Betting - TTH's wisdom of crowds

[Reg]

Alright, thought I'd see if there's much interest in this.

If you haven't heard of the 'wisdom of crowds' have a read of this article (http://www.bbc.com/future/story/20140708-when-crowd-wisdom-goes-wrong). The paragraph below quite neatly sums it up.


The notion that a group’s judgement can be surprisingly good was most compellingly justified in James Surowiecki’s 2005 book The Wisdom of Crowds, and is generally traced back to an observation by Charles Darwin’s cousin Francis Galton in 1907. Galton pointed out that the average of all the entries in a ‘guess the weight of the ox’ competition at a country fair was amazingly accurate – beating not only most of the individual guesses but also those of alleged cattle experts. This is the essence of the wisdom of crowds: their average judgement converges on the right solution.

I thought it'd be interesting to see whether a crowd of TTH'ers would provide a good way of betting on football matches. The plan would be something like this:

We choose a league/team to focus on
Each week, a group of us each put forward our opinion on the chances of a team winning/drawing/losing (or another betting market)
I/we put small bet(s) on the average of our opinions
We find this doesn't work / we enjoy our millions


Thoughts/suggestions?

[Byron]

I'd be up for that. Could be a good laugh.

[Boydy]

Sounds like it'd be interesting to try. My own betting selections are terrible so this couldn't be any worse.

[Luca]

Correct me if I'm wrong, but is this not the basic logic behind Monte Carlo simulations/bootstrapping?

[Reg]

From a very quick Google, I can't see that it has anything to do with the average of a group of peoples' opinions/predictions.

[Toby]

Think we thought about doing this back when we had the weekly predictions of Premier League scores, but it never got going. Hopefully this will be different. The idea that time was just to continue posting score predictions and then taking an average of some sort from what everybody posted.

[Yevrah]

I've always thought about doing this and I think we'd win some money providing we steer clear of 'groupthink'.

[Reg]

Yep, we don't want to be swayed by each other's opinions. Perhaps PMs to me/somebody else each week would work well. Then an average of everybody's predictions would be posted in here for everybody to use however they like.

[Adam]

The last time I won anything worth bragging about was probably November 2013 so I'm well up for this.

[Toby]

Yep, we don't want to be swayed by each other's opinions. Perhaps PMs to me/somebody else each week would work well. Then an average of everybody's predictions would be posted in here for everybody to use however they like.

If you can combine it with a record of who makes the most accurate predictions it'll probably give an added incentive to keep posting them. Also gives the opportunity to do some interesting experiments with weighted averages depending on who has been most accurate.

[Reg]

Good idea that, Toby. :thbup:

[Waffdon]

I'll cover The Dutch and French second division. Easy wins

[Demerit]

Yeah, I'd be up for participating.

[Reg]

I'll cover The Dutch and French second division. Easy wins
Not sure you've got the idea of the thread(?).

[Waffdon]

Is that what you think? What is your name?

[Benny]

Sounds interesting, I assume we'll be sticking to Premier League/Champions League stuff?

[Reg]

Sounds interesting, I assume we'll be sticking to Premier League/Champions League stuff?
I think either/both of those would be the best idea. Unless people think choosing a specific team rather than competition would work better?


Is that what you think? What is your name?
:moop:

[Waffdon]

It doesnt matter what your name is

[Boydy]

A specific team would mean we wouldn't have that much activity though. Whereas if we did premier league and champions league we'd have about four days a week to bet.

[Jimmy Floyd]

Yep, can't be any worse than each of our normal betting and could be fun.

[Reg]

A specific team would mean we wouldn't have that much activity though. Whereas if we did premier league and champions league we'd have about four days a week to bet.
That makes sense to me.

I think this weekend/next midweek would be a bit too soon to start this. I'll give it more time for ideas and suggestions, then we'll start the following weekend.

I also need to sort an excel spreadsheet. I'll look into that as I haven't used excel in a long time.

[Foe]

I'd be up for this. Sounds a bit interesting and it'd be better than me just pissing a few quid a week away on my own.

[Benny]

Reg one of Team Spreadsheet would probably be able to knock you up an Excel template for it I'd imagine.

