play roughly as many home as away matches during a season so it should even out in the long run. 4-1 in favor of Brann is a margin of three goals, but that doesn't explain why a 2 goal margin would be counted as a no change score. More data will be required to fine tune the equation. But for now a seemingly satisfactory equation could be "Predicted goal difference" - "Actual goal difference" = "change in reputation." Given the win is greater than predicted for the favorite, or the outcome (Win, draw loss) better than predicted for the opposition. If this number is positive the favorite manager would lose this much reputation, if it's negative he would gain this much reputation. The opposite holds true for the other manager. This equation predicts that if a draw is the favored result either manager would gain or lose the goal difference as if the equation had no given condition. World reputation behaves unlike the other two variables, but I'll leave it until I get more of an idea what it does. If you let your assistant handle the match you get no change to reputation. Pinning the reputation change equation Game 1: Modum (Player, Rep 2416, home) VS Rosenborg (Computer, Rep 5018, away), Friendly game Player manager1 (Modum) reputation: Home: 2629, Current: 2607, World: 1266 Match odds: Modum: 13-2, Draw: 11-4, Rosenborg: 2-5 (Fav) Modum 0-7 Rosenborg, -5,-5,-1 Modum 0-4 Rosenborg, -1,-1,-1 Modum 0-3 Rosenborg, No change Modum 2-3 Rosenborg, No change Modum 0-0 Rosenborg, +1,+1 Modum 1-0 Rosenborg, +4,+4 Modum 3-0 Rosenborg, +7,+7 Modum 8-0 Rosenborg, +7,+7 It seems like 7 is the biggest amount the reputation variables can increase from one single friendly match, as I didn't get a bigger increase by winning 8-0 than by winning 3-0. The changes does not seem to be quite as linear as I originally thought, but the equation still gives you the general idea. Interestingly the gain was smaller for a draw here than in the Modum-Brann match, possibly this was because the odds were better for a draw in this game, 11-4 (2.75-1) compared with 4-1. Game 2: Modum (Player, Rep 2416, away) VS Brann (Computer, Rep 5476, home), Friendly game Player manager1 (Modum) reputation: Home: 2629, Current: 2607, World: 1266 Player manager2 (Brann) reputation: Home: 3250, Current: 3250, World: 1625 Match odds: Brann: 1-8 (Fav), Draw: 5-1, Modum: 12-1 Modum 0-4 Brann, Manager 1: -1,-1,-1. Manager 2: +2,+2, 0. Modum 4-7 Brann, Manager 1: No change. Modum 1-2 Brann, Manager 1: No Change. Manager 2: No change. Modum 0-0 Brann, Manager 1: +2,+2,0. Manager 2: -1,-1,-1. Modum 2-1 Brann, Manager 1: +7,+7,0. Manager 2: -3,-3,-1. Modum 4-2 Brann, Manager 1: +7,+7,0. Manager 2: -3,-3,-1. Modum 4-0 Brann, Manager 1: +7,+7,0. Manager 2: -8,-8,-1. A pattern seems to be emerging. It seems like the biggest you can loose any match without a reputation loss is by 3 goals, as there was no difference in threshold value in either of my three matches so far (1-4, 2-5 and 1-8). The biggest increase you can get from a draw seems to be 2, as there was only 1 point difference between odds of 5-1 and 11-4 (2.75-1), still to early to say anything certain about this though. Your potential gain from winning against superior opposition seems to be highly correlated to the odds, as a 1 goal win gave a 4 point increase at odds of 13-2 while it gave 7 points at odds of 12-1 (which makes sense, as the odds is nearly twice as bad.). Game 3: Modum (Player, Rep 2416, Home) VS Brann (Computer, Rep 5476, away), Friendly game Player manager1 (Modum) reputation: Home: 2629, Current: 2607, World: 1266 Player manager2 (Brann) reputation: Home: 3250, Current: 3250, World: 1625 Match odds: Modum: 7-1, Draw: 3-1, Brann: 1-3 (Fav) Modum 0-4 Brann, (e), Manager 1: No change. Manager 2: +3,+3, 0. (pinned back, they dominated, 50%, 0-4) Modum 1-3 Brann, (a), Manager 1: No change. Manager 2: +1,+1, 0. (near total domination) Modum 1-3 Brann, (e), Manager 1: No change. Manager 2: +2,+2, 0. (had by far the better of the match and eased through, 50%) Modum 