https://www.bigsoccer.com/forum/showthread.php?s=&threadid=64954 still would really like to see a better venue to discuss some of these things in depth. please post in that thread if you agree.
Sorry, only had a chance to glance at this thread, but why wouldn't you try to break down possession to possession in the attacking half to eliminate the odd results, and/or look at possession in the box.
I've seen the same kind of thing when I looked at some 20-year-old NFL data. By assuming a Poisson distribution for scoring, you could model things pretty well. This kind of result -- obtaining good information from an overly simplistic model -- is why I've become more optimistic about the feasibility of soccermetrics.
I'm sure it could be useful but as I'm sure has already been mentioned somewhere in the previous 8 pages, it's knowing what measurements would be useful and also crucially, knowing how to interpret them. I'm sure many are aware that an old dinosaur at the FA many years ago studied hundreds of matches and found that most goals come from moves of three passes or less, and deduced from that that the long ball game, with it's emphasis on the minimum of passes, was the way to go. It would be far too easy to look at a dry stat like passing completion rates for two central midfield players and assume the one with the higher percentage is better - of course the one with the lower stats could be a creative midfielder trying to pick out players in the box, while the other could be a ball-winning fetcher & carrier type player who only plays safe passes. It would take a staggering amount of breaking down of stats to ensure that only like for like comparisons are being made.
I really think if to be productive at these types of discussions or to maybe gain some greater insight it would be best to do that in our own forum. please pose on this link and maybe we can make that happen.
This won't ensure only like-for-like comparisons, but you could do pretty well by tracking a few simple things: Location of a touch (e.g. attacking third) Type of pass (e.g. crosses) Direction of a pass (e.g. square) This doesn't account for defensive pressure and offensive support, but it does provide a way to filter out "creativity." Lower-percentage passes are only a good move when you're pushing the ball forward or when you're in the attacking third of the field. By excluding those cases, you should be able to make a pretty accurate comparison of midfielders' ability to guard possession.
What's the book's name? Probably what would be more interesting is to get the bibliography of the soccer section and see the sources. There seems to be a lot of work already done, and I hate to reinvent the wheel everytime. The dilemma of a red-card vs. a sure goal is one aspect what this discussion is about: making decisions based on more than just instinct, intuition and/or "common sense".
Chances for getting a forum don't look good... https://www.bigsoccer.com/forum/showthread.php? s=&threadid=64954 I think this topic is just too esoteric, and the all the math speak is off-putting or just over many people's head. There must be someplace- a mailing list or something on this topic somwhere. If not, Bigsoccer can be a pioneer. Build it, and maybe they'll come...
I think perhaps the place to put this topic in is the Coaches forum. I think you'd get a good input from people re: their perceptions/view on what some numbers show. For instance, I've started looking the idea of height in midfield and the role it plays not just winning and possession but the more subjective "level of play" and styles. And comparing it to MLS. Looking at Arsenal for instance Vieria is 6'3" Ljunberg 5'9", Parlour 5'10", and Pires 6ft. Liverpool w/ Gerrard, Riise, and Hamaan, have three 6ft and over, with Danny Murphy at 5'8". Man Utd. is small by comparison, as Giggs, Scholes, Keane, and Butt are all under 6ft. But they all that one big guy up front: Henry, Heskey and Van Nistelrooy at 6'2". Looking at the Fire and Metros as examples: Lisi 6-1, Williams 5-5, Clark 5-10 and Guevara at 5-11. Up front they have no one over 6ft-Mathis and Moreno at 5-10 and Magee at 5-9. The Fire have Armas at 5-7, Whitfield 5-11, Mapp 5-10 and Marsch at 5-11. Razov at 6-1 and Ralph at 6-0 are on top. At the bottom of our tables is Dallas with: Vaca 5-6, Pareja 5-8, Stone 5-9, and Davis 5-11. Up front: Cerritos 5-9 and Rhine 5-10. Hmm...
yeah it's disapointing people haven't have seen fit to post for a new forum about soccermetrics. i'm going to do a little more reasearch as to what rssf.org or whatever the heck was about. microbrew, i really liked your ideas about what that forum would represent, especially because i'm interested in this stuff but i'm a dope when it comes to math. i just think it's disapointing because it appeared to be a general topic that people were pretty interested in then this thread kind of died because of the disorganization that comes with having a topic like this just in one thread. i presonally would be a big fan of continuing to press for a new forum or sub forum on here. i just prefere the set up of a message board to that of a listserver and it'd be nice to start something like that on these boards. maybe this would have better luck in the coaches forum, but if we were looking to do non-traditional "outside the box" stuff i'm not sure if that would be the best place for it.
The below paper is where the author looks at the scoring of goals in the World cup with a poisson distribution, using a data set of 232 games. Chu, S. (2003), "Motivating the Poisson Process Using Goals in Soccer." INFORMS Transactions on Education, Vol. 3, No. 2, http://ite.informs.org/Vol3No2/Chu/
http://www.amazon.com/exec/obidos/A...1499505/sr=2-1/ref=sr_2_1/002-8819408-2287220 If I have a chance I'll get the book from the library and check out the bibliography. I'm pretty sure that the red card article is: G. Ridder, J.S. Cramer, P. Hopstaken, Journal of the American Statistical Society, 1994, 89, 1124-27. For people with access, the link is: http://links.jstor.org/sici?sici=0162-1459(199409)89:427<1124:DTTETE>2.0.CO;2-R
a little cross polination posting From Mr. Elias' latest column. I wonder if he's been sneeking peeks at this board? On a different note. Great find by someof you guys in regards to some stuff that's already been written on this subject. However, as mircobrew suggested for possiable topics of a forum relating to this stuff, probability and statistics for math dummies threads would be a massive help to people like me. Because I'm an absolute moron when it comes to numbers and math when I see Poisson distrabutions I think of being back stage with Bret Micheals.