[Toby]

It's not like it's going to be all that complex. You just have to decide whether you're taking the average results, the average scores, or something in between. Average margin of victory with some sort of cap (e.g. two or three goals) would prevent anybody blowing things out of line by predicting 8-0s.

[Jimmy Floyd]

The average of any decent number of football score predictions will invariably be 2-1 to the favoured team, or 1-1 if a tight matchup.

[Nick]

Sounds interesting. Count me in. :thbup:

[Josh]

Yeah I'm interested too. It will be interesting to see if we do just end up with 2-1's like Jimmy says which seems logical or if we will come up with some stranger results. I do think the knowledge on here is far greater than the average punter but I don't think the bias is any different to your average Joe.

[igor_balis]

Yeah I like the sound of this, nice one Reg.

[Spammer]

Yep, im on this too

[Foe]

It's not really Monte Carlo analysis though. Monte carlo basically just does a huge number of trials of potential scenarios based on previous results, reliability etc. It's very much statistical probability to create a distribution of the 'most likely' outcome.

There's definitely similarities but this is much more gut feel.

[-james-]

It's pretty difficult to find an edge over the market in most leagues by just going off past results and statistics, but this kind of thing is definitely the way forward.

I think it's generally most efficient to follow leagues for which there a lot of information (teamsheets and previews) available for the majority of games, but obviously the more mainstream you go the harder it is to get in there before the big boys and syndicates. Croatian from 2nd tier down and German from 3rd tier down would be a couple of my suggestions.

[Josh]

Great idea, everyone starting learning how frequently NK Imotski lose at home and we are sure winners.

[Pepe]

Not sure this works as well in cases with discrete options that span a small range, essentially making it an either/or decision. Also, I don't think it is meant to be used for prediction. Still something that should be interesting to try and the bigger the number of people participating the more conclusive.

[-james-]

Great idea, everyone starting learning how frequently NK Imotski lose at home and we are sure winners.

Nah, but scan their team sheets every week until Mario Vojkovic and Dominik Radalj are out injured and you might be onto something.

[Baz]

TTH might win me money? It just gets better.

[Adamski]

Yeah, I'm in.

[Reg]

Looking for your thoughts on a few decisions lads.

1. The competition(s). I think the Premier League and the Champions League would work best.

2. Shall we predict results or scores as well? I'm leaning towards just picking results. I think picking a correct score is incredibly difficult and people may have ideas about common scores that aren't actually statistically true. I also feel we'll probably get more involvement the simpler this is.

3. Are there any other things people would like tracked on the spreadsheet, other than each user's success rate?

[Reg]

From starting to play with it on Google Docs, I've quickly realised that this will be beyond my spreadsheet skills...

Would anybody be willing to please set one up? (I realise I've led this and can't actually do the main task. :gs:)

If not, each week I can work out a simple percentage of home win, draw and away win, but there's not much else I can do manually...

[Baz]

Get on it GS John Arne

[Toby]

Can't you just assign values to home win, draw, away win and take the mode average?

Or mean and round to nearest.

[Reg]

It gets too complicated for an excel/google docs novice like me, when you start factoring in all the teams and weeks.

I'm very happy to just take the percentages home win/draw/away win predictions though, if we don't have a spreadsheet.

[Jimmy Floyd]

1 for a home win
2 draw
3 away win

Then anything between 1.67 and 2.33 would be a draw consensus (I think).

Just put all the numbers in a row and then do =AVERAGE(range), where 'range' is the span of cells they are in.

[Toby]

It gets too complicated for an excel/google docs novice like me, when you start factoring in all the teams and weeks.

I'm very happy to just take the percentages home win/draw/away win predictions though, if we don't have a spreadsheet.

It's literally just fixtures going down the rows, posters going along the columns. Then you start a new sheet or scroll down a bit for a new gameweek.

[The Merse]

I'm in.

[Reg]

OK cool, cheers lads. We can start next midweek if people are up for doing the Champions League games (thoughts still welcome on which comp(s) to focus on!).

[Jimmy Floyd]

The top tier games, because that's where everyone has the most knowledge.

[Spoonsky]

Gambling is illegal in Utah, so I'm afraid I won't be able to participate in this.