1-3 Brann, Manager 1: No change. Manager 2: +2,+2, 0. Modum 0-2 Brann, (a), Manager 1: No change. Manager 2: +1,+1, 0. (Near total domination, coasted through, 54%) Modum 1-2 Brann, (a), Manager 1: No change. Manager 2: +1,+1, 0. (Could and should have scored more, dominate game, 54%, 1-1) Modum 2-3 Brann, (a), Manager 1: No change. Manager 2: +1,+1, 0. (Could and should have scored more, dominate game, 54%, 0-0) Modum 2-3 Brann, (h), Manager 1: No change. Manager 2: +1,+1, 0. (Good display, both side had chances, 52%, 2-1) Modum 3-4 Brann, (h), Manager 1: No change. Manager 2: No Change. (entertaining, 51%) Modum 0-0 Brann, Manager 1: +2,+2,0. Manager 2: -1,-1,-1. Modum 1-1 Brann, (a), Manager 1: +2,+2,0. Manager 2: No Change. (massive superiority, 58%) Modum 1-1 Brann, Manager 1: +2,+2,0. Manager 2: No Change. Modum 2-2 Brann, Manager 1: +2,+2,0. Manager 2: No Change. Modum 2-2 Brann, (a), Manager 1: +2,+2,0. Manager 2: No Change. (superiority, 53%) First thing to note here is that the odds are more in my favor than the last time I played Brann at home. I accredit this to the fact that my clubs reputation increased from 2185 to 2416 after the season change, while Brann had their reputation decreased from 5554 to 5476. The ratio difference (5476/2416=2.27) is quite a bit better than it was before (5554/2185=2.54). This seems a plausible explanation for the change of odds in my favor I think. It's becoming increasingly clear to me that this calculation is quite more complex than I originally thought. As you can see I lost 0-4 this time without any change in reputation, both the times I played Brann before (with worse odds) and lost by the same score I also lost reputation. I am going to gather some more data from this match before I try to analyze it further, but any input would be appreciated. Reputation Changes over time For this experiment I created 3 unemployed Malaysian managers, one with International reputation, 70 years old, one with professional reputation, 50 years old and one with semi-professional reputation, 30 years old. I had only the three Spanish leagues running to make this go faster. Immediately after creation I sent all three of them on a 5 year long holiday and let the game alone to work it's way through the years. After five years I checked their reputation values in FMM and the results were unanimous. Time alone does not change reputation. This is given that going on holiday does not make you immune to reputation change, but I can't see why it logically would. So I would say it's safe to falsify hypothesis 6, based on these data. Ingame Reputation Labels I have done some testing to uncover which of the three reputation variables govern the ingame reputation label. For these experiments I created a Manager with English nationality, International footballer reputation, and he took charge of Arsenal. I changed all the reputation variables in realtime with FMM and checked the Label changes ingame straight away. These are the data I found: Data set 1: (uncovering English reputation label thresholds) (home),(current),(world) = (ingame label) 10000,10000, 5000 = World Class 8000, 8000, 4000 = Continental 6000, 6000, 3000 = National 4000, 4000, 2000 = Regional 3000, 3000, 1500 = Regional 2750, 2750, 1375 = Unproven 2500, 2500, 1250 = Unproven 2000, 2000, 1000 = Unproven From these data we get a general idea of what the threshold values of each individual reputation label are. Interestingly it appears that you cannot have Local or Obscure reputation when managing in england. At least not when only managing the Premiere division. Data set 2: (Uncovering the reputation label variable) (home),(current),(world) = (ingame label) 10000,10000, 0 = World Class 10000, 0,10000 = Unproven 0,10000,10000 = World Class 0,10000, 0 = World class 0, 3000, 0 = Regional 10000, 2750,10000 = Unproven From these data we can conclude that the only variable that affect your ingame reputation label is Current Reputation. Data set 3: (applying for the Nigerian national