To see less ads Register or Login ----- Daily Fantasy Sports games 18+

Analysing strategy for the run-in: The wildcard strategy league

A Fantasy Football forum for news on fantasy football games run by the Premierleague (FPL).
JoonaR16
Grumpy Old Man
Posts: 1654
Joined: 14 Aug 2014, 14:34

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by JoonaR16 »

Those rounds between 34 and 37 were also very important. I don't know if I was lucky but late WC felt much better option for me. I had lots of time to plan my GW34 team and in GW35 I still had full squad except for goalkeepers. GW36 was for WC and I managed to get great results from that round too.

User avatar
Mr Clarinet
Grumpy Old Man
Posts: 2253
Joined: 30 Sep 2008, 17:35
Location: Athens

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Mr Clarinet »

Well, as we meander towards the somewhat anticlimactic conclusion of the Season 15-16, I thought I’d contribute a line or two while waiting for the final points to trickle in.

At the outset, it seemed to me that many managers were facing a binary decision relating to use of their second WC, either ‘early’ (GW32/33) or ‘late’ (mostly GW36). This situation arose from the way that most of the postponed fixtures had been re-arranged into just two GWs, 34 and 37 (a situation that may be unlikely to re-occur?). In the lead-up to making the WC decision (and the related deployment of chips), I found myself struggling to work out what was likely to be for the best (for my own team), and it occurred to me that others may be similarly perplexed (in relation to their own teams). Hence the ‘Wildcard Strategy League’. Thanks to everyone who joined in, and to all who have contributed to the thread.

As it happens, my ‘day job’ involves the mathematical evaluation of decision models, and I have in the past (up to the end of Season 10-11) used what expertise I have in this area in playing FPL. But tbh it took the enjoyment out of the game, and these days I just play for fun, just hoping to compete well in my main mini-league. So I’m not going to bring any heavy-duty analysis to the data provided by the League (and there are likely far-better qualified FISOers to do so, anyway). So, any insights gained will have to be clear-cut. And hopefully, some entertainment will also have been had by those joining in.

What I wondered, initially, was whether either of the two main strategies would turn out clearly better than the other, on the basis of what we could see from a league table over GWs 32-38. As was quickly pointed out, there are obvious difficulties with this. In such a data set, there are many factors that are uncontrolled on entry to the trial, itself based on a self-selected sample. As a result, I thought that clear-cut insights were unlikely to emerge, that differences between the signals representing the two strategies would be drowned by the noise from all the extraneous sources of variation. Whether any convincingly clear-cut insight emerges, I think depends (for most) on the addition of the final points, taking into account the following (informally expressed) considerations.

How, then, might you informally evaluate the outcome when the final scores eventually emerge? This is my version. Your initial position (‘early’ or ‘late’ as the better decision) can be represented by your (subjective prior) probability. I could do no better than 50%/50%, but eventually came down in favour of ‘late’ (for reasons connected with the way my main mini-league was going). Others had a prior that expressed their view that of one or the other of the two main strategies was favourable. Some even went as far as a prior of 100% for their adopted strategy.

The league scores represent evidence, which you combine with your prior probability to reach a posterior (conditional) probability of your adopted strategy being the favourable one, given the evidence. Since I started from an initial position of maximum uncertainty, my evaluation depends on the evidence (this is what those of you who saw the movie 'The Imitation Game' were watching happen, although the underlying inference problem was – understandably – not given great prominence). At the other extreme, for those with a prior of 100% for their adopted strategy, no amount of evidence can alter their prior. As can be seen from many of the comments posted in the thread, different subjects have different views of the outcome (so far), and this is because they have different priors and attach different weights to the evidence.

One bit of statistics I thought might be interesting concerned the extent to which the league scores (GWs 32-38) were related to the game score for GWs 1-31. That is to say, were the players with the best WC/chip strategies simply those who were the best strategists anyway, as judged by their performance in the season so far? Up to now (GW38 scores pending), there is a weakly increasing linear trend for league score on GW 1-31 score, but it is very small compared with the residuals about the trend. I will finalise this once the GW38 points are added, but what this seems to suggest is that if you can find and deploy a favourable WC/chip strategy during the run-in, large gains are possible even in a so-far mediocre season. An incentive to keep playing...

