The match engine – part 1: Chances

Before anyone starts a flaming war. The data I based my theories on can be found here: http://wihlah.aleander.no. The experiment is described here: https://wihlah.wordpress.com/2009/01/12/the-match-report-analysis-experiment-the-background/. I am more than happy to receive critical and constructive feedback on these experiments. I cannot guarantee that I didn’t miss something, so If you notice something out of place, don’t hesitate. Suggestions to further experiments or calculations are also more than welcomed.

The chances in a match

In the forum it has been said that each match has 10 set chances. I do not believe this is exactly true, and we have been given examples of matches with more goals. Special events are not a part of “chances” as such, and maybe also penalties. I simply haven’t been able to figure that out. What we DO know is that not all chances are reported on the match report. So IF the max of 10 chances is true, the average number of chances should not be above 10. Because of the crappy (hope no one is offended by this)  teams I have been playing with this season, most of the chances should statistically be on the match report, as there isn’t that much of a chance the defenders/midfielders of my opposing teams would stop anything (except for det beste teamet, which can clearly be seen on the match reports of those two matches as well).

The average number of chances in my 14 matches this season is 9,64. Remembering that only real chances are reported, the claim of max 10 chances in a match seem to hold up.

The distribution

There were 135 chances in the 14 matches of which:

• 88,1 % of those were regular chances
• 5,2 % were free-kicks
• 3,7 % were penalties
• 0,7 % were corners (only one this season clearly a corner)
• 2,2 % were special events (long shots)

This means that 9,6% of the chances were set-pieces chances. HT-Tjecken (I believe, the source of the statement is much appreciated), has claimed that around 10% of the chances are supposed to be of that type, which proves fairly true.

First half vs second half

The chances between the two halves are distributed as follows:

	First half	Second Half
Total chances	50,37 %	49,63%
Regular chances	52,94%	47,06%
Free kicks	14,28%*	85,71 %*
Penalites	20,00%*	80%*

*Note that there were only 7 free kicks and 5 penalties this season, and so this could very likely be lack of data. But it would not suprise me if Hattrick has actually taken this into account either, it is nevertheless a fun fact.

Conclusion on chances in general

The claim that each game has 10 chances distributed between the teams has yet to be called false, but It has not been proven either. The claim that 10% of all chances are set pieces chances, seems valid enough with my data at least.

I will look more closely into chance distribution vs playmaking advantage, effectivity/efficiency, chance vs player position distribution and attack side distribution with the AiM tactic later on.

The match report analysis experiment – the background

I decided, while reading a book actually, to do some experiments on the match engine.

“How do you do that exactly”, i wondered sitting there. And then it struck me, loooong in the back of my head I vaguely remembered TH-Bjorn saying:

On Match report, Line-up (Forum post 2002523.108): 2002523.107 2002523.107 2002523.107

Match report is the only data available for us. If it doesn’t say everything ’bout our team’s quality then what does? Our imagination? ;P
Who says match report doesn’t say everything??? I didn’t say that. I said the ratings doesn’t say everything. The match report has further information to add. Some information is only known to the manager of your opponent. That is how it is supposed to be. It is also how it is IRL

One more time: “Who says match reports doesn’t say everything???”. So I did some searching and found that this link http://club.hattrick.org/KawasakiTigers/default.asp?site=http://mrat.hattricknippon.org, which is dead now. SO I did the next natural thing; I created my own tool.

The tool

The tool is not open to everyone yet unfortunately, I need to modify it somewhat so it will be if there is a need. What I basically did is to break the basics down and collect all the information available to us in the match report. I found this to be:

Matchdetails

For my team:

• Matchid
• Season
• Round
• Matchtype – league, friendly or cup
• Date
• Home/away
• Possession first half (mine only, calculated the rest)
• Possession second half (mine only, calculated the rest)
• Team attitude
• Team tactic

For both teams:

• Team exp
• Team stars
• The midfield/dr/dc/dl/ar/ac/al ratings
• Hatstats
• Loddar stats (not implemented yet, but I will :))

Match Events

From each line of text in the match report, I deducted the following parameters:

• Owner – me or opponent
• Type – chance, corner, freekick, penalty, injury, booking, special event
• Goal – yes / no (if the event resulted in a goal. Typically injuries and bookings dont ever. The others might)
• Minute – is not always given, but the intervals are quite clear so I guessed at some
• Side – from the right, left or center (or none)
• Player – Name or nick or id, doesnt matter
• Player position – Which is quite interesting I think
• Description – The actual wording in the match report

The boring part

I’ll tell you right now. Entering these details when only one match left in the season wasn’t very fun, but nevertheless I did it. All the details. For all of my league games at least. And I started to play with numbers. I will present the results in a separate post, but I will tell you, this is actually very informative. I have been browsing through those reports briefly and concentrated on the stars and player performances. I have never actually looked at the chances for/against, The chance conversion ratio, the efficiency of the chances or the number of chances from possession advantage. I could also from this deduct the number of set pieces chances, the number of chances totally and average in a match.

Why can 14 matches be enough statistical data to be trustworthy?

Actually it cant. Its way too little data. But I do have some advantages which makes it more sensible.

1. I have had the pleasure of winning the possession (often by far) in 100% of my matches
2. I won all the games of the season
3. I played the exact same tactic (almost at least) the entire season
4. My opponents have been really bad through all the matches with almost no active managers

But I did not use the same players, and my hatstats have differed somewhat. But still, some general things can be concluded from these data.

Data source

Those of you who like details, the results can be found here: http://wihlah.aleander.no