Can maths tell us how to win at Fantasy Football? - Joshua Bull

[@rajat_singla]

To save your time, this is what he mentioned: *Team strategy* 1. 5-6 costly players + cheap fillers _over_ 1-2 costly + mid range 2. Bench substitution - players @Home _over_ players @Away. 3. Formation is not really a factor (choose best based on your players) *Tranfer strategy* 1. Form _over_ fixtures 2. Transfer out underperformer _over_ Transfer in last week's best performer/jumping a bandwagon *Captain strategy* 1. Fixtures _over_ form 2. If a tie: home _over_ away

[@sooohaib92]

Big caveat for home vs. away this season is lack of fans. I wouldn't advice considering it as a metric.

[@rajat_singla]

@@sooohaib92 makes sense. I am also a bit sceptical about it. Fixtures makes total sense. Home advantage i am not very sure.

[@ayozero9190]

I think Home advantage dosent exist anymore due to this stop start season? So the better teams usually win regardless of where they're playing.

[@nitinkymar9468]

May god bless you

[@Lewis-vg5uu]

@@sooohaib92 I disagree a little...because of the distance needed to travel on an away day in most cases. Sitting on a train to Newcastle, going to an area, pitch, changing room you don't know...all impacts too.

[@MKronos]

Coworkers passing see me watching an Oxford Mathematics lecture when really I'm just planning FPL strats

[@supernova3658]

😂

[@elvan12581]

What a Boss! hahaha

[@arthurpires8103]

hahahahah

[@neversayxever]

I'm 10 mins into the video and I see myself winning it this season already

[@lackadaisicalll]

He haven't even spoke about fantasy in the first 10 minutes 😑

[@simonnorthover]

not Pablo Neruda - If that what you think, you definitely ain’t winning fantasy this year

[@lackadaisicalll]

@@simonnorthover so speaking about modelling is inspiring to win fantasy .. it didn't hit me this way sorry

[@Shenoizy]

I found this genuinely interesting, not just for the purposes of FPL but because it shows a real world example of how maths can aid us to solve complicated problems. Thanks.

[@HappyCodingZX]

Most people try to find a way to make their work more fun. These guys find a way to make their fun more work.

[@nickwoolley7368]

Glad you are the arbiter of whats fun...

[@HappyCodingZX]

@@nickwoolley7368 each of us must entertain ourselves in our own way

[@bitmixes]

As a Graduate Researcher and an FPL patron myself, this video was right in my sweet spot. Thoroughly enjoyed the video. Also, roasting Norwich on every other slide was epic.

[@kray9438]

“How fantasy football helped me beat cancer” - looking forward to the day you release this talk Josh, I wish you all the best in your fight against this disease.

[@alexmitchell4374]

Good watch, a very humble and funny guy!

[@matthewmcmahon684]

People writing summaries of what he said, but the video was actually very interesting and enjoyable, nice little bit of humour too.

[@paulk4797]

This is by far the best video about Fantasy i have ever seen. And Joshua seems such an enjoyable guy to be around.

[@kemmishb1990]

This is excellent Josh. Two of my favourite things colliding - maths and FPL. Lots of time spent modelling this I am sure! Good kuck 20/21.

[@LvvLndv]

I understand the whole video as I did Maths in my Master. It turns out that's the first time that I have found what I learnt was useful.

[@blehblehbleh3557]

Imagine being Mo Salah and finding out that no one is captaining you in their FPL cuz of a mathematical model

[@saurabhkharbanda3242]

🤣🤣

[@Kvh47]

I am not captaining him because he looked weak during preseason. He is still a great player but i dont expect him to do anything special right from gw1

[@Darwizzy796]

Kvh56 oh no

[@Darwizzy796]

Kvh56 I captained him btw

[@Kvh47]

@@Darwizzy796 hahaha i believe i messed up

[@tkerby91]

Really interesting talk - I'm not very mathematically-minded but I've always been interested in how to use maths within FPL

[@DarraghMurray]

This is a really valuable lecture. Thanks Josh!

[@No-Gumph]

Thank you very very much for sharing so much time and insight. A fascinating talk. What I found most interesting is how playing the game (I.e picking a captain or deciding who to transfer in/out) involves having a hypothesis and reviewing the outcome. New players to the game quickly recognise the factors to consider (price, player form, team form, opposing team form, home/away), but weighting these is where the modelling seems to provide a distinct benefit. One thing I thought about is whether player form is a result of past fixtures and actually isn't a good predictor of future player peformance. I also thought about whether teammate chance creation / conversion is a variable to add in a more complex model. The best of luck (and decision making) next season.