I'll post final positions and my own resulting thoughts later in the week. In the meantime, please continue to add your own thoughts. Thanks again for joining in.

Mr C

User avatar
Mr Clarinet
Grumpy Old Man
Posts: 2253
Joined: 30 Sep 2008, 17:35
Location: Athens

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Mr Clarinet »

Here's the Top 30 (/61) (a nice round score of 500 points over 7 GWs makes the top half of the league). After points, ties were decided on fewest transfer hits over the period of the league. A zero in the data indicates a chip used before the league period. After GW37 it seemed a pattern may be emerging, but although the top 3 in the final standings all used the same strategy, overall there doesn't seem to me to be any clear-cut evidence that either the early or the late WC was obviously the favoured route to success. Of the top 10, 6 were late users, 4 early; of the top 20, 10 were late, 10 early; and of the top 30, 13 were late, 17 early.
I still want to have another look at the relationship, if any, between league scores and scores in GWs 1-31.


1 Surrender Monkeys 592 36 37 34 (Total points, then GW for WC, BB, TC)
2 Niskala FC 576 36 37 34
3 Blacksunrise 567 36 37 34
4 Fuzzy Duck FC 566 32 34 37
5 Armes Tuk Tuks 564 32 34 33
6 ian the girth ©® 559 33 34 37
7 FC Schmengellow 2015 557 36 37 34
8 The Surprise Package 553 32 34 37
9 HQM 551 36 37 34
10 Don'tcallmeSchürrle 550 36 37 34
11 Is your Cresswell? 548 37 0 34
12 Chopin FC Bart Bara 548 36 37 34
13 Lads on Toure 543 33 34 37
14 McSharry's Form 5s 543 33 34 37
15 Steamshovels 542 33 34 36
16 Nottingham Forest 538 36 37 34
17 Jelle & Dyche Dream 536 36 37 34
18 Jacko's Jokers 530 33 34 37
19 Los Goleadores 528 33 37 34
20 CTRL ALT DE LAET 523 33 34 37
21 NorthBeach Dadbodies 518 33 34 37
22 Can Ibe your fantasy 514 36 37 34
23 Forest Badgers 514 33 34 37
24 Red Star Morningside 508 36 34 37
25 Blind Over Mata 508 32 34 33
26 Aldershot Rejects 508 33 37 34
27 Rob's Dogs VII 507 36 37 34
28 AC Mullan 506 33 34 37
29 Away's Count Double 505 33 34 36
30 Ayew Shaw? 500 33 34 37

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Stemania »

Great stuff Mr C. Really good read the last few pages. :D

Notned
FISO Knight
Posts: 11198
Joined: 13 Sep 2013, 12:30

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Notned »

Great stuff, Mr C!

Fourth, and the highest placed early wildcarder.. I'll take that!

User avatar
BobMem
Grumpy Old Man
Posts: 4222
Joined: 24 Dec 2010, 08:54
FS Record: The Graduates - FISO 5AS H2H Champions 13/14
FISO King of Rock and Roll - Elimination game winner 2015/16

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by BobMem »

Great analysis Mr C. Very enjoyable read.

Also thanks for doing the top 27 in you league table ;)

User avatar
Mr Clarinet
Grumpy Old Man
Posts: 2253
Joined: 30 Sep 2008, 17:35
Location: Athens

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Mr Clarinet »

Here's the second half of the league table. For positions 31-61 there's almost a 3:1 split of early:late WCs.


