Pareto rules

So far in this cycle(s)…	…A% of the players…	…(i.e. B…	…out of C players)…	…have scored 80% of the points (i.e. D…	… out of E points)
All	18%	119	677	3924	4905
1st	20%	72	360	2896	3622
2nd to 4th (new scout scoring rule)	18%	56	317	1029	1283
2nd	30%	32	107	787	979
3rd	16%	26	164	220	277
4th	15%	7	47	21	27

Pareto’s rule confirmed!

This is how it looks (all cycles):

Success

When a player you picked score a goal like this, you know it’s good.

When that player is the second worst among the ones you picked in the third cycle, when all the good scouts had joined the game and all the good players had been picked, you know it’s a complete success.

I rock… No, Fantasy Scout rocks.

Fantasy Scout for Dummies

Whenever I explain Fantasy Scout to people who are new to the game, I see that the only part that it is not-so-easy to understand is cycles. “So who has most points at the end of the cycle wins, right?” “But can you keep some players after the end of the cycle?” [insert “That’s now how it works. That’s not how any of this works” meme here]

(By the way, this is the reason why I always use the phrase “picking cycles” on the FS site. I don’t think it’s enough to avoid confusion, but I don’t know what else I could do there.)

Next time I’ll have to explain the game, I’ll try a different sppech, leaving cycles for the late, let’s-tell-the-details part.

Another trick that could make cycles easier to make sense of could be… explaining their sense. 🙂 E.g. one could rhetorically say: “But this way we would never have a final ranking, would we?* So in order to have final rankings, we keep separate rankings for players picked in different years.” (and then “But this way there would be too many rankings to keep track of, don’t you think so?* So we have biennial rankings.”)

It would be even better if the questions marked with * weren’t rhetorical, but real, i.e. asked by the person who is learning about FS.

Indeed this method could be applied to the whole explanation. I imagine something like this:

• N (as in New scout): “So what is this Fantasy Scout thing?”
• O (as in Old scout): “It’s a game where you win if you pick the best football players.”
• N: “Easy: I pick Messi!”
• O: “Well, you can’t pick players who have more than two caps.”
• N: “Then I pick all the players in all the U-21 international squads.”
• O: “But each pick costs you points…”
• N: “Now that you make me think about it… how is being good at football quantified?”
• O: “Well, it’s easy: 1 cap = 1 point, and 1 international goal = 1 point. Nothing else.”
• N: “But this way someone who picks a player who gets 150 caps for Palau would win against someone who discovered Zidane.”
• O: “This is the reason why you can only pick players from eight major nations…”

and so on.

Even though this way someone could get the feeling that the game is more complicated than it really is: there’s always one more rule… Maybe one should start with a premise such as “there are few rules, and that they’re so logical that they will naturally follow from your questions.”

(Final, meta note: in a sense I’ve applied this discovery-by-progressive-question method to the task of writing this post: I started with just the idea “I have to explain cycles in a different way” in mind, and, as I thought about it, I came up with the next paragraph; then, while thinking about the content of the second paragraph, I realized that… and so on, and the result was a complete pitch for Fantasy Scout. A pitch that probably doesn’t work, mind you; but a complete speech nonetheless.)

1st cycle possible futures, end of 2014

Here are Daniele’s data about predicted scout scores in the first cycle. They’re particularly interesting because they tell not just the average score, but also the min and max scores (excluding the 20% most extreme predictions) and the chances to win and to be in the top three.

Columns:

1. Scout
2. Mean score
3. Median score
4. 10th percentile
5. 90th percentile
6. chances to win
7. chances to be in the top three
8. mean rank

Note: some scouts that are here aren’t in the official ranking, and vice versa.