[@qwobify]

Fantastic lecture! Loved the quotes as well - I never knew Ada Lovelace was such a keen FPL player!

[@FishBarney]

And creator of a cryptocurrency. All around good egg.

[@leedsmanc]

She needed to play some sort of computer game after inventing computing itself, FPL was the obvious choice.

[@Velvet_Torpedo]

thanks for doing this Josh really enjoyed it, and congratulations on last years win!

[@akmalrodzi4633]

Enjoyed this, this is the first season I'm actually taking FPL "seriously". I'm taking a conservative approach but trusting my inner scout instincts. Imho, we can't control the uncontrollable. So, we control what we can. The only constant/data that we can control in my humble opinion is : (1) Fixtures (2) Player Prices (3) Team Setup ; tactics - offensive, pragmatic etc. So this season I've set up my team according mostly 'pro fixtures' and considering player swaps with similar prices ranges. I definitely think fixtures play a bigger role than form because teams nowadays tactically better, defending better as a team etc. Other than that is up to the players to perform. I hope I get to top 10k this season. 😁

[@mul25]

good luck mate

[@akmalrodzi4633]

@@mul25 thanks mate, u too 😁

[@shambhav9534]

1 tip from my experience, choose players according to next 5 fixtures. God players: KDB Bruno Any goalscoring midfielder Defender who starts all the games.

[@akmalrodzi4633]

@@shambhav9534 yea, I looked at the fixture, imo Man Utd & Chelsea are nicely spreaded out till end of the season. I'm going KdB, Martial, Werner, Havertz/Pulisic, Greenwood till the end

[@shambhav9534]

@@akmalrodzi4633 Don't keep Werner, it's his first season and he will not score like 30 goals but around 20-25, which is excellent, but not in FPL, forwards only get 4 pts for a goal compared to 5pts of Midfielders and wingers are also considered midfielders. So, My recommend players: Bruno(He will take all penalties) Aubameyang Greenwood Rashford Pulisic Sterling They all are goalscoring who will on average score 20 goals, that is 20×5 = 100 + they will start often like 30 matches = 60 pts, which is 160 pts and assists which will get them to around 200 then bonus pts will get them like 230. And defenders: Maguire Every Everton defender(they have got easy games coming)

[@bw4500]

Really likeable guy, really well explained. Thanks Joshua 👍

[@nialloconnor6477]

Amazingly well put together, thanks so much Joshua!

[@danfrancis7196]

I love the quote attributions!

[@judgecrow1611]

Alright, notes were made, plan is set. I'm keeping this lecture secret from my mini league rivals. :D Time form my revenge because of 19/20 season

[@somhaud]

Basically, what I learned from this talk is that a block of wood is worth the same as a community shield

[@stuartfraser8060]

Not true - it's worth more, but it has the same value per kilo (in this scenario) ;)

[@kray9438]

I love maths now. I used to find it “hard” and this quote 8mins in blew my mind: “If people do not believe that maths is simple, it’s because they don’t understand how complicated life is” Nowadays I love mathematical modelling and solving problems using a good spreadsheet and thinking this way definitely helps. Had my best ever season last year top 25k and won a few MLs. My problem is I take too many unnecessary differential risks and my capt choices have been terrible eg this year gw1 I capt mane instead of Salah - backfired spectacularly.

[@abhiroopdey]

Did you at least make that up in the second gameweek?

[@kray9438]

@@abhiroopdey I’ve been doing a bit better since. Overall 795k but this season is throwing up some very weird results! I’ve WC early and much happier with my team now. How are you doing?

[@gdognz79]

This straight up hurts my brain , and my chances of winning my league

[@ankurakashady]

Josh, what a great informative video.. Thanks for the time and efforts

[@abc-qv1pe]

This was very enjoyable, thanks a lot.

[@jeffjones114]

I think it is important to figure out an exact number of "expected points" a player will get. I'm quite surprised there is no discussion about this. Maybe he didn't want to go into the numbers too closely.