31 everythingbutthegoal 498 36 37 33
32 stuboy united 498 36 37 34
33 Temple's 'Owlers 497 33 34 37
34 Neilers Nailed On! 496 33 34 37
34 Gingangoolygingantoo 496 33 34 37
36 Regretful Monkeys AC 495 36 37 33
37 Half Man Half Okapi 494 36 37 34
38 Ibe gotta feeling 493 33 34 37
39 Gav's Gunners 492 33 0 34
40 Dynamo Jag 490 33 34 37
41 Norfolking Good 488 32 34 37
42 BARLEY FK 488 33 34 37
42 Useless Generation 488 32 34 37
42 Billionaire 488 33 34 37
45 Chamakh My Pitch Up 485 33 34 37
46 No Weimann No cry 485 0 34 37
47 Ayew kidding me 485 33 34 37
48 Schrodinger's Cat 479 33 34 37
49 Bringbackthebiff 473 0 34 33
50 The Little Horse 470 33 34 38
51 Oldale's Army 465 0 34 33
52 Who Let The Hogs Out 464 33 34 37
53 Butter Bridge United 462 36 37 34
54 Norfolk-in-chance 460 32 34 37
55 Obvious Child 457 32 0 34
56 Hodgys Hotshots 453 36 37 0
57 Ali's Allstars 453 32 34 33
58 Sharknado 450 33 34 37
59 Tokyo Sexwale 445 33 34 0
60 Utakadefoekingpearce 420 36 37 33
61 Kuban Krasnodar 393 0 0 33

User avatar
Mr Clarinet
Grumpy Old Man
Posts: 2253
Joined: 30 Sep 2008, 17:35
Location: Athens

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Mr Clarinet »

As I mentioned previously, there was a weak increasing linear trend of league points (GWs 32-38) on points for GWs 1-31. Although weak, the trend was consistent over the weeks that I monitored it, in that it didn't vary much as new data were added, week by week. An interpretation of the league results viewed via this trend is shown in the attachment. To produce this, I calculated the residuals from the linear trend, then plotted the residuals (in units of points) against league rank. To read the graph, looking left to right shows the ordering of teams by league ranking. Looking top to bottom shows the ordering by residual, or "gain over trend". Because the trend is only weak, the two orderings are similar; the teams furthest to the left (higher rankings) tend to be nearer the top of the plot (higher gains over trend). But you can pick out little details too. For example, Fuzzy Duck FC (ranked 4th, reading left to right) were ranked third in terms of residuals (reading top to bottom) - that is, they had a larger gain over trend during the period of the league than the team ranked 3rd in the league. The other thing the graph shows is the extent of gain over trend that was possible. The team that topped the league gained 93 points over trend during the period of the league. The average for the top 5 was 69 points (rounded to the nearest whole point).

[click on pic to enlarge]
WCSL League.jpg
You do not have the required permissions to view the files attached to this post.

User avatar
Mr Clarinet
Grumpy Old Man
Posts: 2253
Joined: 30 Sep 2008, 17:35
Location: Athens

Re: Analysing strategy for the run-in: The wildcard strategy league

Post by Mr Clarinet »

Did we learn anything? This is my own view.

1. The league data provide no clear evidence that either early (GW 32 or 33) or late (GW 36 or 37) WC during the run-in was the definitively favourable strategy.
2. Performance (measured by points scored) during GWs 32-38 depended only weakly on performance during GWs 1-31.
3. Large gains are possible over the run-in even if previous performance has only been moderate - but the idea that there is a simple one-size-fits-all way to realize such gains is chimeric.
4. Not directly linked to WC strategy, but personally I was surprised by the extent of paid-for transfers by league managers. Not counting free transfers, over the 7 GWs of the league, there was a total of 263 hits (average 4.3/team, range 0-12). This, of course, during a period when nearly everyone (58/61) had a wildcard to play, granting unlimited free transfers for one GW. The team at the top of the league took 0 (zero) hits.
5. Personally, I found this an interesting exercise, and it provided me with some useful insights on how my own strategy during GWs 32-38 compared with that of others. To my own amusement, I finished exactly on the trend line (see graph in the above post) with a "gain over trend" of zero!

View Latest: 1 Day View Your posts
Post Reply

Return to “Fantasy PremierLeague.com (FPL)”