Just kidding, let's do it.

[Coffin n Fartin]

Count me in on this please guys

[Offshore Toon]

I'm in.

[Yevrah]

Not sure entirely how this will work, but if we're taking averages I'd urge people who have no feeling on a particular fixture to skip that one rather than just chucking something down.

[Reg]

What would people rather go with to kick this off in midweek?

1) Champions League fixtures where an English team is involved (plus any nominated teams - e.g. Barcelona, Real Madrid)
or
2) All of the Champions League fixtures

[Byron]

I'd say a select number of Champions League fixtures.

[Baz]

Number 1 please.

[Reg]

OK, here are this midweek's fixtures.

Please PM me your predictions (rather than replying in here), with a '1' beside the fixture for a home win, '2' for a draw, and '3' for an away win.

Barcelona v Bayer Leverkusen
Arsenal v Olympiakos
FC Porto v Chelsea
Malmö FF v Real Madrid
Man Utd v VfL Wolfsburg
Borussia Mönchengladbach v Man City

I'll then post in here with an average score of everyone's predictions around 16:00 (UK time) tomorrow.

[Boydy]

PM sent. I just realised this is split over two days though. Don't like that. :(

[Reg]

PM sent. I just realised this is split over two days though. Don't like that. :(
Why not?

[Boydy]

I don't like having bets split over multiple days

[Adamski]

PM sent but it might be less confusing if we did 1, 2, X like the rest of the world.

[Toby]

Have you had many responses? Presumably we'll get the averages before kick off tonight?

[Reg]

Have you had many responses? Presumably we'll get the averages before kick off tonight?
12 people have sent their predictions so far. Hopefully we'll get a few more but that's not too bad for a first round.

I'll post the averages around 4pm UK time. I think it'd be good to give people a few hours as people will be travelling back from work etc.
----


Consensus


Then anything between 1.67 and 2.33 would be a draw consensus (I think).

Having thought about it, I think this is about as close you can get to categorising the results. Any maths whizzes feel free to share your opinion; as far as I can see the above is pretty damn good.

Obviously an average of 2.9 is a more satisfying win consensus than 2.4. Also worth considering that averages in some cases are a truer reflection of a group of predictions (e.g. if half predict a home win and half predict an away win, then the average isn't so helpful - edit: if I'm correct there).

I can upload the spreadsheet, hiding usernames, each time.

[MartinSAFC]

I'm up for this. Great idea!:thbup:

[Reg]

Round 1 Results

Thanks everyone for your predictions (except Benny of course, who mooned me).

Here is the spreadsheet: https://docs.google.com/spreadsheets/d/1s47BHJYet7t97tXvumwiZgHYJucq5-FEogJWh0YRXVI/edit?pli=1#gid=0

Just a reminder that 1=home win, 2=draw, 3=away win.

For simplicity, I have bolded the teams who have a win consensus, and left each team unbolded in the case of a draw consensus (using the consensus categorising Jimmy posted which you can see two posts above).

Barcelona v Bayer Leverkusen - average = 1
Arsenal v Olympiakos - average - 1.066666667
FC Porto v Chelsea - 2.066666667
Malmö FF v Real Madrid - 2.933333333
Man Utd v VfL Wolfsburg - 1.333333333
Borussia Mönchengladbach v Man City - 2.466666667

[Toby]

May as well give this a go.

£5 treble on tonight's games at 6.57.

[Benny]

https://adragonseerie.files.wordpress.com/2015/02/627478-monty-python-life-of-brian.jpg

[Byron]

Go on then, Patent on tonight's games.

[Jimmy Floyd]

I'll put a tenner on each single and see how long it takes me to get down to zero.

[Vim]

Looks like apart from the Man City game, everyone voted for the same thing.

[Pen]

Went for the treble @7.1 too just to try this out.

[Vim]

I've put £5 on the treble. I was going to do doubles, but it's not worth it because if only Arsenal/Barca go through (the most likely outcome) then it's still a net loss.

[7om]

Not being able to bet online in Murica is really shit. I want to join in.

[Vim]

I've also done a 6-fold on all the games and 5-folds. Small stakes of course, just the money I had left over in my account. Can't really afford to deposit more so here's hoping for TTH to be right.