[@stuartfraser8060]

Thanks Josh, enjoyed - bit concerned that all of this is based on one season, it was rather bound to prove your preferred strategies right using the data from the season you finished first no? Would be much more interesting to test these (form/fixture/home&away etc) strategies over multiple seasons, see whether they hold firm throughout and/or whether there are any trends over the years.

[@joecampbelluk]

First, congrats Josh on an amazing achievement! I came to this lecture in the hope of improving my FPL performance (time will tell 😜), but left with a facinating insight into a way of using maths that I haven't come across before. Kudos. Very keen to pick up on the simulation aspect of modelling as this seems to be where the power lies. How do you go about simulating 10,000 results? Does it require specialist software only available to university mathematicions, or is this something that anyone can do using excel or other free software? Also are you just feeding in raw fpl data, or are you inputing odds based data to guide the simulation? Would love to know how to get started with learning this!

[@Nihalmoneymaker]

He will be using programming languages like python to simulate.

[@avinashsekar2206]

This is brilliant

[@joni944]

The most useful and helpful talk on FPL that I've ever listened to 👏🏻👏🏻👏🏻

[@THEchrisse95]

Dear Joshua/Oxford Mathmetics, are you willing to test certain hypothesises by playing the game with multiple accounts? For the science to see how beneficial it is in practice? As I get that with statistical/Mathmatical modeling you can have probabilities which are most likely to perform based on 10000 monte-carlo experiments. However with the data based on last year it would be interesting to see a season played out by comparing your own personal account VS the university accounts who are restricted by set conditions from the start.

[@nigefal]

Multiple accounts are illegal in the game and can get you disqualified. He won't want that!

[@THEchrisse95]

@@nigefal I was suggesting it for science purposes, if they have for example 7 scenario's they can use 7 researchers in the project to avoid breaking the rules. Moreover, they could've contacted FPL to ask if it was possible to do it for research and not for any monetary gain.

[@Banzybanz]

@@nigefal They don't have to actually make accounts on the site. Just write a programme to simulate it all.

[@rennofiraldy8861]

@@nigefal huh really? Some dude in my mini league did this

[@konzi1]

Brilliant video!

[@paulmeahan3667]

Really interesting talk! Lots to consider going forward. Good luck to you this season, Josh!

[@NeonLightification]

Great video, would have liked to see an analysis of set and forget captains vs changing captains.

[@sakriyaneupane706]

My takeaway: I need to get into Oxford to win FPL.

[@sebbes6760]

Hi, great presentation and super interesting. Any thoughts on players new to the premier league? (Transfers from abroad) 😅

[@tletsoalo]

Great video. Thank you

[@danielburgess7101]

Gonna let you lot in on a secret strategy that I devised Chose players in pairs, normally a midfielder and attacker from one team. Son scores? Kane more than likely assisted him. Vice verse. James and DCL. Grealish and Watkins is my inside shout this year. It means when you get a goal it’s likely that it comes with an assist as well. One team has a stinker one week, at least you have other pairs to hopefully compensate.

[@lukesargent3859]

Brilliant, insightful video - thank-you.

[@streetspirit956]

Josh I'm interested to know. What is your FPL history like in terms of rank finishes? How many top 10k finishes, how many top 1k finishes etc?

[@dassteam]

14/15 2.7m, 16/17 3.4m, 17/18 4.2m, 18/19 1.3m, 19/20 1st... So he got lucky, very lucky.

[@Szczych11]

@@dassteam that's interesting but I don't think you can win the entire fpl with luck. I think it's more likely he devised this model and used it in 2019/2020 and didn't take the other years seriously.

[@dassteam]

@@Szczych11 Those are very very poor results prior to winning, I'd find it surprising that he didn't take it seriously at all for 4 years and then suddenly did and wins the whole thing. Getting 1m + ranks and then winning it one year does suggest he was very lucky even if he did start taking it seriously. I would agree though you cannot win fpl with only luck.

[@dassteam]

I think his rank this year will paint more of a clearer picture.

[@dassteam]

As an update Josh is currently 157,909th

[@PEGGLORE]

My dad's best friend from Cambridge university formulated the abc conjecture after his phD, and is now Professor Emeritus of Mathematics and Computer Science at the University of Basel. He's good at Maths. Wonder if he'd be any good at FPL?