The Poisson distribution is (hopefully) not too complicated. If the Quakes score 20 goals in 10 games, we know that they average 2 goals per game, but we don't know how consistent they are. Perhaps they score 2 goals every game, or perhaps they alternate between scoring 4 goals and scoring 0. If the former is true, the Quakes will rarely lose; if the latter is true, they'll lose quite often. When we say "Poisson," we can answer how consistent the Quakes are. The Poisson distribution makes the simple assumption that goals are scored at random time intervals. That is to say, we can ignore all of the complicating factors of tactics and streakiness and fatigue and match-ups and home field advantage. If the Quakes are going to score 20 goals in 900 minutes, we choose 20 random numbers between 1 and 900, and those are the minutes that the Quakes will score. This gives us an easy way to estimate the probability that the Quakes will get shut out, or score one goal, etc. In most cases, it works pretty well. Then again, the assumption is factually incorrect, so it won't be perfect. Every rose has its thorn.
You know how whenever a game hits the 41st minute or so, and it's 0-0, the announcers make a big deal of one team trying to get a goal to seize momentum? Here's an interesting question. Breaking the game into 5 minute blocks, what is the record for the team scoring the game's first goal in each block? Obviously, the teams scoring the first goal of the game in the last block would be pretty near perfect. I would expect that scoring first in the first five minutes would be the least important. Anyway, if, in fact, momentum is a big deal, then scoring a goal in the last 5 minutes of the first half would stick out. Teams would have a much better record than teams scoring in the 36th to 40th minutes, or the 46th to the 50th. Yes, I realize that because of stoppage time, there won't be an even distribution of goals scored. But I don't see that as relevant. I'm interested in the record. Where's an easy place for me to check this out over the last 1 2/3 MLS seasons? I might work on it at work (during lunch only, of course.)
The league breaks it down by 15-minute intervals, but not five. I've often wondered about that end-of-half momentum thing, too. Would be interesting to just pore through the boxscores and see goals scored from 40-45 minutes (and first half stoppage) and see if there really is an effect, either on the team's performance in the second half or its changes of winning the game.
This does sound like a good idea. I'm not aware of anywhere convenient for MLS data, but you could easily get the Premiership here (as well as several other leagues): http://www.rsssf.com/tablese/eng03.html If I may make a suggestion, perhaps you can record a bit of side information for each game, too, in a format like the following: HOME TEAM,AWAY TEAM,Scoring First,Minute, Final Score Charlton,Chelsea,Home,7,2-3 The reason I think side information will be useful is that you'll probably want to know whether the first goal is more important depending on the circumstances.
kenn, that's a different angle to take. I'm just looking at first goals, you're looking at any goal. That would be fun, too. Ask, did a team score in this 5 minute block? What was their record if yes? Now that I've thought a bit more, I think I'll look at all tiebreaking goals, not just all first goals.
No, I was just thinking about goals scored right before halftime, which are supposed to be momentum-shifters. Not first or tiebreaking goals, just goals.
I'm coming in at the end of a conversation (and without reviewing it) but that "Analyze This" at mlsnet.com analyzed goals just before or after the healf and as it turns out they don't effect wins or losses any more than any other goal. Something to that effect. It was a few weeks ago.
You know what would be interesting too? To look at games where opposing teams, or even opposite teams score goals in bunches. As you saw for example with the U-17 Brazil game. How many times are teams getting beat by a quick outburst of goals, even if it's just two. I'd think froma coaching perspective after objectively figuring out how often this happens that a little subjective analysis would go greatly towards hoping to understand why a game can sometimes so easily get blown open. It'd also be an interesting question to ask to gauge the consistency of teams, even if there was frequent 10-15 minute intervals where their opponents were knocking in 2 or more goals by subtracting those you might get an interesting idea if a team is living up to its full potential. Just a thought. P.S. Please please please go to the Suggestions and Announcements forum, and the New Forums section and post if you think it'd be a good idea to have a forum of it's own for this type of conversation. The guys at RSSF.org are great, but they're dealing with historical trivia, and honestly right now we're just bantering around questions and ideas. I think there's a golden opportunity to really get into some substinative analysis and discussion if we could just do it in a more organized format.
Well, there you go. No need to do that one. Doesn't really surprise me. I've never been completely convinced of the true magnitude of "momentum" in sports. I think if it actually exists, it's probably strongest in basketball (or, at least, basketball looks like it has more readily-quantifiable shifts in momentum, and players and coaches act as if it really exists).
I have all of the Crew goals scored and allowed - for all competitions - broken down into the following intervals: 1-5, 6-10, 11-20, 21-30, 31-40, 41-45, 46-50, 51-60, 61-70, 71-80, 81-85, 86-90 and 90+ (overtime matches). They're all referenced by Crew match. If anyone wants to have fun analyzing that, let me know (it's in Excel format right now; hoping to move it into a database during the winter).