job) (home),(current),(world) = (response to application) 10000,10000,10000 = Got the job right away 10000,10000, 0 = No answer after a month 10000, 0,10000 = No answer after a month 0,10000,10000 = Got the job right away It appears from these data that the your chances of getting a national job is a function of your Current reputation and your world reputation combined. Neither of these maxed out by it self was enough to instantly bag the job. Home reputation seems to have no effect. For the record Nigeria had a reputation of 5500 and started the game without a manager. Comparing a friendly to a league game In this experiment I wished to compare a friendly and a league game with the same match odds to see if I could determine just what effect competition reputation plays in the reputation equation. To achieve this I played a friendly against my first league opponent just one week before the season started. Here are the data from the two matches: Game1 Randaberg (Player, Rep 3025, Home) VS Modum (Player, Rep 2416, away), Friendly game Player manager1 (Randaberg) reputation: Home: 3250, Current: 3250, World: 1625 Player manager2 (Modum) reputation: Home: 2631, Current: 2609, World: 1266 Match odds: Randaberg: 4-6 (Fav), Draw: 9-4, Modum: 7-2 Randaberg 7-0 Modum, Manager 1: +2,+2, 0. Manager 2: -5,-5,-1. Randaberg 9-5 Modum, Manager 1: +1,+1, 0. Manager 2: No Change. (Modum 2 red) Randaberg 5-1 Modum, Manager 1: +3,+3, 0. Manager 2: -2,-2,-1. Randaberg 6-3 Modum, Manager 1: +1,+1, 0. Manager 2: No Change. Randaberg 5-4 Modum, Manager 1: No Change Manager 2: No Change. Randaberg 3-2 Modum, Manager 1: No Change Manager 2: No Change. Randaberg 2-1 Modum, Manager 1: No Change Manager 2: No Change. Randaberg 1-0 Modum, Manager 1: No Change Manager 2: No Change. Randaberg 1-1 Modum, Manager 1: -2,-2,-1. Manager 2: +1,+1, 0. Randaberg 0-1 Modum, Manager 1: -4,-4,-1. Manager 2: +4,+4, 0. Randaberg 1-2 Modum, Manager 1: -4,-4,-1. Manager 2: +5,+5, 0. Randaberg 2-3 Modum, Manager 1: -5,-5,-1. Manager 2: +5,+5, 0. Randaberg 0-2 Modum, Manager 1: -8,-8,-1. Manager 2: +7,+7, 0. Randaberg 1-5 Modum, Manager 1: -8,-8,-1. Manager 2: +7,+7, 0. Game2 Randaberg (Computer, Rep 3025, Home) VS Modum (Player, Rep 2416, away), Norwegian first division (rep 7) Computer manager1 (Randaberg) reputation: Home: 3236, Current: 3236, World: 1657 Player manager2 (Modum) reputation: Home: 2636, Current: 2614, World: 1266 Match odds: Randaberg: 4-6 (Fav), Draw: 9-4, Modum: 7-2 Randaberg 7-0 Modum, Manager 1: +5,+5,+2. Manager 2:-10,-10,-2. Randaberg 7-2 Modum, Manager 1: +4,+4,+2. Manager 2: -6,-6,-1. Randaberg 4-0 Modum, Manager 1: +3,+3,+2. Manager 2: -1,-1,+2. Randaberg 2-0 Modum, Manager 1: 0, 0,+1. Manager 2: 0, 0,+3. Randaberg 2-0 Modum, Manager 1: +1,+1,+1. Manager 2: 0, 0,+3. Randaberg 3-2 Modum, Manager 1: 0, 0,+1. Manager 2: 0, 0,+3. Randaberg 2-1 Modum, Manager 1: 0, 0,+1. Manager 2: 0, 0,+3. Randaberg 1-0 Modum, Manager 1: 0, 0, 0. Manager 2: 0, 0,+3. Randaberg 0-0 Modum, Manager 1: -2,-2,+1. Manager 2: +4,+4,+4. Randaberg 2-3 Modum, Manager 1: -9,-9, 0. Manager 2: +10,+10,+4. Randaberg 2-3 Modum, Manager 1:-11,-11,-1.Manager 2: +11,+11,+4. Randaberg 0-2 Modum, Manager 1:-14,-14,-1.Manager 2: +16,+16,+4. Randaberg 1-3 Modum, Manager 1:-15,-15,-1.Manager 2: +14,+14,+4. Notice that the match odds of the two games are equal and that the game is played in the same arena both times, so the only changed factor between these matches are the competition (Friendly, rep 1, VS. Norwegian First Division, rep 7). Competition reputation definitely has an effect, as the reputation changes are significantly bigger in the league game and the limit to the reputation change size is higher. Just how the reputation factors into the equation however is difficult to say. It's not a simple multiplication or addition to the numbers from the friendlies, but simple graph analysis of the numbers do suggest that it multiplies the result in some way. The size of the changes seem to rise exponentially the more unlikely the