[@rgqwerty63]

Hi, I think a nicer way of modelling budget would have been assigning a value of 0 for minimum price - that is 4.0M GK/DEF and 4.5M MID/FWD cost 0, then 1 extra value for each 0.5M extra. For example 4.5 defenders have a cost of 1, 6.0 defenders cos of 4. 10.0 midfielders a cos of 11. This almost perfectly models the actual budget in the game and is easy to implement

[@_uky]

This information is valuable. Thanks for sharing Josh. It'd greatly appreciated if u could also share how u test and simulate your hyphostes ☺️🙏

[@TANK123cs]

The first sir bobby quote had me like Whaaaaat😂

[@jskelley1104]

This is excellent

[@jordieclive]

Don't want to take away, as obviously made some really smart observations. But seems very manual, and not data crunching enough-looking at sns violin plots, rather than letting algos learn these things. Would be really quite interesting to apply Machine Learning with some operations research type constraint theory, added in. I imagine it is similar to how renaissance technologies or successful current algos are doing the stock market. Obviously this would be taking it to an extreme level. I imagine reinforcement learning running simulations on past seasons, and keeping very specific weights for teams current form. But more than the stock market the way football has evolved, rules have changed etc. would be interesting, so would have to weight learning a lot for last few seasons. Also the data avaialble in previous seasons wouldn't be as strong, as I'm sure it is now.

[@husseinmohammed8654]

Science 😍 .. thanx Dr. Jushua

[@jakelamotta456]

Thanks for great content. Regarding the home/away choice as tiebreaker not doing much difference, isn't it be as home/away is already taken into account in the FDR? A home 2 might be a 3 away, so the home team might be playing a worse opponent than the away-team, for them to be facing the same FDR-rated opponents. Hope it makes sense =)

[@lucagastoldi6509]

Thank you Joshua. Very interesting video. I could not catch the name you thank at the end of the video for posting all the data you use on github. Would you mind telling me please?

[@OxfordMathematics]

Hi - the address is on the slide.

[@GAFFRSHORTS]

Is home fixtures over away fixtures really still a factor with no fans in the stands? Is there any research on home vs away fixture during the restart period?

[@nolikeygsomnipresence270]

Fantastic and engaging, thank you from Chili

[@TheSeasonKeepers]

Great video Joshua!

[@bigchicharito4813]

I loved the quotes on the bottom xD. Btw , how did you notice how the bias in the new set and forget model came. Like once u saw that more defenders were doing better in your model, how did you know the reason for the bias? Thanks

[@B4dd3r5]

23:44 .... A violin plot... clearly

[@ElGrebr]

Great video!

[@sarath_sajan]

"I transfer in players who wear prime numbered shirts" 😂

[@fredtaylor8086]

Great video Josh. Thank you for sharing. Did your transfer strategy simulations take into account the 3 players limitation?

[@olliert4840]

Do you have access to the fpl api for this or whatever in order to make these models?

[@indraneeldeshpande3927]

In Vardy's example, can you tell what is the way that you can get the distribution and avoid overfitting when one has less data points?

[@CagedPaps]

There's one main number they're getting wrong though, the intro says around 8 million players, then Joshua says about 7.5 million. The reality is it's a fraction of that. You'll find thousands of teams created by sites that aim to win the highest score for just 1 gameweek and the manager of the month, there's prizes for both these. I'm not saying this makes it easier, but if you model the people you're competing against you need to remove the noise.

[@roym2185]

As a Fantasy Football manager I also have a question. Is there a mathematical model to determine how much time is wasted playing this game? Edit: Enjoyed the talk, the humour, great stuff!

[@europa1387]

Time you enjoy wasting is not wasted.

[@Veaseify]

'Pick Kevin De Bruyne' - attributed to Blaise Pascal (1623 - 1662) an axiom that has stood the test of time...

[@ADAJ3KINGANGEL]

I wonder how COVID-19/No - limited fans affects home/away performance over the course of the season. The problem I see with the team I chose is it’s too balanced.

[@telquel7843]

Let's see if he wins it again before we all put our faith in this. A lot of it is common sense, but the notion that modelling alone could secure you a win in FPL seems specious given the unpredictability of the game. That said, most people would perform far better than they do now if they were to emulate this approach (myself included).

[@gustavjohnsson8074]

33:00 What are those “diagrams” called now again. Completely forgot

[@elamanov]

Box plots

[@juanernestolima1595]

Jeez this guy is smart

[@shaunhenderson3877]

Good video very informative.