+-------------------------------------------------------------------------------------------+
     |        scout   mean_S~0   med~2010   p10~2010   p90~2010 pwi~2010  pto~2010 meanra~0 |
|-------------------------------------------------------------------------------------------|
  1. |          Sid   33.42028   32.8       25.2       42        .2836     .54       4.1804 |
  2. |         Mark   31.95634   31.8       26.2       37.95     .1484     .4596     4.4262 |
  3. |   Michael H.   31.87204   31.72727   28.54545   35.36364  .0718     .4402     4.0502 |
  4. |      Daniele   30.82016   30.6875    27.4375    34.3125   .0452     .3094     4.754  |
  5. | Sid Debgupta   28.95963   28.28572   22.71428   36        .0734     .2374     6.7866 |
|-------------------------------------------------------------------------------------------|
  6. |        Jesus   26.89452   26.5       21.3       32.9      .024      .125      8.388  |
  7. |      Alberto   25.98517   25.33333   18.16667   34.66667  .0532     .159      9.8766 |
  8. |      Tom two   24.7268    21.5       17         37.5      .098      .1786    12.1192 |
  9. |    Andrea V.   24.14705   24         21.56      26.92    0          .0026    10.7272 |
 10. |     Generoso   22.25673   21.91667   16.91667   28.16667  .005      .021     13.336  |
|-------------------------------------------------------------------------------------------|
 11. |        Nigel   21.82115   21         12.5       32        .034      .0944    14.6886 |
 12. |      Tom one   21.74915   21.5       18.41667   25.41667  .0002     .0012    13.6786 |
 13. |  Giovanni B.   21.40871   21.07692   14.38461   28.84615  .0078     .0356    14.6438 |
 14. |   Alessandro   21.3198    20.71428   15.14286   28.28572  .0054     .0322    14.6832 |
 15. |     Cristian   20.68654   20.2       15.3       26.7      .0012     .0108    15.3552 |
|-------------------------------------------------------------------------------------------|
 16. |     Riccardo   20.34554   20         14.18182   26.90909  .0034     .0172    15.784  |
 17. |    Daniel O.   20.30633   19.5       11.66667   30        .016      .0566    16.2688 |
 18. |    Andrea B.   19.70853   18.33333    6.333333  35        .0674     .1322    17.7412 |
 19. |      William   18.87587   18.54545   14.18182   24       0          .0024    17.5842 |
 20. |    Saintjust   16.9012    16.2       11.2       23.6      .0006     .0062    20.1978 |
|-------------------------------------------------------------------------------------------|
 21. |       Pietro   16.62054   16.44828   13.48276   20.03448 0         0         20.5634 |
 22. |       Mattia   16.58378   16.35294   13.41177   20.05882 0         0         20.6184 |
 23. |       Jacopo   15.51964   15          9.3       22.6      .0008     .0032    21.8332 |
 24. |   Michael M.   14.868     11          6.5       29.5      .0376     .0724    23.1248 |
 25. |    Daniel B.   14.68352   13.8        5.8       24.4      .0056     .0194    22.694  |
|-------------------------------------------------------------------------------------------|
 26. |      Tommaso   12.92805   11.75       4.25      23.125    .005      .0158    24.613  |
 27. |    Simone R.   12.72067   12.27273    6.909091  19       0          .0004    25.2212 |
 28. |      Abubakr   12.51684   11.6        8.2       18.2     0          .0004    25.6482 |
 29. |        Mpfat   11.97331   11.57143    9.714286  14.71429 0         0         26.5022 |
 30. |    Gianmarco   10.37502    9.909091   5.909091  15.45455 0         0         27.9414 |
|-------------------------------------------------------------------------------------------|
 31. |      Desiree    9.774      6.5        0         24.5      .0172     .0336    26.9594 |
 32. |    Andrea C.    9.354      8.571428   3         16.57143  .0002     .001     28.4712 |
 33. |       Giulio    7.814769   7.461538   3.846154  12.23077 0         0         30.4566 |
 34. |    Simone G.    7.394367   6.5        2.5       13.5     0         0         30.4142 |
+-------------------------------------------------------------------------------------------+

Of course scouts with more players have a narrower window, so they have higher min scores/ranks (twenty players can’t all fail…) but also lower max scores/ranks (…but neither can they all succeed!). This is quite apparent when you looks at Tom vs. me: our mean score is very close, but…

• The difference between 10th and 90th percentile is more than 20 for him, a little more than 5 for me.
• So I beat him by min score, but he beats me by max score.
• Since it’s in the futures where you get a very high score that you win or however are in the top three, he has more chances to do it than me (actually I have no chance to win! 😦 ).
• But in many futures Tom gets bad scores, so my mean rank is higher than his.

Half of these scouts picked 10 players or more: only 3 times of 10 the winner is one of them. The other 7 times the winner is a scout from the other half: none of them picked more than 7 players. If you check the thirteen scouts who picked 11 players or more, their total chance of winning is lower than 15% (basically only Michael and Daniele have a chance).

The prediction is quite bold: Sid has more than 25% chances of winning and more than 50% of finishing in the top three. The probability of neither Sid nor Mark nor Michael being in the top three is just 14%.

Links to Wikipedia

There’s a new hidden page in the FS site: “Picked players information”. For the time being, there you can find the link to each player’s page on Wikipedia (click on his name).

As you can see if you check the page, I’m planning for more.

No Lucas no party

I’ve always thought that the best way to rate a player is to compare his team’s result with vs. without him: evidence from July 2007, September 2007, June 2009.

So I am not partial at all when I link at a tweet Squawka Football highlighting that “Liverpool have lost just 2 of 21 games Lucas has played in this season. They’ve lost 8 of their 12 games without him.” 🙂

The Fantasy Scout Map

The map

(full-size)

The meaning

Roughly speaking, each player is closer to the players who are more similar to him (by position, nationality, and/or age).

How I drew it

1. I took the top 31 players by predicted final score (i.e. the ones who are predicted to score more than 67 points)
2. I computed the distance of these players from each other, this time using just three dimensions: date of birth, nation, position.
3. I took the top 50% closest couples.
4. I gave these couples as input to TGWikiBrowser.

Note to self: the correct way to do something like this is using a force-directed graph… or so I guess.

Similar players

I’ve computed how distant each picked player is from any other picked player. I’ve considered the following criteria:

• Cycle (0 if same, 1 if different)
• Date of birth (from 0 if same day to 1 if they’re twenty years apart)
• Nation (0 if same, 1 if different)
• Position (from 0 if same to 1 if they’re goalkeeper vs. forward)

(this is obviously related to the “weekly battles). I’ve used Euclidean distance.