result, whereas the most likely results see no change from competition rep. These observations refer to the variables Home and Current rep in this case, World rep is a chapter to itself. As I have noted before world rep can only change negatively from friendlies, and only marginally at that. This does not hold true for league games apparently, it was in fact very difficult for the Modum manager not to increase his world rep, as he had to lose by at least 5 goals not to do so. I will not speculate too much why this is at this point, but it seems like it is enough just to play against a more reputable team to gain reputation. Furthermore it is becoming clearer and clearer that there is another factor in the equation besides Comp rep, goal difference and odds. This factor does not seem to be of critical importance, as it has only accounted for differences of max 2 points so far, but it becomes very difficult to analyze the data until it is identified. Manager reputation and signing players For this experiment I Created a Norwegian manager with rep scores of 500,500,250 who took the helm at Fulham. I the did a player search and asked my assistant to filter out unrealistic targets. I took a screen shot of the players that came up, retired my manager, created a new one with scores of 8000, 8000, 4000 and repeated the procedure. The results where clear: Manager reputation do affect what players you can sign. When I compared the two lists there where clear differences. The first manager was able to hire 53361 players, while the second manager could chose from 54165 players. Considering the fact the the two managers was at both their extremes reputation wise, and that manager 2 only could hire 804 more players than manager 1 (1.5% more), this effect is not that great. Among the 22 most valuable players manager 2 only had 3 more players than manager 1 to chose from. Hypotheses These have been my most important working hypotheses during my experiments. Part of the reason I list them is that I would like input from you, particular on the unresolved ones. I have also listed instructions on how to check those. If you want to double check my findings and tell me that I'm wrong please do so. That's the way science moves forward. Hypothesis 1. You get a boost to reputation after you play your last match of a league or a tournament. (Falsified) You do indeed get a boost to your reputation if you either Win or get promoted in your current league. This boost does not necessarily come at the last day of the season however. You get it with the message that your team has won the league or got promoted. By winning the Norwegian first division my Reputation scores increased from 2765, 2738 and 1410 to 4061, 4034 and 2190, which makes for increases of 1296, 1296 and 780. To get a boost from a Cup or tournament you apparently have to reach the final. By losing the semi-final of the Norwegian Cup I did not get any reputation, but by losing the final my rep scores increased by 442, 441 and 231. By winning the Cup I got increases of 757, 756 and 463, which roughly gives you a ratio of 3-5 (Loose - Win). Hypothesis 2. Manager reputation increases according to league result at end of season. (Verified, given that the result is sufficiently good.) You have to either win or get promoted in the lower divisions, premiere division haven't been tested yet. Hypothesis 3. The size of these reputation increases depend on the reputation of the league. (Falsified?) I would say that this was falsified for certain except there is a genuine possibility that the League increases I got was the sum of a winning boost and a promotion boost. At least you cannot compare winning a league to winning a cup. Hypothesis 4. Reputation can change from match to match, depending on result and opposition. (Verified, preliminary equation) Preliminary equation: "Predicted goal difference" - "Actual goal difference" = "change in reputation" Given the win is greater than predicted for the favorite, or the outcome (Win, draw loss) better