[@vilmervrennhage5552]

Lovely Guy, 💕

[@michaelginting1783]

24:20 anyone would help me to tell whats the diagram called?

[@trevorvanderwoerd8915]

Violin plot

[@joetaylor7363]

Some nice ggplot work there Josh!

[@nickagger534]

Josh, can I have your FPL ID for the new season? I like checking certain peoples teams every week to see how people are doing 😄 thanks

[@ankurakashady]

The Bulldozers

[@nickagger534]

ankur srivastav his ID not club name

[@ankurakashady]

@@nickagger534 oops, sorry mate.. On a side note.. If you find his ID, please do share here :)

[@nickagger534]

ankur srivastav once the season starts I’ll go through the mathematicians league and find it that way for everyone😃

[@Adriano-ni2ol]

Nick Jones found it bro?

[@intense10]

sir i have question to you i want buy a best book for Differential equations. Please tell me which book bast for me .I am 2nd year student

[@intense10]

@@MijeshDeuja Thanks very much sir

[@lughaidhmoutia3589]

HOW MANY POINT DID HE END UP GETTING

[@hazzanfl9814]

What was the code

[@hossam2508]

Whats ur squad this season 👀

[@fluntimes]

He is doing reveals to Fantasy Football Hub members.

[@clipmania1717]

how does he sim the games?

[@ahasiikuyu8001]

To be honest my first GW 20/21 can't BELIVE I captained Werner instead of salah those fpl pundits kind of talked me into it I could have been flying by now

[@coachtechnique]

im a coach im hope that i can learn something

[@dav9766]

De ley salgo campeon ahora, crack!!

[@streetspirit956]

Mmm I have just seen Josh's FPL history which leaves a lot to be desired. In the 4 seasons prior to winning it, there isn't one rank finish inside the top million which makes me think that winning it last year was more down to a lucky one off season rather than some mathematical formula.

[@Banzybanz]

Or possibly he has improved over the years? I've been playing FPL for 12 years and I've been complete rubbish for 9 of those season. Then I had 2 200k finishes and finally a 55k finish last season. This time I am aiming for top 10k. As you improve, you notice all the stupid mistakes you made in the past, and you also get access to better tools to help FPL decision making.

[@paddydoyle2324]

He actually has said many times that for years he never put any real time into FPL except last year and the year before. He made teams because his family wanted him to but never really made any changes or put any thought into them throughout the season. I don’t believe he’s ever said it’s down to a “mathematical formula” he just applies maths to FPL in recent seasons. Let’s see how he does this and decide if it was fluke, but I’m sure he will do well again however it is very unlikely he’ll win. Also he has said that he did get lucky this season and I think anyone that gets rank 1 overall has to have luck. But yeah his history is very poor so let’s see this year.

[@Ayra0731]

Nice video

[@thegoat7360]

great lecture josh, cancer beware.

[@mattiseriksen976]

Watching this 10 times and I will still pick the wrong captain

[@shambhav9534]

16:18 You meant Lundstram.

[@dlon8899]

The Var decision makers, have a higher chance of supporting Manchester United and Liverpool. Biased Penalty decisions proved this theory

[@AmazingGolazo]

I thought this guy was Paul Dano in the thumbnail...

[@nielsus6]

Why am I late to this party... :)

[@thetntun]

GW 1 n 70 points , now give me math strategy 😅 for all weeks

[@shreyanshjain853]

Can we use data to win another game? Cricket

[@shreyanshjain853]

Need help if anyone would be interested

[@maherhussein1999]

I definitely wouldn't have called that a "violin plot"

[@intense10]

hi Oxford Mathematian

[@obafemiadeyemi]

Can you model this into a software?

[@acehighjohn1759]

For a mathematician the F.A cup bit shoulda been easy peasy lol.

[@rameshsoren5197]

Hi ..

[@Mucha.offical]

0:32 sounds like the miaka guy

[@MrSamshy]

He exposed us 🤦🏻. Mathematicians shouldn’t talk about these things 🤫

[@74Jupiter]

No wonder he won it, he had 10000 teams entered.

[@angelgospodinov3215]

if we have to be mathematically accurate you end up in second place. The first one had more points than you, he was just disqualified for mistranslating a word from his language.

[@shambhav9534]

Simply use goalscoring Midfielders. Eg. Rashford, Aubameyang etc.

[@farmer__]

*inserts Unai Emery meme