Here are the twelve closest couples:

Player 1	Player 2	Distance
Ciano (Tommaso 2007-10)	Immobile (Giovanni B. 2007-10)	0.0002840909
Bertolo (Pietro 2007-10)	Valeri (Tom one 2007-10)	0.0004261364
Bonaventura (Daniel B. 2007-10)	Sciacca (Tommaso 2007-10)	0.0004261364
Escudero (Michael H. 2010-12)	Jose Angel (Gianmarco 2010-12)	0.0004261364
Conti (Gianmarco 2007-10)	Schelotto (Pietro 2007-10)	0.0005681818
Marconi (Tommaso 2007-10)	Mustacchio (Simone R. 2007-10)	0.0005681818
Araujo V. (Abubakr 2012-14)	Luan (Tom two 2012-14)	0.0007102273
Kerlon (Mark 2007-10)	Renato Augusto (Michael H. 2007-10)	0.0007102273
Alan Kardec (Pietro 2007-10)	Dentinho (Mattia 2007-10)	0.0009943182
Carlos Eduardo (Gianmarco 2007-10)	Fernando (Abubakr 2007-10)	0.0009943182
Di Santo (Daniel O. 2007-10)	Piatti (Generoso 2007-10)	0.0009943182
Schelotto (Pietro 2007-10)	Sciacca (Tommaso 2007-10)	0.0009943182

The farthest couples all involve Frey and a young forward, or Batalla and an old one.

Let’s check the best (by predicted score) players:

Great player	Closest player to him	Distance
Neymar (Jesus 2007-10)	Maicon (Riccardo 2007-10)	0.1018465909
Muller (Daniele 2007-10)	Baumjohann (Simone G. 2007-10)	0.1369318182
Ozil (Michael H. 2007-10)	Aydilek (Jacopo 2007-10)	0.0132102273
Kroos (Matthias 2007-10)	Ekici (Pietro 2007-10)	0.0113636364
Gotze (Andrea V. 2010-12)	Merkel (Binder 2010-12)	0.0144886364
Ozil (Michael H. 2007-10)	Aydilek (Jacopo 2007-10)	0.0132102273
Busquets (Sid Debgupta 2007-10)	Martinez J. (Michael H. 2007-10)	0.0068181818
Schurrle (Riccardo 2010-12)	Lasogga (Binder 2010-12)	0.0573863636
Higuain (Andrea V. 2007-10)	Zarate (Generoso 2007-10)	0.0379261364
Cazorla (Daniele 2007-10)	Zapater (Mpfat 2007-10)	0.0258522727
Di Maria (Generoso 2007-10)	Blanco (Mattia 2007-10)	0.0042613636

Weekly battles

The latest craze in my email circle.

Daniele vs. Andrea V.:

• Scuffet vs. Perin: Scuffet on the bench, Perin started but had three goals against from Sassuolo… 0-0
• Marquinhos vs. Varane: Marquinhos on the bench, Varane played… 0-1
• Marchisio vs. Montolivo: Marchisio won 4-0 vs. Verona, Montolivo lost against Atalanta… 1-1
• Pogba vs. Koke: Pogba scored a nice goal and won vs. Verona, Koke won vs. Granada… 2-1
• Mueller vs. Goetze: both in winter break… 2-1
• Icardi vs. Dybala : Icardi didn’t score vs. Empoli, Dybala scored away vs. Roma… 2-2

But I have to admit that Cazorla and Lacazette, not included here, shined…

Alberto vs. Andrea V.:

• Bardi+Leali vs. Perin: Bardi on the bench, Leali and Perin had three goals against each (respectively from Torino and Sassuolo)… 0-0
• Banega vs. Gago: Banega was subbed off after 60′ with Sevilla winning 1-0 against Malaga, the Argentinian league is on break but Gago wasn’t even on the bench for the last six matches… 1-0
• El Ghazi vs. Boetius: El Ghazi played just 9 minutes (even though Ajax got their second goal then), Boetius scored in a winning match vs. Twente… 1-1
• Denis Suarez vs. Deulofeu: Suarez came off the bench (for Banega!) and scored the second goal… but Deulofeu had assisted the first one! (and he started)… 2-1 (I feel generous…)
• Lavezzi vs. Higuain: Lavezzi played just the last 10′, Higuain scored the wonderful goal that meant an away win vs. Lazio… 2-2 (but El Pipita deserves more than one point!)

Social play before picking

The social part of Fantasy Scout is usually what happens after you pick a player: you comment his (i.e. your!) highs and lows with your fellow scouts. But you can’t talk about the players that are in your watchlist; otherwise, now that they know that you are interested in those players, the other scouts could have one more reasons to immediately pick those players before you!

It’s a shame that when you really play (i.e. decide whether to pick a player or keep waiting) you can’t really share it with anyone else. And I guess there are ways to change this. A radical example is a bet I had with Daniele some time ago: in the form proposed by Jesper, “the loser must find N eligible players from which the winner chooses which player goes on the loser FS roster”.