than predicted for the opposition. I have done some testing on this and reputation can indeed change after a single match, magnitude of change depends on result and opposition. I also have to retract my previous statement about friendlies not affecting reputation as my data have proven that they can indeed significantly affect manager reputation (Though not club reputation). If you let your assistant handle the match it has no effect on your reputation. The size of the change depend on the odds, which is a function of the club reps, home advantage and current form, and they are apparently "Normally distributed". This last bit means that the more unlikely the result the greater the change to reputation. Hypothesis 5. Home reputation also applies to secondary nationality teams. (Still needs testing) If you want to check it just create a manager with a secondary nationality and apply for jobs in your secondary country. Hypothesis 6. Reputation declines over time. (Falsified) This was pointed out by Tokey in the thedugout.tv forum. Testing has shown that time alone does not change reputation at all over a time period of five years. The only thing that's still uncertain is whether going on holiday makes you immune to reputation change, but until someone can show me data that proves this is so I declare this hypothesis falsified. Hypothesis 7. If team A is predicted to win by x to the power of y (like 4-1), then A is predicted to win by x/y-1 goals. (Falsified) Not that simple. Hypothesis 8. Ingame Reputation Label is correlated to one or more of the three rep variables. (Verified) Testing has proved that the ingame rep label is a categorization of the Current Rep variable. Home and World rep have no influence on this label. Hypothesis 9. Manager reputation affect what players you can sign (Verified) This was pointed out by SCIAG at the sigames.com forum. It was proven by manipulating the reputation of a club manager of a certain club, in a standard game with 50k+ players, to both extremes. Manager reputation only counted for up to 1.5% of the available players to the very reputable manager however, so club rep is a much more important factor here. Hypothesis 10. Reputation changes and pre-match odds are calculated according to the ELO ranking system. (Falsified?) This was suggested by DaveRH in the sigames.com forum and could seemingly explain the normal distribution and ranking mechanisms used in FM08. My test calculations did not give completely accurate predictors but they were quite close. As this system apparently is used in proffessional football it would make sense for SI to use it, but they apparently don't. More information can be found on this link: http://en.wikipedia.org/wiki/Elo_rating_system Hypothesis 11. If you are currently under contract and are offered a new job your new club will have to pay to release you from the contract. (Verified by ZJ) Differences in bank account pre and post signing (recorded by ZJ) has proved this. This means that financial considerations comes into play when a club hires a new manager and that you thusly are more likely to be offered jobs if you are unemployed and be linked to more jobs if you have little to go on your current contract. Final thoughts I would greatly appreciate your input on these hypotheses, especially input which leads to new testable hypotheses, and/or verifies/falsifies the current hypotheses. Or if you doubt my results feel free to try to replicate them and post here to tell me I'm wrong. All credit will be duly given in the main threads to those who contributes to changing it's contents. Credit so far: Vertanno (cmfrenzy.com forum), some data contributions and pointing out that the Local and Obscure rep labels are only available to AI managers. crouchaldinho (sigames.com forum), data contributions on applying for national jobs. ZJ (sigames.com forum), too many contributions to specify!


Next >
 

Add your comment

Your name:
Subject:
Comment:
yvComment v.1.18.4
Enlarge video
Top10*FM*Matches*FM2009*Various