[-james-]

Bovada.lv

[Sir Andy Mahowry]

Looks like apart from the Man City game, everyone voted for the same thing.

It looks like I was the only one to vote against Olympiakos as well.

[niko_cee]

Bad crowd this.

[Jimmy Floyd]

I voted for Porto. You all know nothing.

[-james-]

Any shrewd opinions here are just going to get lost amongst the sea of average picks.

The spreadsheet had Arsenal as 1.06 favourites, and the draw a 1.4 favourite in the Chelsea game. It's probably a better idea to have people choose which side of the asian handicap market they'd rather be on, or something.

[Reg]

dino One bad round doesn't mean we shouldn't write it off.

I think any 'shrewd opinions' will lose out to consensus in the long run.

[Toby]

Yeah it does seem likely that we're going to lean fairly heavily on the stand out favourites (most games last night) or throw up a draw where the teams are more evenly matched (Porto v Chelsea). Tonight will be more interesting, as will the knockout stages and/or Premier League stuff if we do that.

[elth]

Arsenal v Olympiakos - average - 1.066666667
Man Utd v VfL Wolfsburg - 1.333333333


:harold:

[Jimmy Floyd]

Yeah it does seem likely that we're going to lean fairly heavily on the stand out favourites (most games last night) or throw up a draw where the teams are more evenly matched (Porto v Chelsea). Tonight will be more interesting, as will the knockout stages and/or Premier League stuff if we do that.
Might the mode be better than the mean then? For the first time in recorded history.

[Toby]

Wouldn't have made a difference this time around but I suspect when games are tighter it would reduce the prevalence of draws.

[Yevrah]

Averages like this won't work because the average will always be drawn to backing the bookies favourite and doing that will never make anyone any money.

It's more work, but we'd be better off picking a selection of fixtures each day/week and discussing them, out in the open, to find the ones that enough of us think contain value, then betting on them.

[Reg]

It'd be a shame if we ditched this after one round of fixtures, but if others would also be more interested in the above then I'd be up for that.

[SvN]

100% tonight lads.

[Yevrah]

It'd be a shame if we ditched this after one round of fixtures, but if others would also be more interested in the above then I'd be up for that.

No reason why we can't do both.

[Jimmy Floyd]

I won £100 on the treble :cool:

[Reg]

Although like you said earlier in the thread:


I've always thought about doing this and I think we'd win some money providing we steer clear of 'groupthink'.

edit: :cool: Jim

[Boydy]

For tonight's? What were the odds?

[Yevrah]

Fucking hell Jim. :chief:

I thought you only ever bet with a few quid on the short term stuff.

[Jimmy Floyd]

For tonight's? What were the odds?

I was going to do £10 singles as yesterday, but logged on to Coral to see them offering a special price of 3/1 on just these three results, which I stuck all £30 on. So my post joy arithmetic actually reveals the win to be 90 but fuck it.

Minus £15.56 from yesterday's games, so a net profit of £74.44.

[Reg]

It's more work, but we'd be better off picking a selection of fixtures each day/week and discussing them, out in the open, to find the ones that enough of us think contain value, then betting on them.
Yev, what do you think about doing this after we've predicted this weekend's Premier League results?

Other opinions welcome.

I'll ask for the predictions after we've decided what to do for this weekend.

[Reg]

In fact, we've got less than two days till the lunchtime Premier League kickoff. So we better get the predictions in. (Feel free to discuss how best to do this though!)


Round 2 Fixtures

Remember '1' for home win, '2' for draw, '3' for away win. Please copy and paste the list and put a number beside each fixture, then PM me your predictions.

Crystal Palace v West Brom
Aston Villa v Stoke
Bournemouth v Watford
Man City v Newcastle
Norwich v Leicester
Sunderland v West Ham
Chelsea v Southampton
Everton v Liverpool
Arsenal v Man Utd
Swansea v Tottenham

[Reg]

Only 1 PM sent so far lads, read Jimmy's winnings post and be inspired!

[Reg]

Wouldn't like to see this die after only 1 round, so here's another nudge. Only 3 PMs, cheers to those who've sent theirs.

I'll have to post the results when I get in tonight around 10:30/11:00pm, as I won't have much time tomorrow.

[Foe]

These needs to be an individual thread per set of fixtures I think, otherwise people won't think to check back in.

[Boydy]

Sent.

[Reg]

Round 2 Results

https://docs.google.com/spreadsheets/d/1Gdtrv1UtGySltDTCcMjUUv9EmzP7BsUTYlGJLN2EE-M/edit?usp=sharing

Using the consensus categorising from before (between 1.67 and 2.33 is a draw), bold for a winner and neither team bolded for a draw:

Crystal Palace v West Brom 1.375
Aston Villa v Stoke 2.375
Bournemouth v Watford 2.375
Man City v Newcastle 1.125
Norwich v Leicester 2.25
Sunderland v West Ham 2.75
Chelsea v Southampton 1.5
Everton v Liverpool 1.375
Arsenal v Man Utd 2.125
Swansea v Tottenham 1.625

[Sir Andy Mahowry]

Surprised that I was the only one to go for Bournemouth.

Most surprising is how many people picked Chelsea despite their shocking form.

[Yevrah]

Arsenal Man Utd shitting out a draw/0-0 has got to be worth a punt on Sunday.

[Reg]

Thanks for the PMs lads. We had 8 this time.

Good idea, Foe. I'll do a new thread for each round from here on.

[Reg]

Double:
Norwich v Leicester draw (3.6)
Villa v Stoke win (3.1)
£2 returns £22.32

[Boydy]

£2 on a six-fold of today's matches except Palace West Brom as that's already started.

Returns £436. Come on TTH, make me some money.

[Reg]

Quite like the look of the odds on today's games.

Everton @3
Arsenal-Man Utd draw @3.6
£1 on the double, returns £10.35

£5 on Everton single.

[Boydy]

Can't bring myself to be on Everton.

How did we do yesterday?

[Boydy]

Crystal Palace v West Brom 1.375 - Yep.
Aston Villa v Stoke 2.375 - Yep.
Bournemouth v Watford 2.375 - Nope. Draw.
Man City v Newcastle 1.125 - Yep.
Norwich v Leicester 2.25 - Nope. Leicester won.
Sunderland v West Ham 2.75 - Nope. Draw.
Chelsea v Southampton 1.5 - Nope. Southampton won.

3 out of 7.

[Reg]

Crystal Palace v West Brom 1.375 - Correct
Aston Villa v Stoke 2.375 - Correct*
Bournemouth v Watford 2.375 - Incorrect*
Man City v Newcastle 1.125 - Correct
Norwich v Leicester 2.25 - Incorrect*
Sunderland v West Ham 2.75 - Incorrect
Chelsea v Southampton 1.5 - Incorrect

*=narrow consensus

[Baz]

Thanks for the PMs lads. We had 8 this time.

Good idea, Foe. I'll do a new thread for each round from here on.Yeah, I've forgot about it twice. Sorry.

[Reg]

No worries, it'll be interesting to see how we do with 20 or 25 predictions.

[Demerit]

£10 on the treble today and £5 on each single.

Barkley AGS £5 @ 3.6
Naismith FGS 2-1 £5 @ 41
Naismith FGS £4 @ 7.5
Goal scored 3:00-3:59 £4 @ 26
Corner awarded 1:00-1:59 £2 @ 7
Everton BTTS £5 @ 5.5

Everton, Draw, Swansea £10 @ 33.42

Swansea £5 @ 2.87
Draw £5 @ 3.75
Everton £5 @ 3.1

[John Arne]

As a paid up member of Team Spreadsheet, I'll happily create the template/tracker malarkey. Just need to know the specifics.

[Reg]

Thanks, John Arne. Do you mean each user's success rates? I think that'd be useful, but not sure if you'd want to put the work in if it turns out this thing doesn't have legs... (Hope it does.)

[Boydy]

Since when were you in team spreadsheet?

[John Arne]

Since when were you in team spreadsheet?

I stand above Team Spreadsheet, you little shit.

Check out the excel thread on the old forum and bow down to my excel greatness.

[Benny]

You join the cool gang, with GS and Ital.

[John Arne]

Whoopee!